Rogo
面向 Wall Street 的垂直 AI 正在融资和采用上快速复利,但在当前年度 ARR、毛利率或留存未披露的情况下,$2B 估值缺少定价支撑
Rogo 是面向 Wall Street 的垂直 AI 平台,增长很快、融资充足,也有具体机构 traction;但其 $2 billion Series D 价格已经跑在所有公开收入、利润率和留存披露前面,因此应继续研究,而不是给出方向性买入或回避。
封面要素
公司概况
Rogo 是一家总部位于纽约的 AI 公司,构建面向投资银行、私募股权机构和资产管理公司的企业研究与工作流自动化平台,核心是自主智能体 Felix。公司由 Gabriel Stengel(CEO)、John Willett 和 Tumas Rackaitis(CTO)联合创立,三人都是 Princeton 同学;公司从 2024 年种子轮起步,到 2026 年 4 月以据报道 $2 billion 估值完成 $160 million D 轮,自 2022 年成立以来(Rogo 自己的早期材料和一份竞品分析写作 2021 年),融资和用户采用大约每隔几个月就复利推进。
- 官网
- rogo.ai
- 创始人
- Gabriel Stengel, John Willett, Tumas Rackaitis
- 创立地点
- New York, NY (Manhattan)
- 总部
- New York, NY, United States
- 产品
- Felix 是随 2026 年 4 月 D 轮一起推出的自主 AI 智能体,靠微调的金融推理模型,接入客户公司的内部系统(SharePoint、CRM、数据室)和授权外部数据(FactSet、LSEG、S&P Global、PitchBook),执行交易筛选、CIM 生成、买方触达、数据室尽调等多步骤金融工作流。Rogo Research 或 Rogo Comps 等具名产品线无法用公司当前材料交叉证实;Felix 加授权数据连接器,是公开资料能验证的产品结构。
- 客户
- 投资银行、私募股权机构和资产管理公司;具名机构客户包括 Rothschild & Co、Jefferies、Lazard、Moelis、Nomura、J.P. Morgan、Bank of America、Wells Fargo 和 Baird。
- 商业模式
- 面向金融机构直销的企业 SaaS 式许可,叠加智能体插件(Felix);没有来源披露 Rogo 自己的按席位或按合同定价,但可比的 AI 原生研究平台(AlphaSense、Glean)价格落在每席位 / 年度承诺 $10,000-$60,000+ 区间。
- 阶段
- late-stage private (Series D)
- 融资情况
- $160 million D 轮由 Kleiner Perkins 领投,于 2026 年 4 月 29 日完成,据报道估值 $2 billion,较三个月前 C 轮的 $750 million 标记接近翻三倍;从种子轮到 D 轮累计融资超过 $300 million,Sequoia、Thrive Capital、Khosla Ventures、J.P. Morgan Growth Equity Partners 及其他持续跟投的投资人参与。
执行摘要
主要优势
- 具体且量化的机构 traction:35,000+ 名专业人士、250+ 家机构;已发布的 Baird Equity Research 案例显示 85% 周活和 70% 日活,对这个阶段的公司来说是少见强证据。
- 高确信融资财团:同一批核心投资人(AlleyCorp、Khosla、Thrive、Sequoia)连续参与每一轮定价融资,Series D 又加入 Kleiner Perkins,估值三个月内从 $750M 近三倍升至 $2B。
- 工作流集成和授权数据护城河(SharePoint/CRM/data-room 集成,加上 FactSet/LSEG/S&P/PitchBook connectors);独立技术报道认为,这比底层模型本身更可防守。
主要风险
- 没有来源披露 Rogo 自身当年 ARR、毛利率、净收入留存或现金头寸,$2B 估值无法按公司具体经济性对标,只能靠第三方估计和可比公司倍数。
- 推导级准确率缺口:专家编写的基准 BigFinanceBench 显示,表现最佳的 AI agent 在逐步核对金融推导时只拿到 58.8% rubric points,意味着面向客户的 Felix 输出可能看起来完整,却嵌入未被发现的错误。
- 监管边界快速收紧(EU AI Act 自 August 2026 执法、FINRA/SEC/OCC 审查金融 GenAI),且没有独立审计佐证 Rogo 自报的 SOC 2 / ISO 27001 / ISO 42001 / EU AI Act 合规主张。
- 既有厂商(Microsoft Copilot for Finance、Bloomberg 的 agentic Terminal、S&P Global/Kensho)正在把竞争或相邻 agentic 能力塞进已有分发和数据产品,争夺同一工作流层。
- 独立客户验证很薄:相对 250+ 声称客户,所有抓取来源中只有约八家机构被点名或引用,另有一条匿名但具体的流失报告。
未决问题
- 公司确认的当年 ARR、收入运行率、毛利率和现金头寸 / 烧钱速度仍未披露——这是独立承销 $2B 价格的最大阻碍。
- 没有独立审计方、证书注册库或监管机构确认 Rogo 自报的 EU AI Act、SOC 2 或 ISO 27001/42001 合规主张。
- Subset、Offset 和 Plux 收购中的创始人留任、earn-out 和 vesting 条款未披露,近期买入技术能力的执行风险无法量化。
- 数据供应商合同条款(LSEG、PitchBook、FactSet/S&P entitlements)——独家性、定价、续约日期——未公开;LSEG 自身也在推出竞争性研究 agent。
- 当前员工数仅部分披露(Forbes 称约 100 人,Bloomberg 称管理层目标是到 2026 年底约 300 人),且所有已查来源都没有出现具名 CFO、CCO 或 CRO。
目录
01公司概览
1.1 身份、产品与阶段
Rogo 是一家位于纽约州纽约市的 AI 公司,搭建专为金融服务设计的企业平台,卖给投资银行、私募股权机构和资产管理公司,而不是面向所有知识工作者的横向工具。公司材料以及 Forbes 和多篇 2026 年融资报道都这样描述:平台把微调金融推理模型,与客户公司内部系统(SharePoint、CRM、数据室)和授权外部数据(FactSet、LSEG、S&P Global、PitchBook)深度整合,自动化研究、财务建模、可比公司分析和路演材料制作。截至 2026-07-01 运行日,Rogo 明确属于风险投资支持的后期私营公司:已经融到 D 轮,据报估值 $2 billion,用户覆盖 250 多家机构的 35,000 多名专业人士。 公司的成立日期,是记录中少数真正有争议的事实之一。Forbes 2026 年公司简介和 New York Weekly 的一篇特写都把 Rogo 的成立时间写成 2022 年,但 Rogo 自己 2024 年 10 月的 A 轮新闻稿,以及 Hebbia 2026 年的一份竞品分析,都称 Rogo「成立于 2021 年」。Bloomberg 对 D 轮的报道有助于调和这个差异:它写道,联合创始人 Gabriel Stengel 在「2021 年底」离开 Lazard,开始和联合创始人在 Manhattan 的厨房桌旁搭产品,正式成立公司则发生在 2022 年。后续章节应以 2022 年作为主要参考年份,同时注明来源之间的一年差异。 Rogo 的核心产品架构围绕 Felix 展开。Felix 是随 2026 年 4 月 D 轮推出的自主 AI 智能体,用很少的人为介入执行多步骤金融工作流——交易筛选、CIM 生成、买方触达、数据室尽调。尽调材料提到的「Rogo Research」「Rogo Comps」「Rogo Models」或「Rogo Pitchbook」等具名产品线,无法与 Rogo 当前网站或任何已抓取的 2026 年报道交叉证实;公司实际公开的产品结构是 Felix 加一组授权数据连接器,这一缺口保留为开放问题,而不是靠假设补上。[CO001, CO002, CO012, CO013, CO014, CO015]
| 指标 | 数值 / 状态 | 日期 / 版本 | 置信度 | 尽调缺口 |
|---|---|---|---|---|
| 成立 | Forbes 称为 2022 年;独立来源称 2021–2022 年 | 2022 | 中 | 用公司备案核对准确注册日期 |
| 总部 | 美国纽约州纽约市 | 2026-04 | 高 | |
| 最新轮次 | Kleiner Perkins 领投 $160M Series D | 2026-04-29 | 高 | |
| 最新报告估值 | 投后估值 $2B(独立报道) | 2026-04-29 | 中 | 对照已签股本结构表确认准确投后估值 |
| 累计融资 | 从种子轮到 Series D 累计超过 $300M | 2026-04-29 | 高 | |
| 用户 / 机构 | 250+ 家机构的 35,000+ 名专业人士 | 2026-04-29 | 高 | |
| 收入 | Forbes 称约 $2M(2024 年)至 $15M+(2025 年) | 2025 | 中 | 索取 2026 年当前 ARR 和毛利率 |
| 员工数 | 约 100 名员工(Forbes);管理层预计 2026 年底约 300 人 | 2026-04 | 中 | 确认当前员工数和招聘计划 |
| 旗舰产品 | Felix 自主 AI agent,加数据伙伴连接器(OpenAI、FactSet、PitchBook、LSEG) | 2026-04 | 中 | |
| 监管姿态 | 在 2026 年 8 月可执行期限前宣布 EU AI Act 合规准备就绪 | 2026-03-19 | 高 |
成立年份、估值、收入和员工数单元格反映 2024–2026 年抓取到的官方和独立来源中的最新数字;没有使用 null,因为每行至少有部分公开数据点,但多个单元格仍明确标出需要核对的缺口。
[CO002, CO012, CO013, CO037, CO038, CO039]Rogo 的产品逻辑是把公司内部数据和授权外部数据接入微调模型,再由驻场部署的银行家调校工作流,最后产出可审计、带引用的交付物。
[CO001, CO021, CO023, CO024, CO055, CO057]1.2 创始人、领导层与治理
Rogo 由 Gabriel Stengel(CEO)、John Willett 和 Tumas Rackaitis 联合创立,三人都在 Princeton University 读书时相识。Stengel 此前曾在 Lazard 担任投资银行分析师,多方来源都把这段经历描述为 Rogo 的直接起因:作为初级银行家,他凌晨 2 点还在 Excel 和 PowerPoint 里重复做分析,挫败感推动了公司诞生。Willett 此前在 J.P. Morgan Chase 和 Barclays 工作,现在从随 C 轮开设的 London 办公室负责 Rogo 的欧洲扩张。Rackaitis 拥有 Oberlin College 计算机科学学位,担任首席技术官,也是公司公开列出的技术架构负责人。 三位创始人之外,Bloomberg 报道确认 Rahul Rekhi 是 Rogo 的总裁。他在美国财政部工作约一年、在 Lazard 工作七年后加入,为管理层补上政策和资深投行信誉。公开信息看不到的,是更完整的高管或治理结构:公司、投资人或媒体材料中都没有出现独立董事名单、委员会架构,或具名的 CFO/COO 职位。因此,以据报 $2 billion 估值来看,Rogo 仍显得围绕创始人与总裁运转,关键人集中风险真实存在。 New York Weekly 的 D 轮报道还露出另一个治理信号:J.P. Morgan Growth Equity Partners 既是 Rogo 的持续投资人(B 轮到 D 轮),又通过母行 J.P. Morgan 成为 Rogo 平台公开提到的机构客户。这种投资人兼客户的双重关系本身并非不当,但自然引出尽调问题:信息隔离墙如何设置,是否存在优先访问权。后续章节或现场尽调应直接追问。[CO003, CO004, CO005, CO006, CO007, CO008]
| 人物 / 职务 | 公开来源支持的背景 | 创始人-市场匹配 / 职能覆盖 | 关键人物依赖 |
|---|---|---|---|
| Gabriel Stengel / CEO 兼联合创始人 | 曾任 Lazard 投行分析师;Princeton 计算机科学毕业 | 直接亲历 Rogo 自动化的初级银行从业人员工作流痛点;主要公开发言人 | 很高——几乎所有融资和媒体叙事都流经 Stengel |
| John Willett / 联合创始人,负责欧洲 | 前 J.P. Morgan Chase 和 Barclays 银行从业人员;Princeton 毕业 | 用银行运营深度平衡 Stengel 的产品 / 愿景角色;现负责伦敦办公室 | 高——Rogo 欧洲扩张唯一具名负责人 |
| Tumas Rackaitis / 联合创始人兼 CTO | Oberlin College 计算机科学学位 | 负责 Rogo 微调金融推理模型的技术架构 | 高——公开材料中唯一具名技术联合创始人 |
| Rahul Rekhi / 总裁 | 在 U.S. Treasury Department 约一年;在 Lazard 七年 | 在三位创始人之外增加政策 / 监管和资深银行可信度 | 中——近期新增层级,但权限范围未公开详述 |
这只是部分列举,限于 2024–2026 年抓取的公司和独立来源中点名的领导层人物;Rogo 尚未发布完整董事会名单或独立董事阵容。
[CO003, CO004, CO005, CO006, CO007, CO008]1.3 融资历史、估值与利益相关方
Rogo 的融资历史又短又陡。Citybiz 和 Sacra 都报道,Rogo 在 2024 年 2 月完成 $7 million 种子轮,由 AlleyCorp 领投,Company Ventures、BoxGroup 和 ScOp Ventures 参投;SixThirty Ventures 另行估计,该种子轮的投后估值约为 $48 million。2024 年 10 月 1 日,Rogo 自己发布新闻稿,宣布完成 $18.5 million A 轮,由 Khosla Ventures 领投,投后估值 $80 million,Mantis VC、Jack Altman 和前 Google CEO Eric Schmidt 参投,Khosla 普通合伙人 Keith Rabois 加入董事会;累计融资达到 $26 million。约在 2025 年 4 月,Rogo 完成 $50 million B 轮,由 Thrive Capital 领投,J.P. Morgan Growth Equity Partners、Tiger Global 和 Positive Sum Ventures 加入,累计融资升至 $75 million,Forbes 和 Sacra 都把估值放在接近 $350 million 的水平。 随后节奏明显加速。2026 年 1 月,Rogo 宣布完成 $75 million C 轮,由 Sequoia Capital 领投,个人投资人 Henry Kravis(KKR 联合创始人)和 Wells Fargo 参与,使累计融资超过 $165 million,估值达到 $750 million——不到一年就较 B 轮标记翻了一倍多——并为 Rogo 在 London 开设首个国际办公室提供资金。仅三个月后,2026 年 4 月 29 日,Rogo 宣布完成 $160 million D 轮,由 Kleiner Perkins 领投,Sequoia、Thrive Capital、Khosla Ventures、J.P. Morgan Growth Equity Partners、BoxGroup、Mantis VC、Jack Altman、Evantic 和 Positive Sum 加入。D 轮把累计融资推高到 $300 million 以上;Bloomberg 和 TBPN Digest 的独立报道称,Rogo 该轮估值为 $2 billion——三个月里几乎翻三倍。 投资人名单更像一个集中且高确信度的财团,而不是广撒网:AlleyCorp、Khosla、Thrive、Sequoia 这四家机构领投方加上 Kleiner Perkins 覆盖了每一轮定价融资,J.P. Morgan Growth Equity Partners 和 Henry Kravis 则显著地把风险资本世界与传统金融机构连接起来。SixThirty Ventures 在 A 轮同期评论里提出的问题,到了 D 轮价格仍然成立:整个 AI 分析师赛道(包括 Rogo)的估值,相对已展示的牵引力,是否已经按离群级收入倍数定价。[CO028, CO029, CO030, CO031, CO032, CO033]
| 利益相关方 | 角色 / 关系 | 重要性 | 尽调要求 |
|---|---|---|---|
| Kleiner Perkins | Series D 领投方(2026 年 4 月) | 以报道中的 $2B 估值为 Rogo 定价,并将 Rogo 定义为品类定义型操作系统 | 确认董事 / 观察员权利和投后持股比例 |
| Sequoia Capital | Series C 领投方(2026 年 1 月);Series D 参投方 | 设定 $750M Series C 估值基准,并在 Series D 加码 | 审阅 Series C 到 D 的估值跃升条款 |
| Thrive Capital | Series B 领投方(2025 年);持续参投至 Series D | 锚定中期扩张轮,并一路跟到后续轮次 | 确认按比例认购权和后续跟投承诺 |
| Khosla Ventures | Series A 领投方(2024 年 10 月);重复参投方 | Seed 后最早机构领投方;Keith Rabois 任 Rogo 董事 | 明确董事会治理权和投票控制 |
| AlleyCorp | Seed 领投方(2024 年 2 月) | Rogo 首笔机构资本 | 确认种子轮到 Series D 的持股稀释 |
| J.P. Morgan Growth Equity Partners(投资方) | Series B / C / D 投资者,并通过 J.P. Morgan 银行成为被引用机构客户 | 投资者-客户双重关系是治理和利益冲突尽调红旗 | 明确 J.P. Morgan 投资部门与银行部门之间的信息隔离墙 |
| Henry Kravis(KKR 联合创始人) | Series C 个人投资者 | 带来不同于风投基础的私募股权行业可信度 | 确认 Kravis 个人投资条款与任何 KKR 机构关系的区别 |
| 企业客户群(Rothschild & Co、Jefferies、Lazard、Moelis、Nomura 等) | 250+ 家具名和未具名机构客户 | 证明大型银行真实部署,而不只是试点使用 | 要求现场客户访谈和合同 / 续约条款 |
代表性地图来自抓取来源中披露的融资轮投资者和具名客户;它不是完整股本结构表。
[CO028, CO030, CO031, CO032, CO034, CO037]1.4 客户、规模与牵引力
Rogo 的规模指标显示企业采用又快又广。到 2026 年 4 月 D 轮公告时,超过 250 家机构的 35,000 多名金融专业人士每天使用该平台,高于 2026 年 1 月 C 轮公告中提到的约 25,000 名专业人士,也高于 OpenAI 2024 年合作伙伴案例研究中提到的 5,000 名银行家;这条轨迹与 OpenAI 页面归因于 2024 年产品的 27x ARR 增长相符。Rogo 官网展示了 Truist Securities CEO、Nomura 国际投行业务负责人和 Baird Global Investment Banking COO 的具名背书;多个来源列出的机构客户包括 Rothschild & Co、Jefferies、Lazard、Moelis、Nomura、J.P. Morgan、Bank of America、Wells Fargo 和 Singapore 的 GIC 主权财富基金。 最具体的客户级证据,是 Rogo 自己发布的 Baird Equity Research 部门案例研究:超过 100 名专业人士活跃使用平台,约 85% 每周活跃、70% 每日活跃。Rogo 把这种参与度定义为真正的工作流依赖,而不是尝鲜式采用。Forbes 的公司简介又为规模图景补上财务维度:收入从 2024 年约 $2 million 增至 2025 年超过 $15 million,同时估计截至 2026 年 4 月员工约 100 人;Bloomberg 另称,Rogo 管理层预计到 2026 年底员工接近 300 人。 即便如此,本年度财务图景仍然明显不完整:已抓取来源没有披露 2026 ARR、毛利率、净收入留存或盈利能力,导致据报 $2 billion 估值与外部尽调可获得的投资判断细节之间存在真实缺口。PeerSpot 2026 年中竞争心智份额数据把 Rogo 放在 2.3%(FactSet 为 18.7%),这个交叉检查虽不完整但有用:即便 Rogo 份额快速增长,它的规模主张仍处在一个由传统龙头主导的市场里。[CO016, CO017, CO018, CO019, CO020, CO041]
公开证据显示规模和估值快速复合增长,但相对于 $2B 估值,当年收入、利润率和准确员工数仍未披露。
[CO039, CO038, CO016, CO041, CO042, CO045]1.5 里程碑、监管姿态与负面信号
Rogo 自 2022 年成立后的里程碑显示,公司大约每隔几个月就在融资轮次、产品和合作发布之间交替推进。2024 年 2 月种子轮和 2024 年 10 月 A 轮之后,Rogo 在 2025 年收购 Subset,以补入电子表格智能体的 Excel 建模技术,随后在 2025 年 4 月完成 $50 million B 轮。2026 年节奏进一步加快:1 月完成 $75 million C 轮并开设新的 London 办公室,3 月收购由 Raj Khare 和 Shiv Shrivastava 创立的 AI 智能体初创公司 Offset,同月稍晚宣布 EU AI Act 合规工作,赶在该法案 2026 年 8 月可执行期限之前,4 月则完成 $160 million D 轮并公开推出 Felix。此后 Rogo 继续叠加数据和分发合作——Daloopa(5 月)、深化 PitchBook 集成(5 月)、SS&C Intralinks(6 月),并成为 Microsoft Copilot in Excel 的首发合作伙伴(6 月)——把触达延伸到相邻的金融数据和生产力生态。 监管与信任层面,Rogo 表示不会用客户数据训练或更新模型,并维护 SOC 2、ISO 27001、ISO 42001 和与 GDPR 对齐的文档,2026 年 3 月的 EU AI Act 合规工作又进一步正式化这些承诺。对一家卖给全球监管最重机构之一的公司来说,这些承诺有意义,也可以验证。 公开记录中最清晰的负面信号不是诉讼或监管行动,而是 D 轮报道本身突出带出的结构性质疑。Bloomberg 报道称,一些 AI 行业观察者认为 Rogo 是不必要的中间层,因为金融专业人士原则上可以直接查询大型通用 AI 模型;这是一个关于竞争护城河的问题,也被 SixThirty Ventures 的分析师评论独立呼应,后者同样质疑 AI 分析师估值整体是否反映了离群级收入倍数。另一个来源是竞争对手自己的 2026 年竞品指南(该来源显然有竞争动机,但提出了具体、可核查的主张),它指出了具体技术限制:Rogo 难以把分析扩展到数千份文档,引用粒度停留在回答级而非句子级,团队协作工具也有限。最后,Bloomberg 描述了初级银行家的真实焦虑:Rogo 式自动化可能减少入门级招聘。Rogo 创始人的回应是,技术会让银行增加更多资深交易人员,而不是减少初级人员;但这项主张仍未用银行实际招聘数据验证。[CO025, CO026, CO050, CO051, CO052, CO053]
| 日期 | 事件 | 类型 | 金额 / 估值 / 状态 | 参与方 | 含义 |
|---|---|---|---|---|---|
| 2022 | Rogo 成立(据 Forbes);Bloomberg 称创始人于 2021 年末开始搭建 | 成立 | Gabriel Stengel、John Willett 与 Tumas Rackaitis | 启动公司的公开运营时钟,不同来源之间存在一年报告差异 | |
| 2024-02 | Seed 轮 | 融资 | AlleyCorp 领投 $7M | AlleyCorp、Company Ventures、BoxGroup 与 ScOp Ventures | 隐身模式产品开发后获得首笔机构资本 |
| 2024-10-01 | Series A | 融资 | Khosla Ventures 领投 $18.5M,估值 $80M | Khosla Ventures、Mantis VC、Jack Altman 与 Eric Schmidt | Keith Rabois 加入董事会;累计融资达到 $26M |
| 2025 | 收购 Subset | 产品 | Rogo;Subset | 增加电子表格 agent 技术,用于审计复杂 Excel 模型并向前滚动 | |
| 2025-04 | Series B | 融资 | Thrive Capital 领投 $50M,估值约 $350M | Thrive Capital、J.P. Morgan Growth Equity Partners、Tiger Global 与 Positive Sum | 累计融资达到 $75M;JPMorgan 同时以投资者和客户身份进入 |
| 2026-01-28 | Series C 与伦敦办公室开设 | 扩张 | Sequoia Capital 领投 $75M,估值 $750M | Sequoia Capital、Henry Kravis 与 Wells Fargo | 不到一年内估值较 Series B 翻倍以上;开启欧洲扩张 |
| 2026-03-13 | 收购 Offset | 产品 | Rogo 与 Offset(Raj Khare、Shiv Shrivastava) | 嵌入学习型 agents,直接在工作流中维护和更新金融模型 | |
| 2026-03-19 | 宣布 EU AI Act 合规准备就绪 | 监管 | Rogo;外部审计师 | 使 Rogo 先于该法案 2026 年 8 月欧洲客户可执行期限完成定位 | |
| 2026-04-29 | Series D 与 Felix agent 发布 | 融资 | Kleiner Perkins 领投 $160M,报道估值 $2B | Kleiner Perkins;Sequoia;Thrive Capital;Khosla Ventures;J.P. Morgan Growth Equity Partners;其他 | 累计融资超过 $300M;三个月内估值接近翻三倍 |
| 2026-04-29 | 围绕估值和初级银行从业人员替代的公开质疑随 Series D 报道浮现 | 反向 | 独立 AI 行业评论者;初级银行从业人员(据 Bloomberg) | 在融资里程碑旁边凸显估值溢价和劳动力市场风险 | |
| 2026-05-20 | 宣布 Daloopa 合作 | 合作 | Rogo;Daloopa | 扩展 Rogo 的数据摄取伙伴生态 | |
| 2026-05-27 | 深化 PitchBook 数据合作 | 合作 | Rogo;PitchBook | 在 Rogo 内原生呈现 PitchBook 的交易、基金、公司和投资者数据 | |
| 2026-06-15 | 宣布 SS&C Intralinks 集成 | 合作 | Rogo;SS&C Intralinks | 将 Rogo 延伸到数据室和尽调工作流 | |
| 2026-06-26 | Microsoft Copilot in Excel 首发伙伴 | 合作 | Rogo;Microsoft | 为共同客户把 Rogo 银行和研究工作流直接嵌入 Excel |
供后续章节引用的标准按日期编排时间线;2022 年创立日期这一行记录的是跨来源真实分歧(2021 年 vs. 2022 年),而不是武断给出结论。
[CO012, CO013, CO014, CO028, CO025, CO030]Rogo 把从种子轮到 $2B 估值的增长压缩到大约两年机构融资里,产品和合作发布与每一轮融资同步推进。
[CO012, CO028, CO030, CO025, CO032, CO034]1.6 图表与证据
02市场分析
2.1 市场定义与边界
Rogo 竞争的是专为投资银行、私募股权机构和资产管理公司设计的 AI 研究与工作流自动化市场,比「金融服务 AI」或通用 AI 办公生产力软件更窄。第三方追踪机构把该产品描述为面向银行、PE 和对冲基金用户的垂直研究助手,整合内部与外部金融数据源,区别于通用大语言模型聊天助手(CM042、CM021)。这个品类要自动化的服务收入底座很大:2024 年全球企业与投资银行收入达到 $3.0 trillion(CM004),全球股票研究行业即使在美国卖方分析师人数自 2015 年以来约下降 18% 后,每年仍产生约 $8.7 billion 收入(CM017、CM018)。该品类夹在两类截然不同的替代品之间。一边是传统金融数据终端——Bloomberg(约 33% 份额,约 $27,660/席位/年)、Refinitiv Eikon(约 20% 份额)、Capital IQ 和 FactSet($12,000/席位/年)——它们把数据交付与越来越强的分析层打包,代表根深蒂固的现状(CM014、CM015、CM016)。另一边是 Microsoft 365 Copilot 等通用 AI 办公助手,开箱并不懂金融,但会争夺同一条企业 AI 预算线,同时缺少核心品类所需的授权数据集成和交易专属工作流(CM041)。法律和咨询导向的 AI 研究工具被明确排除:它们服务的是不同买方(律所、咨询公司)和不同工作流,尽管底层模型技术相似。[CM042, CM021, CM004, CM017, CM018, CM014]
| 细分 / 类别 | 纳入支出 | 排除支出 | 买方 / 付款方 | 对 Rogo 的意义 |
|---|---|---|---|---|
| 面向 IB/PE/AM 的 AI 研究与工作流自动化软件(核心类别) | 面向银行、PE 机构、对冲基金和资管公司销售的智能体式 AI 工具,用于拉取可比公司数据、综合尽调、起草推介书、自动化建模 | 通用消费者 AI 聊天;非金融垂直行业 | 交易 / 客户覆盖团队(买方);CTO/COO 或交易团队预算(付款方) | Rogo 竞争的核心可服务类别 |
| 金融数据终端与分析师工作站(Bloomberg、FactSet、Capital IQ、Refinitiv) | 市场 / 公司数据交付、筛选,以及部分嵌入式 AI 分析层 | 跨文档和工作流的完整智能体式任务执行 | 同一批买方,且常常是同一条技术预算线 | 根深蒂固的现有替代品,也带来捆绑风险 |
| 通用 AI 办公 Copilot(Microsoft 365 Copilot、通用 LLM 助手) | 跨行业通用生产力、起草和摘要 | 金融专用数据集成、授权市场数据、交易专用工作流 | 企业 IT / Microsoft 365 预算 | 相邻替代品,争夺同一条 AI 预算线 |
| 法律与咨询 AI 研究工具(如 Harvey 及同类) | 面向律所和咨询公司的 AI 研究自动化 | 投行、PE 与资管专用工作流 | 法务运营或咨询公司预算 | 排除:买方不同,工作流不同 |
| 传统卖方股票研究生产 | 银行分析师撰写并分发给客户的研究报告 | AI 原生、由智能体运行的研究自动化 | 研究部门预算,通常打包在交易佣金里 | 相邻且萎缩的类别,AI 工具会替代其中一部分 |
| 面向对冲基金 / 资管公司的另类数据与分析平台 | 投资决策所需的数据获取与分析工具 | 银行 / PE 交易工作流自动化(用例不同) | CIO 或投资组合团队预算 | 买方重叠相邻,集中在资管细分 |
纳入 / 排除支出的边界,是作者参考所引分析师和竞争对手来源搭建的类别框架;相邻类别的金额在规模测算视角表中另行引用。
[CM041, CM016, CM042, CM014, CM015, CM018]2.2 市场规模:TAM、SAM、SOM 与相互冲突的测算视角
没有一个单一且无争议的数字能描述这个市场,本章保留这种分歧,而不是强行制造虚假的共识。Precedence Research 将全球金融服务生成式 AI 市场估为 2026 年 $2.51 billion,到 2035 年升至 $17.88 billion,CAGR 为 24.81%(CM001、CM002)。The Business Research Company 划出更窄的「银行与金融」品类:2025 年 $1.75 billion,到 2030 年增至 $7.71 billion,CAGR 为 34.5%(CM003)——这个增长率和基准年份都明显不同,本章把它标为明确冲突(CM003 与 CM001/CM002 矛盾),而不是调和成一个数字。买方侧锚点提供了更多背景,但也不能消除缺口:2026 年全球私募股权 AUM 约 $8 trillion,最大平台 Blackstone 单独管理 $1.3 trillion(CM007、CM008);更广义资产管理行业管理 $147 trillion,其 2025 年收入增长超过 80% 来自市场升值而非净新资金流入(CM009)。已审阅的分析师报告没有一份单独隔离出 IB/PE/AM AI 研究自动化软件的可服务可触达市场——所有已发布数字都把这个窄品类与更广的金融服务生成式 AI 支出或传统数据终端收入混在一起(CM043)。因此,Rogo 可实际拿到的收入(SOM)无法用公开数据计算;它需要公司披露尚未公开的 ARR、席位数和按席位定价。本报告把这记录为尽调缺口,而不是靠假设估算。[CM001, CM002, CM003, CM007, CM008, CM009]
| 视角 | 定义 | 数值 / 估算 | 依据 / 方法 | 置信度 |
|---|---|---|---|---|
| 广义金融服务生成式 AI TAM(2026) | 银行、保险、资管和资本市场中的全部生成式 AI 软件收入 | 2026 年 $2.51B,到 2035 年升至 $17.88B | Precedence Research 自下而上模型;2026-2035 年 CAGR 为 24.81% | 中 |
| 较窄的银行与金融生成式 AI TAM(替代视角) | 专门限定在银行和金融领域的 AI 软件,不含保险 | 2025 年 $1.75B,到 2030 年升至 $7.71B | The Business Research Company;CAGR 为 34.5% | 中(范围和增长率与第 1 行冲突) |
| CIB 服务收入基数(背景,不是 AI 专属数字) | AI 工具正在自动化的全球公司与投行业务服务收入 | 2024 年 $3.0 trillion | McKinsey 年度 CIB 报告 | 背景参考可信度高;不是软件 TAM |
| 全球 PE AUM(买方规模锚点) | 约 100 家最大 PE 平台管理的总资产 | 2026 年约 $8 trillion | 基于 2025 年 Q4 申报文件、财报和新闻稿汇编排名 | 中 |
| 全球资管行业 AUM(买方规模锚点) | 资管行业总 AUM | 2025 年 $147 trillion | BCG Global Asset Management Report 2026(报告) | 中 |
| 面向 IB/PE/AM 专用 AI 研究自动化的 SAM(Rogo 所称类别) | 狭义软件收入:银行、PE 机构和对冲基金购买的智能体式研究 / 工作流工具 | 所审阅公开分析师报告均未单独拆出 | 没有分析师报告把这一小块从更广义的金融生成式 AI 支出或数据终端现有收入中拆出来 | 低 / 证据缺口 |
| SOM(Rogo 实际可获得收入) | Rogo 在上述未解 SAM 中可获得的份额 | 无法用公开数据计算 | 需要 Rogo 尚未披露的私有 ARR、席位数和单席价格 | 低 / 证据缺口 |
第 1-2 行采用的范围和计量口径有实质差异,因此保留为相互矛盾的公开视角,而不是合并成单一数字;第 6-7 行明确记录规模测算缺口,不按本章尽调哲学臆测数字。
[CM001, CM002, CM003, CM004, CM007, CM009]从最宽的金融服务 GenAI TAM,到尚未解决的 Rogo SAM/SOM,按已发布市场规模视角搭出金字塔。
单位为 $B。底部两层没有披露数值;纳入它们是为了可视化测算缺口,而不是暗示任何数字,也遵循保留缺口、不猜测的要求。
[CM001, CM002, CM003, CM043]低 / 基准 / 高区间对比两个分析师口径在共同 2026 年和约 2030 年时点披露或按 CAGR 推算的市场规模,单位为十亿美元。
约 2030 年低端($6.09B)按 Precedence Research 披露的 2026 年数值($2.51B)和其声明的 24.81% CAGR 复合 4 年计算;高端($7.71B)为 The Business Research Company 直接披露的 2030 年数值。两个口径均为年度软件收入,单位为十亿美元;未混用百分比或指数单位。
[CM001, CM002, CM003]2.3 买方、用户与付款方分层
可触达市场由五类买方组成,每类的预算归属和采用触发点都不同。大型综合投行和全球投行是最早、资金最充足的买方,背后有平均约占收入 9-11% 的 IT 预算,以及与 Bloomberg、FactSet、LSEG 已存在的采购关系(CM044、CM019、CM020);它们采用这类工具,主要受同行压力和初级员工成本经济性驱动。中型市场和精品投行采用更晚、价格更敏感,通常会等大型同行验证 ROI 后,才投入合伙人层级预算(CM045)。私募股权机构正在把 AI 嵌入尽调、交易生命周期自动化和被投公司运营,但工具预算在基金层面的普通合伙人手里,不在单个被投公司手里,并且越来越多地被拿来向有限合伙人证明价值创造能力(CM046、CM010)。对冲基金和资产管理公司会把 AI/另类数据工具预算与纯数据采购分开,软件和技术通常吃掉这笔独立预算的三分之一到一半(CM047);94% 受访基金经理预计 2026 年增加这项支出(CM026),60% 机构投资人表示,他们更愿意把资本分配给对生成式 AI 研究投入可观预算的基金(CM028)——这是由 LP 驱动的采用触发点,区别于单笔交易 ROI。企业发展和内部 M&A 团队则补全版图:这是一个更小、由 CFO 预算负责的细分市场,动机是减少对付费顾问费用的依赖。供应商和竞品评论印证了这个框架:Rogo 和 Hebbia 面向大型金融机构以及企业级银行 / PE 团队定位,而更小的新进入者用更精简的交易团队切入私募信贷基金和家族办公室(CM021、CM022)。[CM044, CM045, CM046, CM047, CM010, CM026]
| 细分 | 买方 | 用户 | 付款方 | 工作流 | 预算负责人 | 采用触发因素 |
|---|---|---|---|---|---|---|
| 大型全能 / 全球投行 | 交易与客户覆盖团队 MD、技术采购 | 分析师、Associate、VP | 全公司技术预算 | 推介书、可比公司分析、CIM、业绩会准备 | CTO/COO 加交易团队负责人 | 同业采用叠加初级员工成本压力 |
| 中型 / 精品投行 | 管理合伙人、交易负责人 | 小规模分析师 / Associate 团队 | 合伙人层面或公司技术预算 | 可比公司分析、尽调支持、轻量建模 | 管理合伙人 | 价格敏感;等待大型同业证明 ROI |
| 私募股权(并购 / 成长型) | 交易合伙人、投后运营负责人 | Associate、VP、被投公司运营团队 | 基金层面 GP 预算,与被投公司运营支出分开 | 尽调综合、交易生命周期自动化、投后监控 | GP 运营合伙人 / 基金 CTO | LP 要求证明 AI 赋能价值创造 |
| 对冲基金 / 主动资管机构 | 基金经理、CIO、研究负责人 | 分析师、量化研究员 | 公司研究 / 数据预算,与数据采购预算分开 | 另类数据综合、业绩会分析、研究自动化 | CIO 或研究负责人 | 机构投资者偏好 AI 赋能管理人 |
| 企业发展 / 内部 M&A 团队 | 企业发展 VP | 企业发展分析师 | 企业财务 / IT 预算 | 标的筛选、可比公司分析、内部交易备忘录 | 企业发展负责人或 CFO 组织 | CFO 要求减少对付费顾问费用的依赖 |
采用触发因素和预算归属来自对供应商 / 竞争对手定位及行业 IT 支出基准的综合,而不是单一列举式调查;把它当作方向性地图,不是覆盖每类买方的完整普查。
[CM044, CM045, CM046, CM047, CM021, CM022]基于证据的 1-10 顺序评分,覆盖五类买方在四个采用相关维度上的表现。
评分是基于证据的顺序判断,来自已引用的 IT 预算占收入比例基准、终端定价,以及 PE / 对冲基金 AI 支出调查数据;不是单一直接调研排名,应视为方向性而非精确测量。
[CM044, CM045, CM046, CM047]2.4 增长驱动因素与采用路径
几股结构性力量正在把预算拉向这个品类。最清晰的是初级银行家的成本和产能压力:Anthropic 在 2026 年推出 10 个可直接运行的金融智能体模板,覆盖路演材料、财报回顾监控、建模和可比公司检查——这些任务过去主要由初级投行人员完成(CM024);一些可靠性不一的评论称,部分银行因此把分析师入职批次最多削减了三分之二(CM025)。LP 和机构投资人的压力是第二个驱动因素,很大程度上独立于已经证明的单笔交易 ROI:58% 另类投资管理人预计前台会更广泛集成生成式 AI,高于 2023 年的 20%(CM027);McKinsey 估计 AI 与运营模式杠杆可把 CIB 盈利能力提高 20-30%,这给银行自上而下拨预算提供了理由(CM005)。采用路径很清楚:公司先评估和试点工具,再把工具嵌入 M&A 等特定工作流(截至 2025 年,86% 采用者这样做,见 CM038),最后只有少数越过试点进入可衡量 EBIT 影响(约三分之一实现规模化,只有 6% 算作 AI「高绩效者」,见 CM037)。采用与规模化价值之间的缺口,本身就是下文继续讨论的约束,也会直接影响 Rogo 这类供应商把试点席位转化为扩张收入的速度。[CM024, CM025, CM027, CM005, CM038, CM037]
| 驱动因素 / 约束 | 方向 | 时间 | 影响 | 尽调问题 |
|---|---|---|---|---|
| 初级投行员工成本和产能压力 | 驱动因素 | 现在(2026) | 银行有直接动机购买自动化,替代培训成本最高的初级员工 | 用银行披露的分析师入职班缩编情况核验 Rogo 实际席位增长 |
| LP / 投资者要求 GP 展示 AI 赋能价值创造 | 驱动因素 | 现在至 2027 年 | AI 工具预算获得一定强制性,不完全依赖已验证的单笔交易 ROI | 询问 GP 如何向 LP 披露 AI 支出,以及募资尽调是否会审查 |
| 前沿模型商品化(如 Anthropic 金融智能体模板) | 约束(压缩供应商护城河) | 现在至 2026 年 | 通用模型提供商可以复制狭窄工作流模板,挤压点解决方案定价 | 评估 Rogo 价值有多少来自专有工作流 / 集成,而不是基础模型能力 |
| 现有巨头捆绑(Bloomberg、FactSet、Microsoft Copilot) | 约束 | 现在至 2027 年 | 买方可能从现有终端或 Office 支出中获得「足够好」的 AI 功能,而不是新增采购科目 | 跟踪现有巨头 AI 功能是否在蚕食原本留给点解决方案的试点预算 |
| FINRA/SEC 对生成式 AI 的监管监督 | 约束 | 现在(2026 考试周期) | 合规成本上升;缺少成文治理控制时,采购可能变慢 | 对照 FINRA Rules 3110/4370 和 SEC 记录留存预期,确认 Rogo 合规姿态 |
| EU AI Act 执法阶段触达金融服务 | 约束 | 2026 年 8 月 | 全球运营的银行和 PE 机构要承担更多跨境合规负担 | 评估 Rogo 欧洲客户的 EU 专属治理要求 |
| AI ROI 质疑 /「试点疲劳」(MIT 95% 失败率发现) | 约束 | 现在(2026) | 买方可能收缩预算,或要求更快、更硬的 ROI 证据后才扩大席位 | 索取 Rogo 已实现的生产力 / ROI 案例数据,而不只是试点数量 |
| AI 部署人才 / 技能短缺 | 约束 | 现在至 2027 年 | 即使买方愿意买,也可能缺人运营和治理新 AI 工具,拉慢上线速度 | 询问 Rogo 销售周期有多少受买方 AI / 治理人员缺口影响 |
方向和时间基于作者对所引 2025-2026 年分析师、监管和新闻来源的综合,而不是单一排名调查;时间估算按方向性理解。
[CM024, CM028, CM029, CM016, CM032, CM040]以 100 家正在评估的公司为指数基准,展示进入采用、M&A 工作流集成、试点规模化和高绩效阶段的比例。
第 2-5 阶段把企业 / PE AI 采用者的报告比例(blott.com,引用 McKinsey 的 State of AI 2025 和 Deloitte 2025 M&A 生成式 AI 研究)套用到 100 家公司的指数基准;跨阶段复合只是近似漏斗形态,并非长期跟踪的单一队列。
[CM037, CM038]2.5 采用约束、监管摩擦与尽调缺口
三股力量限制预算转化为已签合同的速度。第一是传统厂商打包。Microsoft 在 2026 年推出金融专用 Copilot 能力,包括嵌入 Excel、Outlook 和 Teams 的 Finance Agent,面向 FP&A、会计、税务、合规和资金管理工作流(CM029、CM030),让 IT 买方在已经批准的支出里拿到一个「足够好」的打包替代品——即便 Microsoft 自己的 Copilot 付费席位渗透率估计只有约 3.3%,对应其已安装基数和 $190 billion 的 2026 年 AI 资本开支,差距已经引发投资人对传统厂商 AI 货币化的普遍怀疑(CM031)。第二是监管监督。FINRA 的 2026 Annual Regulatory Oversight Report 首次加入专门的生成式 AI 章节,点名幻觉风险、偏见、网络安全暴露,并把监督扩展到自主「智能体式」AI(CM032);SEC 的 FY2026 检查重点也提高了对 AI 治理、可解释性和营销声明中「AI-washing」的审查,要求保留 AI prompt/output 日志,作为受监管的账簿和记录(CM033)。供应商不能把这项合规负担外包给买方,也不能声称不知情(CM034);EU AI Act 最重要的执行阶段将在 2026 年 8 月触达金融服务业(CM040)。第三是需求侧怀疑。MIT 的 GenAI Divide 研究发现,按六个月 ROI 测试,企业生成式 AI 试点失败率为 95%(CM035);61% 企业领导者表示,证明 AI ROI 的压力比一年前更大(CM036);人才 / 技能短缺是 71% 企业评估 AI 但没有实施的首要原因(CM039)。这些约束没有哪一项会致命打穿 Rogo 的论点,但它们与尚未解决的 SAM/SOM 缺口(CM043)、缺少当前精确的美国 经纪交易商可触达数量一起,定义了投资人在支撑具体市场份额假设前应走的尽调路径。[CM029, CM030, CM031, CM032, CM033, CM034]
2.6 图表与证据
03竞争对手
3.1 竞争格局图:直接同业、传统厂商、相邻玩家与现状替代
Rogo 的竞争集合,比简单列出「还有谁把 AI 卖给银行家」要宽。Hebbia、F2、Marvin Labs 等直接 AI 原生同业,争夺同一笔企业交易团队预算,但产品押注有重叠也有差异:Hebbia 押注大批量文档智能,F2 押注面向承销的原生 Excel 计算,Marvin Labs 则押注更轻量的股票研究自动化。第二层是 Bloomberg、FactSet、S&P Capital IQ/Kensho 和 LSEG;AI 原生新进入者融资时,它们并没有停在原地,而是各自把原生生成式 AI 研究智能体(ASKB、FactSet AI for Banking、ChatIQ、Deep Research)直接塞进银行已经许可的终端或数据平台,把过去单纯的数据订阅变成打包 AI 竞争对手。第三层是横向 AI 助手,包括 Microsoft 365 Copilot 和 Glean,缺少金融专用微调,但会争夺同一条企业 AI 预算线,而且两者已经在大多数银行现有软件版图内。Intapp DealCloud 占据相邻且大体互补的赛道:Celeste AI 面向交易发起和关系情报,而不是研究与分析生成,所以更常与研究工具搭配,而不是替代研究工具。Harvey 是一个相邻的法律 AI 平台,正在扩展到资产管理和基金设立工作流;它更可能是未来进入者而非当前竞争对手,但其智能体基础设施和资本底座值得跟踪。最后,无论供应商竞争如何,两类现状替代仍然存在:大型银行(JPMorgan、Goldman Sachs、Morgan Stanley)自建内部 GenAI 工具,以及 SP2 Analytics 等公司把人工分析师外包明确营销为 AI 研究平台的低幻觉风险替代品。这张竞争格局图(TP001)有意定义为一个有证据支撑的代表性样本,而不是该品类每一个点解决方案的完整索引。[CP001, CP002, CP003, CP006, CP007, CP009]
| 实体 | 类别 | 主营业务 | 与 Rogo 的关系 |
|---|---|---|---|
| Hebbia | 直接 AI 原生同业 | 面向批量尽调的 Matrix 文档智能网格 | 争夺同一条企业金融 / PE / 信用交易团队预算 |
| F2(F2.ai,竞品) | 直接 AI 原生同业(相邻) | 带原生 Excel 计算引擎的智能体式 AI 承销 | 与 Hebbia 争夺私募市场承销场景;也与 Rogo 的 Excel 自动化野心重叠 |
| Marvin Labs | 直接 AI 原生同业(细分) | 基于申报文件、新闻稿和业绩会的股票研究自动化 | 规模较小的同业,瞄准单个分析师而非企业级交易团队 |
| o11 | 相邻替代品 | Excel/Word/PowerPoint 原生 AI 执行层 | 明确把自身定位为 Rogo 和 Hebbia 的「最后一公里」补充或替代 |
| Bloomberg Terminal / ASKB / BloombergGPT | 现有巨头捆绑 | 带原生对话式 AI 智能体的金融数据终端 | 如果银行的 Terminal AI 能匹配 Rogo 研究工作流,就会形成替代;多数目标账户已授权使用 |
| FactSet AI for Banking / FactSet AI(竞品) | 现有巨头捆绑 | 带 MCP 服务器和银行 AI 工作流层的金融数据平台 | 直接现有巨头回应,正面瞄准 Rogo 的投行买方 |
| S&P Capital IQ Pro / Kensho / ChatIQ(竞品) | 现有巨头捆绑 | 带生成式 AI 文档智能和聊天助手的金融数据平台 | 数据层和工作流均有重叠;ChatIQ 明确瞄准银行和买方分析师 |
| LSEG Workspace / Deep Research 智能体 | 现有巨头捆绑兼数据伙伴 | 带原生 AI 研究智能体的金融数据 / 分析平台 | 同时是 Rogo 授权数据供应商和直接研究智能体竞争对手 |
| Microsoft 365 Copilot / Copilot for Finance(竞品) | 相邻替代品(横向) | 嵌入 Excel/Teams/Outlook 的通用办公 AI Copilot | 争夺同一条企业 AI 预算线,但没有金融专用微调 |
| Glean | 相邻替代品(横向) | 带金融服务 MCP 生态的企业搜索和智能体平台 | 聚合 Rogo 所集成的同一批第三方数据提供商,但缺少银行专用工作流智能体 |
| Harvey | 可能进入者(相邻垂直) | 正在扩展到资管和基金设立的法律 AI 智能体平台 | 尚非直接竞争对手,但共享资本充足的智能体基础设施,且资管客户群有重叠 |
| Intapp DealCloud (Celeste) | 相邻捆绑 | 带智能体式 AI 的交易与关系智能 CRM | 补充 Rogo 而非替代 Rogo;瞄准交易拓展 / 关系跟踪,不是研究生成 |
| Daloopa | 相邻替代品(细分) | 从申报文件自动抽取金融数据并填入模型 | 更窄的点解决方案;CB Insights 将 Rogo 列为 Daloopa 替代品 |
| 银行内部自建生成式 AI(JPMorgan LLM Suite、Goldman GS AI Assistant) | 现状 / 内部自建 | 银行内部构建和托管的专有生成式 AI 工具 | 在资源最强的大型全能银行账户中压缩可服务市场 |
| 人类分析师外包(如 SP2 Analytics 离岸 CA/CFA/MBA 分析师) | 现状 / 替代品 | 外包人类研究分析师团队 | 明确作为 AI 研究平台的低幻觉风险替代方案销售 |
代表性而非穷尽:金融 AI 供应商版图还包括数十个较小点解决方案(如 Fiscal.ai、Quartr、Visible Alpha、Fintool、Metal),本文未逐一画像;见 enumerationScope。
[CP001, CP002, CP003, CP006, CP007, CP009]3.2 竞品画像:规模、产品范围与战略方向
在直接 AI 原生同业里,Hebbia 资本最充足:两轮累计融资 $160 million(包括 2024 年 a16z 领投的 $130 million B 轮),截至 2026 年 5 月估值约 $700 million。它的 Matrix 产品优化的是大型、混乱数据室里的单元格级、多文档尽调,而不是标准化交易工作流生成。F2 切的范围更窄——私人市场承销——但它用公开且独立验证的基准分数支撑原生 Excel 计算主张(SpreadsheetBench Verified 得分 95.25%),Rogo 和 Hebbia 都没有披露自己的对应结果。传统厂商打包层则从相反方向往同一终点移动:FactSet 首席 AI 官把公司战略定义为基于 Anthropic、Google 和 OpenAI 模型,为 9,000+ 客户构建「开放、灵活且安全的解决方案」;S&P Global 的 ChatIQ 明确「为支持银行和买方分析师需求而定制」;Bloomberg 的 ASKB 依托 363 billion token、很大程度上自有的训练语料,比多数 AI 原生金融初创公司早了多年。LSEG 的角色尤其双重:它既是授权数据伙伴,把公司基本面、估算和超过 1.5 million 笔交易的 M&A 数据库送入 Rogo 自己的平台;也是在 Workspace 内出售自有竞争产品 Deep Research 智能体的卖方。Harvey 的战略方向,是未来 12-24 个月最值得观察的一条线:它以 $11 billion 估值融资 $200 million,资金用于把「长周期智能体」扩展到基金设立,并且已经拥有超过 50 家资产管理公司客户——这是贴近 Rogo 核心投行业务流程的滩头阵地,尽管尚未真正进入核心区。[CP007, CP008, CP010, CP012, CP014, CP015]
| 公司 | 规模 / 融资(2026) | 目标细分 | 相比 Rogo 的差异化 | 关键限制 |
|---|---|---|---|---|
| Hebbia | 2 轮共融资 $160M;估值约 $700M(2026 年 5 月) | 做高风险尽调的 PE、信用、银行和资管机构 | 跨大型非结构化数据室的多文档、单元格级 Matrix 分析 | 没有原生就地 Excel 公式计算;生成新模型,而不是编辑现有模型 |
| F2(F2.ai,竞品) | 所审阅来源未披露 | 私募市场承销交易团队 | 原生、确定性 Excel 公式引擎(SpreadsheetBench Verified 得分 95.25%) | 相比 Rogo 更广的研究 / CIM / 可比公司工作流,F2 更聚焦承销数学 |
| Bloomberg (ASKB / BloombergGPT) | Bloomberg L.P. — 私营,规模未单独披露 | 覆盖所有买方 / 卖方角色的现有 Terminal 订阅用户 | 多年、大规模专有金融训练数据(363B tokens)和 800+ 提供方研究网络 | 打包在现有 Terminal 订阅中;不是独立、交易工作流原生的智能体 |
| FactSet(FactSet AI for Banking,竞品) | 9,000+ 客户;241,000+ 个人用户;NYSE/NASDAQ:FDS | 覆盖买方、卖方、财富、PE、企业的现有 FactSet 客户 | 行业首个 MCP 服务器,加上与 Finster AI 共建的专用银行 AI 工作流生态 | AI 层叠加在现有数据订阅之上;采用节奏绑定 FactSet 续约周期 |
| S&P Global(Capital IQ Pro / Kensho / ChatIQ,竞品) | NYSE:SPGI 旗下 S&P Global Market Intelligence 部门 | 现有 Capital IQ Pro 客户;银行和买方分析师 | ChatIQ 基于 Capital IQ Pro 专有表格 / 文本语料训练,来源可完整追溯 | 定位为数据平台附加层,而非独立智能体式交易工作流产品 |
| LSEG (Workspace / Deep Research) | LSEG Group;股票代码 LSEG;全球 26,000+ 员工 | 现有 Workspace 订阅用户;也是 Rogo 数据授权伙伴 | Deep Research 智能体基于 LSEG 自有可信、可审计内容集 | 竞合风险:LSEG 既向 Rogo 供应数据,也销售竞争性研究智能体 |
| Microsoft(365 Copilot / Copilot for Finance,竞品) | 15M+ 付费 Copilot 席位;席位同比增长 160%(2026) | 所有企业知识工作者,包括财务职能 | 多数银行已通过 M365 授权;Agent 365 编排平台(2026 年 5 月 GA) | 开箱不具备金融专用能力;40% 受访组织因治理顾虑将上线推迟 3+ 个月 |
| Glean | 金融服务 MCP 生态于 2026 年 6 月推出,合作方包括 CB Insights、Crunchbase、Daloopa、FactSet、S&P Global | 企业全员知识工作者,包括受监管金融团队 | 权限感知企业图谱,聚合了 Rogo 同样集成的许多数据伙伴 | 横向搜索 / 智能体平台,不是为银行家交易工作流(CIM、可比公司、备忘录)专门打造 |
| Harvey | 2026 年 3 月以 $11B 估值融资 $200M;总融资 >$1B | AmLaw 100 律所、500+ 企业内部法律团队、50+ 资管公司 | 25,000+ 自定义智能体;正把长周期智能体扩展到基金设立和资管工作流 | 尚未在核心 IB 交易工作流(CIM、可比公司、推介书)中证明;产品根基先在法律 AI |
融资、估值和规模数字来自各供应商自有材料或本章审阅的独立 2026 年报道中已公开披露的信息;未披露指标的私营公司标为未披露,而不是估算。
[CP007, CP008, CP010, CP012, CP014, CP016]按智能体式交易工作流范围(x 轴,从窄文档搜索到完整交易工作流生成)和银行 / 机构金融专注度(y 轴,从横向通用到为银行专门打造)对 Rogo 与九家竞品做顺序定位。分数是有证据支撑的顺序判断,并非厂商披露指标。
轴位置是本章对 TP001-TP003 证据的顺序综合,不是厂商发布或分析师评分的指数;相对聚类只能作方向性参考。
[CP002, CP007, CP009, CP012, CP014, CP017]3.3 能力、定价与 GTM 对比
已审阅供应商里,没有一家在企业金融买方看重的所有采购标准上都领先。智能体式交易工作流生成——CIM、可比公司分析、备忘录——Rogo 和 FactSet AI for Banking 走得最远;大批量文档智能上,Hebbia 和 Glean 的权限感知企业图谱领先;原生 Excel 计算上,F2 的已跑基准引擎和 Microsoft Copilot in Excel 是唯二有可验证主张的供应商;任何已审阅来源都没有给出 Rogo 自己来自 Subset/Offset 的 Excel 自动化基准测试。定价透明度同样分化很大:AlphaSense 发布五档结构,从每用户每年 $10,000-$15,000 到 Enterprise Intelligence 团队许可的 $50,000-$100,000+ 不等,Marvin Labs 披露 $89/月档位,Daloopa 提供免费入门层;而 Hebbia、Glean 和 Rogo 自己都没有公开按席位定价,让买方在最可能争夺大型企业合同的平台上只能盲谈。GTM 和分发方面,传统厂商的优势是结构性的,而非产品驱动的:FactSet、Bloomberg、S&P Global 和 LSEG 已经在几乎每一个目标账户的续约谈判里,Microsoft 的 15 million 个付费 Copilot 席位和 160% 同比增长也意味着,银行家评估 Rogo 之前,横向替代方案往往已经拿到许可。信任和监管姿态则有利有弊:Rogo 已披露的 EU AI Act 合规(第一章已说明)相对 Microsoft Copilot 这类供应商是真差异化,后者在 2026 年初确认发生过数据丢失防护绕过;但 Bloomberg、FactSet 和 S&P Global 等传统厂商带着数十年的既有合规基础设施,新进入者无法很快复制。[CP010, CP021, CP022, CP023, CP024, CP035]
| 采购标准 | Rogo | Hebbia | AlphaSense | Bloomberg (ASKB) | FactSet AI | Microsoft Copilot | Glean |
|---|---|---|---|---|---|---|---|
| 智能体式交易工作流生成(CIM、可比公司、备忘录) | 强(核心重点) | 新兴(财务建模智能体,Sep 2025) | 不是重点 | 未披露 | 新兴(FactSet AI for Banking) | 不是重点 | 不是重点 |
| 批量文档 / 数据室智能 | 通过集成支持 | 强(核心重点,Matrix) | 强(券商 / 专家研究检索) | Document Search(测试版,85K+ 用户) | AI 文档搜索测试版 | 仅限通用办公文档 | 强(权限感知企业图谱) |
| 原生就地 Excel 计算 | 公司借 Subset/Offset 收购宣称具备;独立基准测试尚未验证 | 不支持(改为生成新模型) | 未披露 | 面向 Excel/BQuant 输出 BQL 代码 | 未披露 | 原生(Copilot in Excel) | 不适用 |
| 授权外部市场数据广度 | 集成 LSEG、PitchBook、S&P Global、FactSet | 不是数据供应商;摄取客户提供的文档 | 1,000+ 家卖方 / 独立研究供应商 | 800+ 家研究供应商,加上自有 Bloomberg Intelligence | FactSet 自有数据集,47+ 年 | 原生没有;取决于接入的数据源 | 通过 MCP 聚合:CB Insights、Crunchbase、Daloopa、FactSet、S&P Global |
| 来源引用 / 可追溯性 | 审阅来源未能独立验证 | 对源文档逐单元格引用 | 片段级引用 | 归因至原始研究 / 新闻来源 | 为公开 / 私有财务数据添加源文档链接 | 未披露 | Enterprise Graph 引用有权限的来源 |
| 横向办公套件集成(Word/PowerPoint/Teams) | 未披露 | 不是重点(浏览器端) | 不是重点(浏览器端) | 仅 Excel/BQuant | 未披露 | 强(M365 原生覆盖) | 与 Teams 等企业应用集成 |
| 受监管行业合规姿态(SOC2/EU AI Act/DLP) | 已达到 EU AI Act 合规(据公司披露,见第 1 章) | 未披露 | 宣称企业级数据保护 | 长期企业合规基础设施 | 长期企业合规基础设施 | 已确认 DLP 绕过事件(Jan-Feb 2026) | 宣称符合 SOC2/HIPAA/ISO27001/GDPR |
标为“未披露”的单元格,表示本章审阅来源中没有可验证的公开表述,并不等于该能力确认不存在;不支持的单元格刻意保留为缺口,而非估算。
[CP009, CP010, CP012, CP013, CP014, CP021]| 公司 | 定价模型 | 已披露区间 | 包含内容 | 备注 |
|---|---|---|---|---|
| AlphaSense | 分层报价制,年度最低消费 | $10,000-$15,000/user/yr(Core),最高 $50,000-$100,000+/yr(Enterprise 团队许可);Expert Transcript Library 为 $25,000-$50,000+/user/yr | 券商 / 独立研究、申报文件、新闻、情绪分析、仪表盘;更高层级增加内部内容托管和 IT 支持 | 第三方成本分析网站公开对五个层级做了基准比较 |
| Marvin Labs | 按席位订阅 | Standard 层级 $89/month;可免费试用 | 自动研究申报文件、新闻稿、业绩会,并附源链接引用 | 供应商自行披露;本比较中规模最小的工具 |
| Daloopa | 免费增值 | 提供免费层级;审阅来源未披露付费层级 | 自动抽取财务数据进模型 | 独立比较将其定位为同行中成本最低的入门点 |
| Hebbia | 仅企业版,定制报价 | 未公开披露 | Matrix 文档智能网格、多智能体编排 | 审阅的任何来源都未找到公开按席位价格 |
| Microsoft 365 Copilot / Agent 365 | 按席位加购,加上智能体编排层 | Agent 365 $15/user/month(GA May 2026);基础 M365 许可证价格将于 July 2026 上涨 | Copilot Chat、Copilot in Excel/Word/Teams、面向自定义智能体的 Agent 365 编排 | 自 2024 年以来多次重构定价;应视为动态目标 |
| Glean | 企业定制报价 | 审阅来源未公开披露 | 权限感知企业搜索、Glean Assistant、金融服务 MCP 集成 | 本章获取的官方或分析师来源中未找到数字定价 |
| Rogo | 未公开披露 | 未公开披露 | 智能体式交易工作流平台,集成授权数据(LSEG、PitchBook、S&P Global、FactSet) | 证据缺口:获取的来源没有披露 Rogo 按席位或按合同定价 |
金额来自各供应商披露,或本章审阅的第三方定价 / 成本分析网站;标为“未公开披露”的行是明确证据缺口,而非估算。
[CP022, CP034, CP035, CP036, CP037, CP046]把 TP003 的七项核心采购标准与七家厂商交叉,展示基于本章审阅证据,各厂商在哪些标准上优势明确、刚开始成形,或未披露 / 非重点。
单元格标签把 TP003 压缩成便于视觉扫描的短分类评级;每项评级背后的完整证据链接细节见 TP003。
[CP047, CP010, CP013, CP016, CP023]3.4 切换成本、锁定、多栖使用与分发权力
这个市场里的分发权力,往往在正式供应商评估开始前就已经决定。最清楚的证据是一个反例:Ensis Partners 是一家 2026 年 2 月成立、聚焦重组的投行,当它需要交易和关系管理基础设施时,创始人没有评估其他竞品平台,直接选择了 Intapp DealCloud,因为两人此前在 PJT Partners、Perella Weinberg、Blackstone 和 Citigroup 都亲自部署过 DealCloud。这个模式——银行家校友网络默认选择自己已经熟悉的平台——有利于在实际采购人群中现有足迹最广的供应商,而这是 Rogo 作为新进入者无法完全复制的分发优势。多栖使用是这个品类的常态,而不是例外:第三方比较指南明确建议评估 Rogo 或 Hebbia 的买方,同时并行运行 AlphaSense 或 FactSet 等特定来源平台,因为没有单一工具同时覆盖交易工作流生成和深度市场数据广度。LSEG 既是 Rogo 的数据授权伙伴,又出售自己的 Deep Research 智能体,这种竞合角色带来特定锁定风险:如果 LSEG 优先支持自己的原生 Workspace 智能体,Rogo 持续数据访问的条款就会变成一个重大合同风险问题,本章无法仅凭公开来源解决。与此同时,已经投资自有内部 GenAI(JPMorgan 的 LLM Suite、Goldman 的 GS AI Assistant)的大型银行,也在相反方向面临自己的切换阻力:既然已经围绕自研工具搭建内部基础设施和治理,它们就没有那么强的理由再为 Rogo 所瞄准的同一批工作流增加一个完全外部供应商,至少对有内部工程资源建设和维护的大型投行账户来说如此。[CP004, CP005, CP009, CP018, CP019, CP031]
3.5 护城河耐久性、商品化风险与负面证据
Rogo 最耐久的护城河不在模型层,而在工作流集成层:深度接入银行自己的 SharePoint、CRM 和数据室,再叠加授权外部数据合作(LSEG、PitchBook、S&P Global、FactSet),创造出横向智能助手无法复制的切换成本,除非后者也逐家银行完成同样的金融专用集成工作。但这道护城河在三处比看起来更暴露。第一,在最大、最赚钱的账户里,大型投行越来越倾向于自建而不是采购:JPMorgan 的 LLM Suite 已经能用银行自己的专有数据在约 30 秒内生成完整路演材料,并且每八周刷新一次,直接替代 Rogo 所售工作流。第二,Hebbia 约 $700 million 估值和重叠的企业客户基础,使它不是小众陪跑,而是资本充足的直接竞争对手;F2 独立基准测试过的 Excel 引擎,则为确定性计算设下了公开可靠性门槛,Rogo 和 Hebbia 都没有用披露结果追上。第三,也是对整个品类最关键的一点,独立可靠性基准确实不利:BankerToolBench 研究由 Goldman Sachs、JPMorgan、Morgan Stanley 和 Evercore 的 502 名投行人员共同构建,发现 9 个被测试 AI 模型没有一个能在无人修改下产出客户就绪内容;配套基准还测得领先模型的幻觉率为 86%。Deloitte Australia 2025 年 7 月向澳大利亚政府提交含伪造引用报告的真实失败案例,说明这不是理论风险。包括 Rogo、Bloomberg 的 ASKB、LSEG 的 Deep Research 和 Kensho 的 API 在内,所有严肃供应商的缓解模式都相同:让每个答案都落在有来源链接的引用上,使错误可以被审计,而不只是减少错误数量。Rogo 自己对这一原则的实现,能否承受与传统厂商同样的审视,仍是本章公开来源无法回答的开放尽调问题。[CP038, CP039, CP040, CP041, CP042, CP043]
| 护城河主张 | 威胁 | 严重性 | 缓解措施 / 尽调问题 |
|---|---|---|---|
| 与银行系统(SharePoint、CRM、数据室)和授权数据伙伴(LSEG、PitchBook、S&P Global)深度集成 | 数据 / 平台在位者(Bloomberg、FactSet、LSEG、S&P Global)正把原生 GenAI 直接嵌入银行已经授权并付费使用的终端 | 高 | 将 Rogo 多年合同续约节奏与各在位者 AI 功能上线时间线对照;索取命名账户重叠数据 |
| 面向银行家特定任务(CIM、可比公司分析、备忘录)的工作流原生智能体式自动化(Felix) | 横向 AI 副驾(Microsoft 365 Copilot、Glean)已获企业级授权,可能以很低增量成本加入金融专项微调 | 中 | 跟踪 Microsoft Copilot for Finance 面向银行的智能体上线、赢单率,以及相对 Rogo 的治理驱动采纳延迟 |
| 客户信任和参考客户基础(Baird、Moelis、Nomura、Tiger Global) | 大型全能投行(JPMorgan、Goldman Sachs、Morgan Stanley)正在自建专有内部 GenAI,缩小最大、最赚钱客户中的可服务市场 | 高 | 尽调 Rogo 在大型全能 / 全球银行与中型市场、精品投行、PE 公司之间的渗透分布;后者缺少内部 AI 建设预算 |
| 数据室 / 尽调文档智能与 Hebbia Matrix 重叠 | Hebbia 资金充足(估值约 $700M),是直接竞争者;它拥有重叠的企业金融服务客户基础,并在扩展建模智能体能力 | 高 | 索取 Rogo 与 Hebbia 正面对抗的赢 / 输数据和命名账户重叠 |
| Rogo 收购 Subset/Offset 后提出的 Excel 自动化主张 | F2 原生 Excel 计算引擎经独立基准测试达到 SpreadsheetBench Verified 95.25%,为公开可靠性设定门槛;Rogo 尚未用披露的基准结果追平 | 中 | 索取 Rogo 自己的 SpreadsheetBench 风格基准结果,或第三方对其 Excel 自动化准确率的审计 |
| 全品类 AI 可靠性与信任 | 独立基准(BankerToolBench、JurisTech)显示,当前没有 AI 模型能在不经人工修订时稳定产出可直接交付客户的投行业务材料;这可能拖慢企业级普及,或把银行推回分析师外包 / 内部自建 | 中 | 索取 Rogo 自己的准确率 / 幻觉基准结果,以及人在回路审阅流程证据 |
| LSEG 数据伙伴关系 | LSEG 同时是授权数据供应商,又通过自有 Deep Research 智能体直接竞争;如果 LSEG 偏向原生产品,将带来竞合风险 | 中 | 澄清 Rogo-LSEG 合作的合同保护(数据访问连续性、竞业限制条款) |
严重性评级是本章基于已审阅证据,对可能性和潜在影响做出的定性判断,不是供应商披露或第三方审计的风险分数。
[CP003, CP007, CP008, CP010, CP018, CP019]紧凑记分卡从六个维度评估 Rogo 竞争护城河的准备度,基于本章审阅证据按 1-10 打分(越高 = 护城河越耐久 / 暴露越低)。
分数是本章对 TP005 证据的定性综合,不是披露或第三方审计的指数。
[CP003, CP008, CP014, CP017, CP038, CP044]3.6 图表与证据
04财务
4.1 收入模型、定价与 ARR 可见性缺口
Rogo 的收入模型更适合按企业 SaaS 式授权来理解:直接卖给投行、私募股权公司和资产管理人,再叠加 Felix 这个智能体插件,自动跑交易筛选、CIM 生成、买方触达等多步工作流。本章和第 3 章竞争对手定价表审阅的资料,都没有披露 Rogo 的按席位或按合约标价;一页竞品对比把 Rogo 归为「企业定制定价」产品,隐含地把它同更轻的自助式工具区分开。Rogo 自己没有价目表时,可比 AI 原生研究平台就是唯一可用的定价锚:AlphaSense 披露每席位每年 $10,000-$20,000,平均企业合约规模 $50,000-$100,000+;Glean 估计为每用户每月 $45-50+,最低年度承诺 $50,000-$60,000——两者都明显高于 Microsoft 365 Copilot 每用户每月 $30。历史披露收入既少又不一致:Forbes 称 Rogo 收入从 2024 年约 $2 million 增至 2025 年超过 $15 million,CB Insights 却另列 Rogo 2024 年收入约 $1 million;没有公司确认数字,本章无法消化这处差异。OpenAI 自己的案例研究称,Rogo 自 2024 年走出隐身以来 ARR 增长 27x;但一篇独立 LLMOps 技术评审提醒,这个数字和反复出现的「每周节省 10+ 小时」一样,都是自报口径,缺少独立验证前应按营销材料打折看。已审阅资料没有一份把披露的历史收入与当前 2026 年 ARR 区分开;这是本章必须点明、不能抹平的核心可见度缺口。行业定价分析还指出,固定按席位定价会让 AI 产品的成本和价值错位,因为重度用户在同一订阅费下产生的推理成本可能约为轻度用户的 100x;这一机制很可能也适用于 Rogo 自己未披露的定价结构,不只适用于同行。[CI001, CI002, CI003, CI004, CI005, CI006]
| 收入流 | 机制 | 披露状态 | 估算作用 | 置信度 | 尽调问题 |
|---|---|---|---|---|---|
| 核心平台订阅 | 面向金融机构的企业许可,可能按席位或按公司签年度合同 | Rogo 未公开定价,也未拆分披露 | 主要收入流(推断) | 低 | 向管理层索取 Rogo 收入流拆分 |
| Felix 智能体式工作流加购模块 | 自主多步工作流(交易筛选、CIM 生成、买方触达、尽调) | 未作为单独项目披露 | 扩张 / 加购收入(推断) | 低 | 确认 Felix 是否与核心订阅分开定价 |
| 前线部署实施服务 | 由银行家牵头上手,并与 SharePoint/CRM/数据室做定制集成 | 未披露 | 服务收入(推断) | 低 | 索取服务收入与订阅收入拆分 |
| 数据伙伴集成转嫁(如 LSEG、PitchBook) | 数据授权成本 / 费用可能打包,也可能转嫁 | 未披露 | 推测 | 低 | 澄清数据伙伴成本计入 COGS,还是向客户转嫁收费 |
所有行均由 Rogo 产品机制(第 1 章和第 3 章)以及一页竞品比较推断而来;本章获取的任何来源中,Rogo 都没有披露收入流拆分。
[CI005, CI006, CI041]| 供应商 / 细分市场 | 定价模型 | 估算价格 | 合同期限 | 置信度 | 依据 |
|---|---|---|---|---|---|
| Rogo(企业级,全细分市场) | 定制企业合同 | 未披露 | 未披露 | 低 | 第 3 或第 4 章获取的任何来源都未找到公开按席位或按合同价格 |
| AlphaSense(可比公司) | 按席位年度订阅 | $10,000-$20,000/seat/yr;平均交易 $50,000-$100,000+ | 年度 | 高 | 公司通过 Sacra 档案披露 |
| Glean(可比公司) | 按席位月度订阅 | $45-50+/user/month;约 ~$50,000-$60,000/yr 最低承诺 | 年度最低消费 | 中 | 独立分析师估算(AgentMarketCap) |
| Microsoft 365 Copilot(参考) | 按席位月度加购 | $30/user/month | 月度 / 年度 | 高 | 公开标价(见第 3 章) |
| AI 优先 SaaS 行业常态 | 平台费 + 用量消耗的混合模式 | $500-$2,000/month 平台费,另加基于用量的超额费用 | N/A | 中 | 行业基准(GetMonetizely),非 Rogo 专属 |
Rogo 自身定价在第 3 和第 4 章获取的所有来源中都完全未披露;第 2-5 行是用于约束合理企业合同价值的间接可比项,不是已确认的 Rogo 价格。
[CI005, CI006, CI007, CI008, CI009]示意从签约企业合同,经平台使用,到确认订阅和服务收入的流程,因为 Rogo 未披露实际收入确认瀑布。
节点顺序根据 Rogo 披露的产品机制(第 1 章和第 3 章)以及可比企业 AI 收入模式推断;Rogo 未披露实际收入确认瀑布。
[CI005, CI009, CI041]4.2 GTM 规模、招聘与销售效率代理指标
Rogo 最具体的 GTM 代理指标不是收入,而是披露的用户数和机构数:2026 年 4 月官方 Series D 公告称,平台服务 250+ 家机构的 35,000+ 名金融专业人士,包括 Rothschild & Co、Jefferies、Lazard、Moelis 和 Nomura。三个月前的 2026 年 1 月 Series C,独立报道给出的同类数字约为 50+ 家一线机构的 25,000 名专业人士——说明具名用户数明显增长,接触过平台的机构数跳升更大;不过机构口径很可能比付费企业席位更宽。供给侧,Rogo 正在激进扩招:Growjo 估算(低置信度、算法生成)员工接近 292 人,同比增长 143%;截至本章访问日,Rogo 实时招聘页列出 56 个开放岗位,覆盖纽约、伦敦、新加坡的工程、销售、产品、安全和客户成功。纽约州 Empire State Development 机构另行确认,将向 Rogo 提供最高 $6.5 million、与绩效挂钩的 Excelsior Jobs Program 税收抵免,条件是新增 422 个全职岗位并向曼哈顿总部投入近 $14 million——这是一项公开承诺,只有 Rogo 达到招聘和投资里程碑才会兑现。作为销售效率代理指标,Rogo 自称每用户每周节省 10+ 小时,高于一项跨供应商 2026 年 AI 智能体生产力基准的 6.4 小时中位数;该基准汇总 McKinsey、Gartner、Forrester、Bain、Deloitte、BCG 和 MIT Sloan 研究,还发现只有 41% 的 AI 智能体部署在第一年达到 ROI 目标,平均回本期 6.7 个月。这些外部标尺可用来检验 Rogo 的自报数字,而不是照单全收。[CI010, CI011, CI012, CI013, CI014, CI015]
4.3 成本结构与利润率驱动因素
已审阅资料没有披露 Rogo 的毛利率、COGS 构成或模型推理开支,本章只能借可比公司和行业基准框出可能区间。一篇独立 LLMOps 技术评审称,Rogo 分层使用 OpenAI 模型:GPT-4o 处理面向用户的问答,o1-mini 处理数据语境化,更昂贵的 o1 留给高风险评估和推理。这是一套有意的成本优化模式,把最贵模型层限制在最需要它的工作负载上。该模式与 2026 年 GPU FinOps 基准一致:推理如今占企业 AI GPU 支出的 55-80%,每百万 token 成本从 A100 GPU 上约 $1.67 到 H200 GPU 上 $4.54+ 不等;单个 70-billion-parameter 模型服务真实企业流量,光算力一年就可能约 $347,000。更广泛的 AI 优先型 B2B SaaS 经济性分析估算,AI 原生软件毛利率为 55-70%,传统 SaaS 为 78-85%;差异来自可变推理 COGS,占收入 20-40%,而经典 SaaS 低于 5%。这与 Rogo 竞争的传统金融数据巨头不是同一种经济模型。FactSet Research Systems 于 2025 年 10 月向 SEC 提交 2025 财年 10-K(截至 2025 年 8 月 31 日),以及基于同一文件汇总的第三方数据,显示 FactSet 在 $2.32 billion 收入上的过去十二个月毛利率约 52.7%——这是 AI 推理成本结构出现前的传统基础设施基准,不应读成 Rogo 专属数字。Rogo 自己的服务交付模型又放大了成本问题:一页竞品对比称 Rogo 提供「白手套、银行家主导实施」,暗示每个企业账户的服务交付成本显著高于自助式产品,但没有资料量化这项成本。[CI018, CI019, CI020, CI021, CI022, CI023]
| 指标 | Rogo 数值 / 状态 | 可比基准 | 置信度 | 尽调问题 |
|---|---|---|---|---|
| 毛利率 | 未披露 | AI 优先 SaaS 同行区间 55-70%(GetMonetizely);传统数据平台同行 FactSet FY2025 约 52.7%(SEC/Macrotrends) | 低 | 向 Rogo 管理层索取 COGS/毛利率拆分 |
| 本年度(2026)ARR | 公司未披露;Growjo 估算年化约 ~$43.2M(低置信度算法估算) | AlphaSense $600M ARR(2026)与 Glean $200M ARR(2026) | 低 | 索取经审计或经管理层确认的 ARR 明细表 |
| 已披露历史收入增长 | 据 Forbes,~$1-2M(2024)升至 >$15M(2025);CB Insights 显示 2024 基数较低,约 ~$1M | n/a | 中(两个追踪器对 2024 年基数不一致) | 与公司确认,调和两个追踪器的 2024 年收入基数差异 |
| 已披露用户 / 客户规模 | 35,000+ 名专业人士、250+ 家机构(Apr 2026,公司表述);较 Jan 2026 约 25,000 名用户、50+ 家公司增长 | n/a | 中高(公司通过新闻稿披露) | 独立核验注册用户数与活跃席位数 |
| 员工数 | 约 292 名员工(Growjo,低置信度);管理层目标 YE2026 约 300 人(据公司概览);截至访问日有 56 个在招岗位 | n/a | 中 | 确认实际全职员工数及承包商占比 |
| CAC 回收期 / 销售效率代理指标 | 未披露 | 行业 AI 智能体中位回收期 6.7 个月;41% 在第一年达到 ROI(DigitalApplied) | 低 | 索取 Rogo 专属 CAC 与回收期数据 |
| $2B 估值隐含 ARR 倍数 | 无法计算——ARR 未披露;将可比区间套用到 Growjo 估算,隐含约 46x-93x | AlphaSense 约 12.5x ARR;Glean 约 36x ARR;2026 年后期 AI 中位数约 25.8x(Qubit Capital) | 低 | 需要公司披露 ARR,才能有把握计算 |
多数行按设计留空,因为 Rogo 没有披露底层指标;可比公司只是来自相邻 AI 原生和传统数据平台公司的方向性参照,不是 Rogo 专属数据。
[CI001, CI002, CI010, CI011, CI013, CI014]示意成本 / 利润率桥:从企业合同,经模型推理和服务交付成本,到估算毛利区间,因为 Rogo 未披露实际 COGS。
节点顺序和 55-70% 毛利区间来自行业 AI-first SaaS 基准,以及对 Rogo 模型架构的独立技术审阅,不是 Rogo 披露的 COGS。
[CI018, CI019, CI023, CI039]4.4 资本充足性与融资依赖
公司概览已经梳理 Rogo 从 2024 年种子轮到 2026 年 4 月 Series D 的逐轮融资时间线;本章不再复述,转而看前瞻性的资本充足性。两个独立数据跟踪器对累计融资总额大体一致:CB Insights 列示六轮共融资 $310.5 million,Growjo 列示 $314 million,都与公司概览已确认的「超过 $300 million」相符。2024 年 7 月 25 日,「Rogo Series A ScOp SPV Jul 2024, a Series of CGF2021 LLC」向 SEC 提交 Form D;该 Delaware 实体主张 Section 3(c)(1) 私募基金豁免,独立确认至少一名 Series A 投资方 ScOp Ventures 通过特殊目的载体投资,而不是直接投向 Rogo。该文件本身没有披露 Rogo Series A 的估值或轮次规模,因为该馈入型 SPV 的 Form D 报告覆盖的是其自身发行,而不是底层被投公司的条款。资金用途上,Rogo 官方 Series D 公告称资本将深化机构合作、扩展 Felix 智能体平台并加速全球扩张,但没有在这些用途之间给出具体金额分配;此前 Series C 报道称,当时 Rogo 员工刚过 100 人,融资用于欧洲增长、扩大 R&D 能力,以及为北美合作伙伴提供跨境支持。已审阅资料均未披露 Rogo 手头现金、月度烧钱或现金续航月数,因此资本充足性只能靠推断:仅已披露的招聘承诺(422 个新岗位、约 $14 million 总部资本开支、约 $40 million 计划 R&D 支出)就指向一个资本密集的近期增长阶段;在任何可见盈利拐点前,它会消耗已募资金的相当一部分。Axios Pro 2026 年 1 月 28 日独家报道独立佐证了公司概览已使用的 $750 million Series C 估值;尽管正文在标题之后付费墙不可见,它仍为该数字提供了额外独立数据点。[CI024, CI025, CI026, CI027, CI028, CI029]
| 项目 | 估算 / 状态 | 置信度 | 备注 |
|---|---|---|---|
| 迄今累计融资总额 | 截至 April 2026 Series D,6 轮合计约 ~$310-314M(CB Insights 为 $310.5M;Growjo 为 $314M) | 中 | 完整逐轮时间线见公司概览;此处重述生命周期融资总额,仅用于资本充足性框架 |
| 最近估值 | $2B(April 2026 Series D) | 中 | 为提供背景,重述公司概览信息;此处未独立重算 |
| 在手现金 | 未披露 | 低 | 私营公司无披露义务;向管理层索取 |
| 估算现金跑道 | 没有现金和烧钱数据,无法计算 | 低 | 需要披露烧钱率和现金余额 |
| Series C + D 募资计划用途 | 全球 / EMEA 扩张、Felix 平台 R&D、深化机构伙伴关系(Series D);欧洲增长、R&D 扩张、北美跨境伙伴支持(Series C) | 中高 | 资金用途表述来自公司,未经独立审计 |
| 与招聘挂钩的资本承诺 | $6.5M 纽约州 Excelsior 税收抵免,挂钩 422 个新增全职岗位、约 ~$14M 总部资本开支、约 ~$40M 计划 R&D 支出 | 中高 | 绩效挂钩;只有达到招聘 / 投资里程碑,抵免才归属 |
| 融资结构细节 | 至少一名 Series A 投资者(ScOp Ventures)通过 Delaware SPV 投资,依据 Section 3(c)(1) 私募基金豁免(SEC Form D,July 2024 提交) | 高 | 证实存在 SPV 投资结构;本身不披露估值或轮次规模 |
| 下一次资金需求 / 触发条件 | 未披露;未找到 IPO 或追加融资时间线 | 低 | 询问管理层下一轮融资触发条件,以及目标 ARR / 利润率门槛 |
各行只在需要构建前瞻资本充足性框架时重述公司概览事实;每项都由本章本地来源支持,而非公司概览声明 ID。
[CI024, CI025, CI026, CI027, CI028, CI029]把 Rogo 累计融资额映射到其披露的资金用途类别,并与公司概览中已覆盖的逐轮融资时间线区分开。
这是资金用途 / 资本强度视角,不同于公司概览中的融资时间线图;总部资本开支和 R&D 的美元数字来自 Excelsior Jobs Program 承诺,不是已确认的 Series D 专项分配。
[CI015, CI027, CI028, CI030]4.5 财务结论:收入质量、利润率路径与尽调阻断项
Rogo 没有披露当前 ARR,因此其 $2 billion Series D 估值无法用公司确认的收入倍数来校准——这是本章最关键的缺口。可比 AI 原生企业研究平台显示,合理倍数区间可以很宽:AlphaSense 在 $7.5 billion 估值下,对应披露的 $600 million 2026 ARR 约 12.5x;Glean 在 $7.2 billion 估值下,对应披露的 $200 million 2026 ARR 约 36x;更广泛的 2026 年 AI 初创公司估值数据则显示,后期轮次倍数中位数接近 25.8x,整体范围为 10x-50x。若把该区间套到 Growjo 对 Rogo 的低置信度 $43.2 million ARR 估算上,隐含倍数约 46x-93x——高于本章引用的所有可比对象。由此可见,任何估值结论都高度依赖未经验证的第三方估计,而不是公司确认数字。独立分析师对这类倍数是否合理存在分歧:SixThirty Ventures 认为,包括 Rogo 在内的 AI 分析师品类融资,相对于仍在发展中的经常性收入牵引,计入了「离群级收入倍数溢价」;Finro 2026 年 Q1 估值数据库发现,投资者越来越奖励「变现清晰度、利润率质量和持久性」,同时重估仍在卖叙事、未证明单位经济性的公司。放在这个框架里,Rogo 未披露 ARR、利润率和流失数据,无法证明自己属于被奖励的一组,而不是被重估的一组。收入质量上,Rogo 披露的历史增长(按 Forbes,2024 年约 $2 million 增至 2025 年超过 $15 million;但 CB Insights 给出更低的 2024 年基数)确认了小基数上的强劲百分比增长,却没有说明 2026 ARR、毛利率、净收入留存或客户集中度——这些仍是任何承销判断的阻断性缺口。尽调应按顺序优先拿到:公司确认的当前 ARR、经审计或管理层提供的毛利率与 COGS 拆分、已披露现金头寸与烧钱率。没有这三项数据,本章任何估值倍数结论都只是有边界的估算,而不是经验证的判断。[CI032, CI033, CI034, CI035, CI036, CI037]
| 数据点 | 可得性 | 重要性 | 尽调路径 |
|---|---|---|---|
| 经审计 GAAP 收入 / ARR | 未公开 | 阻断项 | 直接向 Rogo 财务团队索取经审计财务报表或 ARR 明细表 |
| 毛利率 / COGS 拆分(含模型推理成本) | 未公开 | 重要 | 索取 COGS 明细,包括 OpenAI 使用成本和基础设施支出 |
| 现金余额和烧钱率 | 未公开 | 阻断项 | 索取最新资产负债表和月度烧钱趋势 |
| 按席位 / 按合同定价 | 未公开 | 重要 | 索取价目表,以及按细分市场划分的实际 ACV 分布 |
| 客户集中度、流失率和净收入留存 | 未公开 | 重大 | 要求提供按队列拆分的 ARR 和留存数据 |
| 收入确认政策(订阅 vs. 用量 vs. 服务) | 未公开 | 次要-重大 | 要求提供会计政策备忘录 |
| 2024 年收入基数对账(Forbes 约 $2M vs. CB Insights 约 $1M) | 第三方估算相互冲突 | 次要 | 要求提供公司确认的 FY2024 收入数据 |
每一行对应本章研究中识别出的证据缺口;底层尽调理由见 localEvidence.evidenceGaps。
[CI022, CI029, CI042]区间图框定 Rogo 已披露和估算收入、间接毛利率基准,以及可比公司 ARR 倍数,因为 Rogo 未直接披露当前 ARR 或利润率。
所有数字要么是单一披露数据点,要么明确标注为低置信度第三方估计和可比公司基准;除 2024/2025 年收入数字外,没有任何数据得到 Rogo 确认。
[CI001, CI020, CI035, CI037]紧凑记分卡汇总截至本章研究时,Rogo 关键经济维度上的财务尽调信号质量。
[CI021, CI024, CI029, CI033, CI036]4.6 图表
05产品与技术
5.1 Felix 与核心产品入口
Rogo 的产品已从研究助手演进为 Felix:一个 AI 智能体,可把单条提示词转成可直接交付客户的 PowerPoint 演示文稿、Excel 模型、Word 文档、看板和带来源的研究。Felix 可通过邮件、聊天和原生 Excel 插件调用,也可承接异步、定时任务——例如监控一家公司,并在每次发布财报后重跑报告。具体金融工作流包括交易筛选、CIM 生成、买方触达、数据室尽调,以及可比公司分析、模型、推介书和投委会备忘录。 Rogo 2026 年 5 月发布把平台整合成更宽的产品界面:Rogo Agents 让机构把专有模板、方法论和重复工作流编码成可复用的自定义自动化;Custom MCP 让机构把自有或第三方 Model Context Protocol 服务器接入为额外工具;Slides Annotator 将演示文稿标注直接映射为可追踪、带版本的 Felix 修订;Memory 在多轮聊天中保留用户惯例和偏好;Library 集中管理 Felix 产出的所有工件。合在一起,这些功能把 Rogo 从单一用途研究聊天机器人,推向机构可按自身固定流程配置的工作流平台。[CE001, CE003, CE004, CE005, CE006, CE007]
| 模块 / 资产 | 主要用户 | 状态 / 成熟度 | 差异化 | 尽调缺口 |
|---|---|---|---|---|
| Felix(核心智能体) | 投行人员、PE/VC 投资经理、股票研究分析师 | 已普遍可用,2026 年推出 | 模型无关的编排框架,一条提示词生成完整交付物 | 没有针对 Felix 的独立审计准确率基准 |
| Rogo Agents(自定义工作流) | 机构工作流负责人、团队负责人 | 已普遍可用,2026 年 5 月发布 | 让机构把专有模板 / 方法论封装成可复用智能体 | 未公开各机构部署的智能体数量 |
| Excel Plug-in | 搭建 / 审计模型的分析师 | 已普遍可用,2026 年 5 月发布 | 在 Excel 工作簿内原生调用 Felix,并以机构模板为依据 | 没有公开基准测试对比 Microsoft 自有 Copilot in Excel |
| 自定义 MCP 连接器 | IT / 数据工程团队 | 已普遍可用,2026 年 5 月发布 | 让机构把自建或第三方 MCP 服务器接成工具 | 未公开在生产环境使用自定义 MCP 的机构名单 |
| Slides Annotator | 修改推介材料的投行人员 | 已普遍可用,2026 年 5 月发布 | 演示稿批注直接映射为可追踪、有版本的 Felix 修改 | 未披露使用量或采用数据 |
| Memory | 所有平台用户 | 已普遍可用,2026 年 5 月发布 | 跨对话保留用户惯例 / 偏好,并保留可编辑出处 | 未披露 Memory 的数据保留或删除控制细节 |
| Library | 所有平台用户 | 已普遍可用,2026 年 5 月发布 | 集中存放 Felix 生成的所有交付物 | 未披露保留期限上限或机构间数据隔离细节 |
状态 / 成熟度和发布日期来自 Rogo 自家 2026 年 5 月产品更新帖;各模块的独立采用或使用数据未公开。
[CE001, CE003, CE006, CE007, CE008, CE009]| 用户任务 | 当前人工流程 | Rogo 方案 | 可量化收益 | 局限 |
|---|---|---|---|---|
| 交易筛选 | 在多个数据终端按投资假设人工筛选标的清单 | Felix 从已连接数据源中按既定标准筛选标的 | 每笔交易少花数小时手工建清单 | 筛选质量取决于投资假设是否写得足够清楚 |
| CIM 生成 | 分析师手工起草 50-100 页卖方备忘录 | Felix 基于已连接交易数据和机构模板起草 CIM 章节 | Rogo 和独立叙述称,起草时间从数天压到数小时 | 产出仍必须贴合机构固定文风,才能真正可用 |
| 买方触达 | 分析师手工整理联系人清单和触达文案 | Felix 生成个性化买方联系人清单和首轮沟通稿 | 加快早期买方清单搭建 | 个性化质量没有独立基准测试 |
| 数据室尽调 | 分析师手工审阅交易数据室里的大量文件 | Felix 汇总数据室文件,提炼关键发现 | 减少手工审文件时间 | 超大或结构混乱文件集的覆盖率没有独立验证 |
| 电子表格模型搭建 / 维护 | 分析师搭建并滚动更新 40 个工作表的模型,链接脆弱 | 来自 Subset 的电子表格智能体搭建、滚动更新并审计模型,同时追踪驱动项 | 按 Rogo 说法,模型更新从数天缩到数秒 | 高度定制或遗留模型结构上的可靠性没有独立测试 |
| 数据变化后的模型维护 | 新数据到来后,分析师手工复核并更新假设 | 来自 Offset 的智能体跟踪模型假设和公式如何变化,并自动更新 | 新信息到来时减少手工返工 | 刚整合(2026 年 3 月宣布);生产环境记录有限 |
| 工作簿内分析 | 分析师在终端、浏览器和 Excel 之间切换来填模型 | Excel Plug-in 让 Felix 在工作簿内填充财务数据并搭建分析 | 减少工具之间的上下文切换 | 前提是机构数据源已经接入 Rogo |
收益数据来自 Rogo 自家公告和独立评论;没有一项来自独立审计的工时研究。
[CE005, CE006, CE039, CE041]银行家的请求如何通过 Felix,从受理流转到经审阅的交付输出。
流程由 Rogo 自身产品描述和一份独立架构深度解析重建;Rogo 未正式绘制内部步骤顺序。
[CE001, CE004, CE005, CE006]5.2 数据、模型与编排架构
独立技术报道把 Felix 描述成一套编排外壳,而不是单个微调模型:编排脚手架、工具层、引用系统、输出格式化器和审计轨迹共同构成外壳;Rogo 再接入内部基准上表现最佳的前沿模型,目前覆盖 OpenAI、Anthropic 和 Google。另一份独立案例研究描述了这套外壳下的分层路由:主模型处理基于聊天的分析,小模型处理数据语境化和搜索结构化,顶级模型留给评估和合成数据生成,在不同任务类型之间权衡质量、延迟和推理成本。 数据层覆盖授权外部数据商,核心包括 LSEG、S&P Capital IQ/FactSet 和 PitchBook,也包括 Moody's、Daloopa、Affinity;内部连接器则接到 SharePoint、Salesforce 和机构专属文档库。Rogo 维护自己由从业者编写的评估套件 Big Finance Benchmark,并在 GitHub 发布了参考框架;它只保留最小四工具脚手架(网页搜索、SEC EDGAR 搜索、URL 抓取、沙箱化 Python 执行),有意排除高级数据源和向量库检索,让评估衡量底层模型,而不是 Rogo 自己的工具。[CE002, CE011, CE016, CE017, CE026, CE027]
| 层 / 组件 | 角色 | 依赖 | 风险 |
|---|---|---|---|
| 基础模型(OpenAI、Anthropic、Google) | Felix 背后的主要推理和生成引擎 | 前沿模型 API 的可用性、定价和能力 | 供应商集中与定价风险;OpenAI 和 Anthropic 也各自在切入面向金融的 AI 产品 |
| 智能体编排框架(Felix) | 编排脚手架、工具层、引用系统、输出格式器、审计轨迹 | 依赖内部工程投入,而不是某个单一模型 | Rogo 称编排框架质量是护城河,但没有与竞争对手做独立基准测试 |
| 分层模型路由 | 把聊天 / 分析、数据上下文化、评估任务路由到不同尺寸模型 | 持续获得主要供应商不同尺寸模型的访问权限 | 分层逻辑和阈值未公开记录 |
| 数据连接器层 | 集成 LSEG、Capital IQ/FactSet、PitchBook、Moody's、Daloopa、Affinity、SharePoint、Salesforce | 与各数据提供商持续保持商业数据授权关系 | 任一数据伙伴(如 LSEG、PitchBook)都可能限制访问或重新定价 |
| Big Finance Benchmark / 评估框架 | 内部评估套件,用来决定哪个前沿模型驱动 Felix | Rogo 自家从业者编写的任务集和评分标准 | 配套研究论文显示,当前最佳前沿智能体也只拿到评分标准 58.8% 的分数 |
| 安全智能体(Sisyphus) | 对 Rogo 自家基础设施做自动化、持续的攻击性安全测试 | 内部红队工具,并用人类安全团队校准 | 披露的漏洞数量和真阳性数据未经独立审计 |
| 部署 / 分发界面 | 电子邮件、聊天、Excel Plug-in,以及(自 2026 年 Q3 起)Microsoft Copilot Marketplace 技能 | Microsoft 的 Excel 和 Copilot Marketplace 作为联合分发渠道 | Microsoft 同时在给 Copilot in Excel 增加自家的竞争性金融数据连接器 |
架构层根据 Rogo 自家产品帖、独立 LLMOps 案例研究和技术博客重构;这些来源披露之外的内部实现细节未验证。
[CE002, CE011, CE022, CE023, CE026, CE030]分层展示 Rogo 产品架构:从数据连接器,经智能体框架,到应用界面和治理层。
层级边界由 Rogo 分散的产品和信任公告,以及一份独立技术案例研究重建;Rogo 尚未发布达到此细节级别的单一官方架构图。
[CE002, CE011, CE026, CE031, CE032]Felix 依赖的模型、数据和分发关系,每一类都带来不同的集中度或竞争风险。
依赖关系集合仅限本章审阅的 Rogo、合作伙伴或独立来源已确认关系;可能还存在其他未披露供应商关系。
[CE002, CE012, CE013, CE014, CE019, CE020]5.3 可靠性、引用行为与评估证据
可靠性上最清晰的第三方证据来自 BigFinanceBench,这是 Rogo 自有基准框架的配套研究论文,2026 年 6 月共同发布在 arXiv。它用 36,000 多个评分点评估 928 个专家编写、嵌入工作流的金融研究任务;评分检查推导的每一步——来源选择、期间和会计定义选择、假设、计算——而不只看最终答案。在十个当前前沿和开放权重智能体中,表现最佳的系统也只拿到可用评分点的 58.8%;作者认为,最终答案准确率是推导质量的有损代理指标,能力在各工作流阶段分布也不均匀。这直接关系到引用风险:智能体可以产出一个看似可信、引用充分的答案,但底层推导的某些步骤仍然出错,而审阅分析师必须逐项审计这些步骤。 另一篇独立 LLMOps 分析对 Rogo 自己公开指标的怀疑更直接:服务银行家数量、每周节省小时数、ARR 增长倍数等头部数字都是自报口径,在其审阅材料中没有独立验证;Rogo 的微调方法,以及吸收上游模型更新的方式,也没有公开细节。数据室尽调中规模很大或结构很差的文档集处理,同样缺少独立基准。超出 Rogo 自选基准后,平台在真实世界大规模场景下的推导准确率,仍是本章的证据缺口。[CE022, CE023, CE024, CE028, CE029, CE048]
Rogo 核心模块和收购能力的成熟度、主要买方与证据基础。
成熟度标签反映公开披露的正式发布与近期集成状态;没有按模块拆分的独立采用率数据。
[CE039, CE041, CE043, CE031, CE032]5.4 信任、安全与监管合规作为产品助推器
Rogo 的信任中心列出 SOC 2 Type I 和 Type II、ISO/IEC 27001、ISO/IEC 42001:2023 认证;最后一项是 AI 管理系统的首个国际标准,覆盖模型训练与编排管线、数据集成血缘和内部 AI 治理结构。信任中心还列出符合 CCPA、VPAT 无障碍报告和欧盟 AI 法案合规文件,并点名 Jefferies、Rothschild & Co、Nomura、Moelis、Lazard 等参考机构已审查 Rogo 的安全姿态。自 2026 年 8 月 2 日起,欧盟 AI 法案对高风险 AI 系统的义务将在全欧盟全面执行,要求形成文件化风险管理、技术文档、人工监督和上市后监测。Rogo 自身认证和信任中心披露意在帮助跨过这道合规门槛,但本章未找到监管机构确认其状态的证据。 独立技术报道还描述了第二个内部智能体 Sisyphus:它每天约一到两次针对 Rogo 自身基础设施跑自动化攻击性安全演练,把认证滥用、授权绕过、注入、SSRF 和 LLM 特有利用类别的发现串起来;同一来源称,它在第三方渗透测试后一周内又发现 18 个可利用漏洞,高置信度发现被校准到超过 95% 的真阳性率。Rogo 公开状态页另报 2026 年 4 月至 6 月窗口 API 正常运行时间 100%,应用正常运行时间 99.97%,一次局部故障约 18 分钟内解决。合在一起,安全与合规构成销售助推层;Rogo 的买家属于企业软件里最厌恶风险的一群客户。[CE031, CE032, CE033, CE034, CE035, CE036]
| 控制 / 认证 | 状态 | 范围 | 缺口 |
|---|---|---|---|
| SOC 2 Type II | 已认证 | 客户数据的安全性、可用性、处理完整性和保密性 | 不独立验证模型输出准确性 |
| SOC 2 Type I | 已认证 | 单一时点的控制设计 | 严谨度已被 Type II 覆盖;Trust Center 仍并列展示 |
| ISO/IEC 27001 | 已认证 | 信息安全管理体系 | 覆盖信息安全流程,不专门覆盖 AI 模型行为 |
| ISO/IEC 42001:2023 | 已认证 | AI 管理体系:模型训练 / 编排、漂移监控、数据血缘、AI 治理 | 认证确认治理流程存在,但不直接衡量模型准确性 |
| EU AI Act 合规文档 | 已在 Trust Center 发布 | 公司在 2026 年 8 月高风险义务执法日期前建立的自有合规计划 | 本章未看到独立监管机构确认其合规状态 |
| CCPA 对齐 | 列为合规 | 加州消费者数据隐私要求 | Trust Center 列示之外的对齐范围未说明 |
| VPAT(无障碍) | 已发布 | 无障碍符合性报告 | 未说明主张符合哪个级别 |
| Sisyphus 持续渗透测试 | 内部运行,约每天 1-2 场 | 针对 Rogo 自家基础设施的认证滥用、授权绕过、注入、SSRF、LLM 特定攻击 | 披露效果(18 个额外发现、>95% 真阳性率)未经独立审计 |
| 公开状态页(uptime SLA) | 100% API / 99.97% 应用可用率,2026 年 4-6 月 | Rogo API 和 Rogo Application 可用性 | 除 2026 年 4 月披露的一次部分宕机外,未发布更多历史事故细节 |
认证和状态条目来自 Rogo 自家 Trust Center 和产品公告;本次未直接审阅独立审计报告本身。
[CE031, CE032, CE033, CE034, CE036, CE037]5.5 自建还是收购:并购与平台扩展
Rogo 没有把所有能力都放在内部自建,而是靠收购快速补能力。Subset(2025 年收购)带来一个电子表格智能体,能理解复杂金融模型公式和区域,并连接 Capital IQ、FactSet、PitchBook、LSEG 及机构私有数据,用来搭建、滚动更新和审计模型;其创始人为 Jason Chan(曾任 Bank of America、Providence Equity Partners)和 AJ Nandi(曾任 Insight Partners)。Offset(2026 年 3 月宣布)由 Raj Khare 和 Shiv Shrivastava 创立,构建能记住特定财务模型假设、公式和输出如何随时间演变的智能体系统,让模型在新信息到来后自动维护——这与一次性生成模型是结构上不同的问题。Plux AI(2026 年初)由 Deepak Guneja 和 Pratyush Chaudhary 创立,监测申报文件、贷款方更新、法院文件和公司披露,提炼长文本市场信号,并扩大了 Rogo 在英国 / 欧洲的覆盖和工程团队存在。 分发扩张既靠合作,也靠收购。2026 年 6 月,Microsoft 将 Rogo 与 LSEG、Ramp、Vena 一起列为 Copilot for Excel 合作伙伴构建技能的首批伙伴,并计划自 2026 年第三季度起通过 Microsoft Marketplace 销售——这把 Rogo 放进了一个分发渠道,而 Microsoft 自己也在用同一渠道销售其独立扩张的金融数据连接器组合。Rogo 创始人已直接承认这种动态:Rogo 同时是 OpenAI 和 Anthropic 的客户、分发伙伴、潜在竞争目标,因为两家模型供应商也分别在自建面向金融的 AI 产品。[CE039, CE040, CE041, CE042, CE043, CE044]
| 日期 / 阶段 | 功能 / 里程碑 | 状态 | 含义 | 来源 |
|---|---|---|---|---|
| 2024 年 12 月 | S&P Capital IQ 数据集成 | 已发布 | 在后来深化 LSEG 和 PitchBook 之前,先扩展基本面 / 预测覆盖 | Rogo;AiThority |
| 2025 | 收购 Subset(电子表格智能体) | 已完成 | 把经过金融训练的电子表格智能体纳入内部,而不是从零自建 | Rogo |
| 2025 年(8 月 / 9 月) | LSEG 战略合作 | 已宣布 | 增加实时基本面、预测和 1.5M+ 笔交易的 M&A 数据库 | Rogo、LSEG 与 FinTech Global |
| 2026 年初 | 收购 Plux AI | 已完成 | 增加欧洲申报文件 / 法院文件监控,并带来英国 / 欧盟工程团队布局 | Rogo;Tech Funding News |
| 2026 年 3 月 | 收购 Offset(模型维护智能体) | 已完成 | 增加可跟踪模型假设 / 公式变化的智能体,减少手工模型维护 | PR Newswire 与 Tech Funding News |
| 2026 年 4 月 | Kleiner Perkins 领投 $160M Series D | 已完成 | 融资用于深化数据集成,并扩大 Forward Deployed Banker 覆盖 | Tech Funding News |
| 2026 年 5 月 | Felix、Rogo Agents、自定义 MCP、Excel Plug-in、Slides Annotator、Memory、Library、PitchBook Premium | 已发布 | 将 Rogo 的产品界面从研究助理整合为多界面智能体平台 | Rogo |
| 2026 年 6 月 | BigFinanceBench 论文发布(arXiv) | 已发布 | 提供公开、可由第三方检查的证据:即便领先前沿智能体,在真实推导任务上也只拿到评分标准 58.8% 的分数 | arXiv |
| 2026 年 6 月 | Microsoft Copilot in Excel 首发伙伴技能 | 已宣布,自 2026 年 Q3 起推出 | 通过 Microsoft Marketplace 分发 Rogo 构建的技能,与 LSEG、Ramp、Vena 并列 | Digital Trends |
| 2026 年 8 月(前瞻) | EU AI Act 高风险义务开始可执行 | 已排期 | 抬高 Rogo 平台在欧盟触及任何高风险 AI 用例的合规门槛 | Netguardia |
日期结合了已确认公告日期和月份级近似;当来源只披露季度或年份时,本表使用近似值。具体精度见单项证据条目。
[CE013, CE039, CE012, CE043, CE041, CE015]5.6 技术优势、限制与尽调缺口
Rogo 最清晰的技术优势不在某个模型,而在架构:一套跨前沿供应商、模型无关的编排框架;一组很深且持续扩展的授权金融数据连接器;基于嵌入式前线部署银行家的 GTM 模型,把机构特定工作流需求翻译成产品配置;以及一套安全 / 合规计划(SOC 2、ISO 27001、ISO 42001、Sisyphus 连续测试)。考虑到公司仍主要卖给金融服务中最保守的客群,这套计划成熟度并不常见。既然 Rogo 宣称的护城河是编排框架,而不是任何单一模型,前沿模型升级原则上就是配置变更,不是重写。 限制也正集中在尽调最该加权的地方。BigFinanceBench 自己发布的结果显示,最佳前沿智能体在嵌入工作流的推导任务上只拿到 58.8% 的评分点;这直接说明,引文级和步骤级准确率远落后于精美、格式漂亮的最终输出。即便还没考虑极大或混乱真实文档集上的规模问题,这个差距已经存在。另一篇独立 LLMOps 评审指出,Rogo 自己的头部使用和增长指标都是自报且未验证,并注明微调方法和模型更新处理未披露。一篇独立产品评审还补充说,上线导入(数据源集成、权限映射、模板配置)往往并不简单,定价也面向大型企业客户,而不是小机构。这些限制并非 Rogo 在垂直金融 AI 智能体中独有,但它们会收窄本章证据当前能弥合的营销话术与独立验证表现之间的差距。[CE045, CE049, CE047, CE028, CE023, CE050]
5.7 图表
06客户
6.1 金融采购层级中的客户分层
Rogo 自有材料描述了三个主要买方群体——投行、私募股权公司和资产管理人;Forbes 独立画像则用对冲基金替代资产管理人。不同来源对买方基础的框定存在细小但真实的差异。最清晰、最具体的证据集中在大型投行和全球投行:Rogo 官方客户页逐一标注 Jefferies、Lazard、Moelis、Nomura 和 Rothschild & Co 的标识位,2026 年一篇独立融资报道又把同一批机构列为活跃用户,形成佐证。Truist Securities 还提供了 CEO 具名背书。离开头部标识墙后,证据很快变薄:一名精品顾问银行分析师和一名管理 $300B AUM 的资管高级董事总经理,只以匿名证言出现在第三方评论聚合器上;GTCR 是唯一具名私募股权账户,也只是 Series C 融资报道中顺带提到,而不是专门案例研究。通过 Baird Equity Research 案例研究,股票研究是唯一具备账户级深度的细分。企业财务和企业发展团队只作为 Rogo 自身产品定位推断出的用例出现;抓取到的 2026 年来源没有点名任何具体企业内部账户。大型投行的采购结构也不同于精品机构或买方,因为 Jefferies 自己发布的 Rogo CEO 领导力访谈,把核心销售描述为围绕实时交易数据,在信息隔离墙和 MNPI 受控环境内搭建合规基础设施,而不是一次普通软件采购。[CU001, CU002, CU003, CU004, CU005, CU006]
| 分群 | 买方 / 用户 / 付费方 | 主要用例 | 规模 / 证据(2026) | 收入 / 战略价值 | 缺口 |
|---|---|---|---|---|---|
| 大型综合 / 全球投行 | 交易团队和研究部门;机构 IT / 采购付费 | 推介材料组装、可比公司分析、尽调备忘录、业绩综合 | 具名客户标识:Jefferies、Lazard、Moelis、Nomura、Rothschild & Co | 可见度最高的标杆账户;撑起同业可信度 | 未披露单家银行席位数或收入占比 |
| 中型 / 精品投行 | 小型顾问机构的交易团队 | 交易规模更小,但核心工作流相同 | 一位精品投行分析师的客户证言 | 大型投行的验证给小机构采用降风险,因此是扩张方向 | 未找到具名精品投行标识;证言匿名 |
| 私募股权 | 交易和投后团队;GP 层面预算 | 交易筛选、尽调综合、IC 备忘录起草 | 2026 年 Series C 报道将 GTCR 列为客户 | 证明 PE 是真实客群,不只是营销话术 | 仅一个具名 PE 账户;未披露使用量或席位数据 |
| 对冲基金 / 资产管理人 | 研究和投资组合团队;引用了一家 $300B AUM 机构 | 研究监控、申报文件综合、模型支持 | FeaturedCustomers 上一位匿名 $300B AUM 高级董事总经理证言 | Forbes 明确将对冲基金列为服务客群 | 未找到具名对冲基金或资管标识 |
| 股票研究 | 银行研究部门内的分析师和高级研究员 | 财报季电话会文字稿处理、模型基准测试、覆盖报告起草 | Baird Equity Research:100+ 活跃用户,约 85% WAU,约 70% DAU | 整个客户群中最深入、量化最充分的账户级证据 | 仅一个具名股票研究部署;尚未在第二家银行得到佐证 |
| 公司财务 / 公司发展 | 内部 M&A 和战略团队(根据产品定位推断) | 标的筛选、市场监控、董事会材料(推断用例) | 已抓取的 2026 年来源中未找到具名账户或使用量数据 | 相比卖方 / PE / 买方基本盘,这是规模最小、证据最弱的客群 | 完全由 Rogo 产品描述推断;出现具名账户前应视为未验证 |
分群行混合使用具名 / 引用账户,以及在没有具名账户时使用 Rogo 与 Forbes 自身的分群表述;公司财务行明确为推断,并非来自具名部署。
[CU001, CU002, CU003, CU004, CU005, CU006]Rogo 的客户旅程从顶级投行标杆客户关系出发,经过合规建设,进入驻场部署生产使用;扩张分支之外,也有真实但匿名的流失分支。
[CU002, CU005, CU041, CU046, CU048, CU050]6.2 采用轨迹与规模信号
约十八个月里,Rogo 公开报告的规模至少经历了四个不同数据点:OpenAI 2024 年合作伙伴案例中的 5,000+ 名银行家和 27x ARR 增长;2026 年 1 月 Series C 前后的约 25,000 名专业人士、每日处理 50,000 次查询;Forbes 公司画像仍显示的 25,000 用户 / 150 家机构;以及 Rogo 自己 2026 年 4 月 29 日 Series D 新闻稿中的 250+ 家机构、35,000+ 名专业人士,并由独立复盘报道佐证。截至本轮访问日,Forbes 仍未刷新到新数字,这本身就是一个有用的新鲜度信号:快速增长的私营公司公开画像,至少会落后真实使用量一个融资周期。在这条大趋势中,Baird 的 Equity Research 部署是唯一原则上可对齐使用细节的账户;但即便如此,市面上也流通三套指标:Rogo 自己的案例研究称 100+ 用户中约 85% 周活、70% 日活;一篇独立 Series D 复盘报道则称每周 10,000+ 个工作流、95% 参与度;客户评论聚合器又引用累计 250,000+ 个工作流。三组数字没有显而易见的共同分母,也没有抓取到的来源解释它们如何互相关联;这里应把它当作冲突数据的证据缺口,而不是硬合成一个头条指标。[CU009, CU010, CU011, CU012, CU013, CU014]
| 指标 | 数值 | 日期 | 来源 | 置信度 | 含义 |
|---|---|---|---|---|---|
| 服务的投行人员(早期基线) | 5,000+ | 2024 年(伙伴案例研究) | OpenAI | 中 | 为 Series C/D 扩张前的 2024 年规模建立下限 |
| 2024 年发布以来的 ARR 增长 | 27x | 2026 年报道 | OpenAI | 中 | 增长率很陡,但 2024 年收入基数未披露 |
| 使用平台的专业人士(Series C 窗口) | ~25,000 | 2026 年 1 月 | SuperbCrew(Series C 报道) | 中 | 2026 年 4 月 Series D 跳升前的中间时点数据 |
| 每日平台查询量 | 50,000 | 2026 年 1 月 | SuperbCrew(Series C 报道) | 中 | 找到的唯一公开查询量数据;没有按用户拆分 |
| 用户 / 机构数(2026 年 4 月仍展示) | 25,000+ / 150 家机构 | 截至 April 2026 资料页 | Forbes | 中 | 似乎落后于下方较新的 Series D 数据 |
| 用户 / 机构(Series D 公告) | 35,000+ / 250+ | April 29, 2026 | PR Newswire;独立综述 | 高 | 截至本次运行,这是最新披露的规模数据 |
| Baird Equity Research 活跃用户和参与度 | 100+ 用户;~85% WAU;~70% DAU | 截至 2026 年案例研究发布时 | Rogo(官方案例研究) | 中 | 目前可得最深的单一账户参与度代理指标 |
| Baird 每周工作流量和参与度(另一口径) | 10,000+ 工作流 / 周;95% 参与度 | April 2026 综述 | ai2.work | 中 | 与 Rogo 自己案例研究里的指标定义不一致 |
| Baird 累计工作流量(另一口径) | 250K+ AI 工作流 | 截至 mid-2026 | FeaturedCustomers | 中 | 第三个仍未对齐的 Baird 指标 |
这里刻意并列表径不可比的指标(用户、查询、工作流、参与度 %),因为已抓取来源之间没有一个对齐后的采用指标;参见链接的 Baird 指标对账证据缺口。
[CU009, CU010, CU011, CU012, CU013, CU014]平均 8 个月销售周期和共同合规建设先于驻场部署导入;之后,具名账户要么显示可衡量参与度(Baird),要么仍只停留在客户标识层级(多数其他具名银行)。
[CU038, CU041, CU046, CU047, CU015, CU045]6.3 具名客户证据:标识、Baird 与直接声音
本章最需要区分标识级营销与已验证生产使用。Rogo 客户页原始标记确认了六个逐一标注的标识——Jefferies、Lazard、Moelis、Nomura、Rothschild & Co 和 Truist;一篇独立融资报道又把其中五个名字列为活跃用户,这比单独一页营销页面更有佐证力。Jefferies 跨过了更高门槛:发布访谈的是 Jefferies 自己的公司网站,而不是 Rogo;访谈中 Rogo CEO 描述了在真实交易数据上的生产部署,因此属于独立域名证据,而不是供应商控制的展示。Truist Securities CEO 也单独具名引用,称赞集成、生产率和降风险结果;不过 Truist Ventures 同时是 Series C 投资方,重复了前几章已在 J.P. Morgan 身上标出的投资方兼客户重叠模式。GTCR 是唯一具名私募股权账户,只在融资报道中顺带出现,而非专门故事。Anthropic 和 Google Cloud 两个伙伴案例研究补充了间接信号:Google Cloud 自己的案例研究称,Rogo 客户曾私下表示,出于安全原因更信任 Google 而非其他 AI 供应商;这是通过供应商而非客户直接传来的罕见终端客户供应商情绪。相对于 250+ 家机构的声称基数,本章所有抓取来源里独立点名或引用的账户总数为 8 个;尽调叙事中,单靠标识墙不应承担过多证据重量。[CU017, CU018, CU019, CU020, CU021, CU022]
| 客户 | 细分 | 部署 / 使用场景 | 生产部署 / 试点 | 结果 / 证据 | 限制 |
|---|---|---|---|---|---|
| Baird Equity Research | 股票研究(银行) | 业绩纪要整合、电话会文字稿处理、覆盖笔记起草、共享项目文件夹 | 生产部署(具名案例研究) | Rogo 称 100+ 活跃用户、~85% WAU / ~70% DAU;独立综述另称每周 10,000+ 工作流、95% 参与度 | 不同来源的指标无法对齐成一个数字 |
| Jefferies | 头部投行 | GIB 工作流中,AI 处理实时交易数据 | 生产部署(Logo 加 Jefferies 主办的 CEO 访谈) | Jefferies 自有网站给出独立域名佐证,不只是 Rogo 营销材料 | 未披露使用比例或席位数 |
| Lazard | 头部投行 | 交易工作流(根据 Logo 位置和创始人的 Lazard 背景推断) | Logo 级(具名,未独立量化) | 出现在官方客户墙,也出现在独立融资报道中 | 未披露案例研究、引述或使用数据 |
| Moelis | 头部投行 | 交易工作流(根据 Logo 位置推断) | Logo 级 | 出现在官方客户墙,也出现在独立融资报道中 | 未披露案例研究、引述或使用数据 |
| Nomura | 全球投行 | 交易工作流(根据 Logo 位置推断) | Logo 级 | 出现在官方客户墙,也出现在独立融资报道中 | 未披露案例研究、引述或使用数据 |
| Rothschild & Co | 全球顾问型投行 | 交易工作流(根据 Logo 位置推断) | Logo 级 | 出现在官方客户墙,也出现在独立融资报道中 | 未披露案例研究、引述或使用数据 |
| Truist Securities | 接近头部投行层级的银行 | 交易工作流,并聚焦银行家产能和客户关系 | 生产部署(具名 CEO 引述) | CEO 公开引述提到生产率、风险降低和产能提升 | Truist Ventures 也是 Series C 投资方,存在投资方兼客户关系 |
| GTCR | 私募股权 | 交易和投后组合工作流(根据融资轮报道推断) | Logo 级(仅在融资报道中具名) | 已抓取来源中唯一具名的私募股权细分账户 | 未披露案例研究、引述或使用数据 |
覆盖范围有限:本章已抓取来源中,只有这些客户被具名或引用;Rogo 和独立报道给出的机构总数超过 250 家,参见 enumerationScope。
[CU017, CU018, CU019, CU020, CU021, CU030]只有 Baird 同时具备相对独立的佐证、生产环境状态和量化使用度;多数具名 logo 仍未量化,最强的负面信号来自匿名来源。
[CU015, CU018, CU019, CU020, CU027, CU031]6.4 留存、参与度与持久性证据
抓取到的来源没有披露 Rogo 的净收入留存、总收入留存、客户标识流失率或平均合同期限,因此本节只能围绕代理指标和一个真实负面数据点展开,而不是正式留存经济性。Baird 约 85% 周活、70% 日活,仍是唯一绑定具名账户的量化重复使用数字;放在受监管金融机构内部的企业软件语境下,这个数字有意义。不过最具体的持久性信号来自一条匿名 Wall Street Oasis 论坛帖:一名发帖人称其公司使用约一年后从 Rogo 流失,并称这是自己用过的最过度炒作的产品;同一帖子里,其他发帖人则说 Felix 发布后平台改善很多,并称同事每天依赖它。另有一名董事级发帖人称,该工具在副总裁级别以上基本没用;这符合一种模式:初级和中级分析师是最重度用户,资深银行家更怀疑。上述内容都无法绑定到具体具名公司,因此只能作为方向性情绪,而不是已验证流失统计;但它是 Rogo 账户使用一年后会发生什么的唯一公开证据,无论正面还是负面。[CU026, CU027, CU028, CU029, CU030, CU031]
| 指标 | 数值 / null | 细分 | 置信度 | 尽调要求 |
|---|---|---|---|---|
| 净留存率(NRR) | null - 未披露 | 全部 | 低 | 向公司或投资方索要 NRR |
| 总留存率(GRR)/ Logo 流失率 | null - 未披露 | 全部 | 低 | 索要 Logo 级流失历史 |
| 平均合同期限 / 续约节奏 | null - 未披露 | 全部 | 低 | 索要标准 MSA 条款和续约日期 |
| Baird 周活 / 日活使用 | ~85% WAU / ~70% DAU | 股票研究(Baird) | 中 | 确认发布后这一水平是维持还是变化 |
| 单一账户流失轶事 | 1 个匿名账户称 ~1 年后流失 | 未说明(论坛发帖者) | 低 | 无法核验到具名公司;仅作为方向性风险信号 |
| 按资历分化的满意度 | 分化:初级 / 中级用户称日常依赖;一名 director 级发帖者称 VP 以上「完全没用」 | 未说明(论坛发帖者) | 低 | 调查资深银行家的采用是否结构性落后于初级银行家 |
除 Baird 的 WAU/DAU 和匿名论坛轶事数据点外,其他量化行都是 null,因为已抓取来源没有披露正式留存经济性;这些 null 是刻意保留,并非遗漏。
[CU026, CU027, CU028, CU029, CU030, CU031]6.5 扩张路径、集中度风险与负面信号
Rogo 的扩张叙事穿过它具名的大型投行标识:独立评论明确称,中小型和中端市场银行正在获得「此前只有大型投行机构独享」的分析能力,并把这些头部账户用作参考证据。同样集中在少数具名标识上,也构成风险:如果 Jefferies、Lazard、Moelis、Nomura 或 Rothschild & Co 中任意一家流失或降级,帮助精品银行销售的信誉光环可能被不成比例地削弱。供给侧集中度从另一角度加重了这一图景。一篇 2026 年融资报道把 Rogo 自身依赖第三方基础模型提供商列为公司层面风险;另一篇复盘报道则从客户角度提出镜像风险:关键交易工作流过度依赖 Rogo 这一个供应商。再往上叠一层,Rogo 面向客户的产品同时依赖与 LSEG、PitchBook 和 Daloopa 的数据授权合作,而这些都不是 Rogo 完全可控。初次对话到生产部署平均八个月的销售周期被描述为行业标准,这会拖慢新增客户增长,也给既有终端提供商——Bloomberg、FactSet、S&P Capital IQ——留出时间,把竞争性 AI 功能叠到自己的既有数据优势上。Truist 和 J.P. Morgan 的投资方兼客户双重关系,又增加了推荐偏差问题;尽调团队应专门寻找非投资方客户做客户访谈来控制这一偏差。[CU032, CU033, CU034, CU035, CU036, CU037]
| 扩张驱动 | 集中度风险 | 影响 | 尽调路径 |
|---|---|---|---|
| 前置部署的银行家模型降低新头部投行账户的切换摩擦 | 同一模型依赖一支规模小、难扩张的前银行家部署团队 | 增长可能超过前置部署团队的产能,导致新账户无法享受同样的白手套导入流程 | 索要前置部署人数与机构账户数的对比 |
| 多模型架构(OpenAI、Anthropic、Google Gemini)降低 Rogo 自身的供应商锁定 | 采用 Rogo 的机构会在关键交易工作流上依赖单一供应商 Rogo | 若 Rogo 宕机、涨价或质量回落,客户交易时间表可能受冲击,交易中途也很难替换 | 向 Rogo 索要业务连续性和多供应计划 |
| 数据合作伙伴集成(LSEG、PitchBook、Daloopa)扩大共享客户的工作流覆盖 | Rogo 的客户价值主张取决于同时维持三项数据授权关系 | 任何一个数据合作丢失或重谈,都可能削弱面向客户的产品;结果不由 Rogo 完全控制 | 确认每项数据合作的最低合同期限和续约日期 |
| 借头部投行背书向精品投行 / 中端市场扩张 | 背书基础仍集中在少数具名头部投行 Logo(Jefferies、Lazard、Moelis、Nomura、Rothschild & Co) | 若一个标杆 Logo 流失或降级,精品投行销售所依赖的信誉光环可能被不成比例地削弱 | 跟踪五到六个具名标杆 Logo 的续约状态,将其作为领先指标 |
| 投资方兼客户关系(Truist、J.P. Morgan)可加速采购 | 同一重叠也会带来背书偏差和优惠准入问题,影响客户证据可信度 | 外部尽调无法彻底拆分这些账户的真实产品市场契合与投资方善意带来的采用 | 安排非投资方客户访谈,以控制这一偏差 |
| 平均八个月从 POC 到生产部署,说明企业销售动作一旦启动可重复 | 同样的长周期会拖慢扩张,也给现有终端(Bloomberg、FactSet、S&P Capital IQ)留下上线竞争性 AI 功能的时间 | 新增账户增长放慢,可能压缩当前采用叙事支撑的估值倍数 | 跟踪机构 Logo 净新增的环比变化 |
每行把扩张驱动与镜像集中度风险配对,而非分开列示,反映同一机制在这组证据中如何双向作用。
[CU032, CU033, CU034, CU035, CU036, CU037]6.6 部署模型、采购动作与合规摩擦
Rogo 的 GTM 动作明确是高接触,而不是自助式。其 2026 年 4 月 Series D 公告称,新资本将投入「更多嵌入我们服务机构的前线部署银行家和工程师」;一篇独立复盘报道更详细描述了前线部署银行家模型:曾任金融从业者的人进入合作机构现场推动采用,并被明确类比为 Palantir 的前线部署工程师。Rogo 自己的 Baird 案例给出最清晰的具体例子:在 Baird 密尔沃基总部进行线下一对一会议,同时配合答疑时段和面向单个用户的直接触达。OpenAI 伙伴案例研究描述了类似的「由前银行家和投资人组成的部署团队」,与客户实时打磨功能;Anthropic 客户故事则引用 Rogo 自己产品负责人的说法,称可用输出被验证后,用户情绪从怀疑转向兴奋——这是变更管理模式,不是即时采用模式。高接触动作之下是真实合规摩擦:Jefferies CEO 主持的 Rogo 创始人访谈称,在 AI 触碰实时交易数据前,双方共同搭建信息隔离墙、MNPI 控制和基于角色的访问基础设施;Deloitte 2026 年银行智能体风险指引也独立描述了受监管银行在扩展任何智能体部署前应建立的智能体登记、审计轨迹和披露层。Rogo 自己的信任中心发布正式的可接受使用、访问控制和资产管理政策,作为采购材料来缩短审查;这与本章其他地方描述的约八个月 PoC 到生产销售周期一致。[CU003, CU040, CU041, CU043, CU046, CU047]
| 维度 | 证据 | 来源立场 | 尽调缺口 |
|---|---|---|---|
| 前置部署银行家导入 | 售后由前银行家和工程师直接嵌入客户机构 | 证实(公司 + 独立综述) | 未披露部署人员与活跃机构账户的比例 |
| 现场手把手推广先例 | Baird 推出时包括在 Baird Milwaukee 总部的一对一线下培训、答疑时段和直接触达 | 证实(公司案例研究) | 不清楚这种白手套导入是标准动作,还是只给旗舰账户 |
| MNPI / 信息隔离墙合规建设 | Jefferies 访谈描述了双方为接入实时交易数据共同建设合规基础设施 | 证实(客户主办访谈) | 未披露该合规建设的时间线或成本 |
| 正式信任 / 安全材料 | Trust Center 发布 Acceptable Use、Access Control 和 Asset Management 政策,供采购审查 | 证实(公司) | 本章来源未引用独立审计或认证机构确认 |
| 销售周期长度 | 从初次沟通到生产部署大约八个月,被描述为行业标准 | 中性(独立分析师) | 未披露围绕这一均值的分布(最快 / 最慢) |
| 从业者对成熟度和可靠性的情绪 | 匿名论坛发帖者一方面称产品是 Felix 之后快速改进的工具,另一方面称它被过度吹捧,有些公司已经流失 | 反向与证实(混合,匿名) | 无法对应到具体具名公司,也无法独立核验 |
该表单独列出购买动作和合规摩擦证据;细分、采用和集中度表只简要引用这些证据。由于没有量化的留存 / 重复使用队列时间序列数据可画图,它替代了该图。
[CU040, CU041, CU043, CU046, CU047, CU048]6.7 图表
07风险
7.1 产品与输出风险——幻觉、引用粒度与规模化
Rogo 的核心暴露点不是头条准确率,而是推导质量。BigFinanceBench 是一个 928 项、由专家编写的基准,2026 年 6 月由 Rogo 自有框架团队共同发布在 arXiv;它用 36,241 个评分点检查金融推导的每一步——来源选择、期间和会计定义选择、假设、计算——而不只看最终数字,并发现十个当前前沿和开放权重智能体中,表现最佳的系统也只拿到可用评分点的 58.8%;最终答案准确率只是推导质量的「有损代理指标」[CR001]。落到 Rogo 自己的 Felix 智能体上,这意味着一个可直接交付客户的可比公司集、模型或备忘录,可能看起来完整、引用充分,却仍埋着错误期间、误用的会计定义或未说明假设;审阅人员必须抓住这些问题,而这种工作流错误风险会随 Felix 为每名分析师生成的演示文稿、模型和备忘录数量一起放大 [CR002]。 引用粒度会进一步放大这一问题:Rogo 推销审计轨迹和带来源研究,但引用粒度(页级、段落级、数据集级)决定了银行家能否在数字进入实时交易文件前真正核验;公开资料中没有独立且 Rogo 专属的引用到来源精度基准 [CR003]。品类层面的先例很明确:FINRA 的 2026 Annual Regulatory Oversight Report 将「幻觉」——AI 生成的不准确或误导性信息——列为会员机构的首要 GenAI 风险,并警告说,把 GenAI 作为监督系统一部分的机构,必须权衡「AI 模型的完整性、可靠性和准确性」[CR004]。报道同一报告的行业媒体也把监管者立场概括为:日常运营中「敦促机构警惕幻觉风险」[CR005]。 文档级和组合级规模处理是第二个相关暴露点。Felix 的定位是同时摄入数据室、多年申报文件和机构私有资料库;独立技术报道把产品描述为一套跨多个前沿模型路由的编排框架,而不是单个微调模型。该设计用一致性换灵活性,也把可靠性工程推到 Rogo 自己的评估框架上,而不是供应商认证基准上 [CR006]。金融之外,最刺眼的警示先例是 Deloitte Australia's AU$440,000 政府报告:2025 年末,外部学者发现其中有捏造参考文献,以及一段由未披露 GPT-4o 工作流生成、交付前未被发现的 Federal Court 判决虚构引文,Deloitte 因此不得不部分退款 [CR007]。该事件不涉及 Rogo,但它是最清晰的公开证据:一家有声誉的专业服务公司,自己的内部审阅流程没能抓住付费客户交付物中的 AI 捏造内容。这正是 Rogo 的前线部署模型试图用人类银行家在环来防止的失败模式,也正是尽调应直接测试、而不是从营销中推断的失败模式 [CR008]。另一个领域也在重复同一模式:2026 年法院针对 AI 幻觉法律引用的制裁潮中,仅 2026 年第一季度,美国因捏造判例法相关罚款就至少 $145,000;这是另一个高风险、依赖引用的专业领域 [CR009]。
| 故障模式 | 可能性 | 严重性 | 缓解成熟度 | 残余风险 | 未解决缺口 |
|---|---|---|---|---|---|
| 面向客户的模型 / 备忘录出现推导层面幻觉(期间、假设或会计定义错误) | 中 -- 品类基准显示,即便最佳智能体也有 41.2 个百分点的评分差距 | 严重 -- 输出进入实时交易决策 | 中 -- 引用系统和审计轨迹存在;未发现独立的 Rogo 专项准确率审计 | 中高 | 没有公开的 Rogo 专项推导准确率基准,可与 BigFinanceBench 对照 |
| 引用颗粒度不足,银行家使用前难以核验 | 中 -- 颗粒度水平未被独立记录 | 高 -- 削弱核心「可审计」价值主张 | 未知 -- 未发现独立测试 | 中高 | 未披露页码 / 段落级引用精度指标 |
| 文档 / data-room 规模处理失败(遗漏或错误加权大型资料库中埋藏的事实) | 中 -- 任何摄取大型非结构化语料的 LLM 编排层都有这一内生风险 | 高 -- 尽调中漏掉一项负债或契约条款,会直接带来交易风险 | 中 -- 编排层按基准表现把任务路由到不同模型 | 中 | 未披露面向 data-room 规模任务的大语料召回 / 精确率基准 |
| 未披露的平台宕机或事故影响实时交易工作流 | 未知 -- 公开事故历史没有正反证据 | 若发生在活跃交易期间则为高 | 未知 -- Trust Center 提到 BC/DR 和事件响应政策文件,但未独立发布 | 中 | 缺少经独立核验的 uptime / 事故记录;本章来源未由第三方佐证状态历史 |
| 安全 / 合规声明(SOC 2、ISO 27001、ISO 42001、EU AI Act)依赖 Trust Center 自报披露 | 中低 -- 认证项目真实存在,但证据由供应商发布 | 中等 -- 夸大合规姿态本身会构成接近「AI washing」的敞口 | 中 -- 具名参考客户(Jefferies、Nomura、Lazard、Moelis)据报道审查过该姿态 | 中 | 已审阅来源中未发现独立审计师确认函或公开证书注册条目 |
可能性 / 严重性是作者基于品类基准(BigFinanceBench)、FINRA / 法律行业幻觉报告和 Rogo 自身 Trust Center 披露作出的定性评估;未发现独立的 Rogo 专项事故或准确率审计。
[CR001, CR002, CR003, CR006, CR007, CR009]可能性和影响是作者基于监管指引、品类基准以及媒体 / 分析师评论作出的定性判断;并非来自量化建模。
[CR001, CR004, CR015, CR021, CR025, CR038]7.2 监管与合规风险——SEC、FINRA、欧盟 AI 法案与诉讼姿态
美国证券监管机构已经从讨论 AI 风险转向直接执法。自 2024 年 3 月以来,SEC 已因夸大或误导性的「AI washing」表述,对投资顾问、上市发行人以及至少一名创业公司创始人采取执法行动;现在,SEC 要求公司像论证业绩或《投资顾问法》营销规则下的风险因素披露一样,严谨支撑 AI 相关表述 [CR010]。2025 年 12 月,SEC 投资者咨询委员会投票建议要求发行人定义「人工智能」、披露董事会对 AI 部署的监督,并单独报告内部和面向消费者的 AI 影响——如果这套草案标准落地,Rogo 的银行和资产管理客户(若 Rogo 接近公开上市,也可能包括 Rogo 自身)描述 AI 驱动结果的门槛都会被抬高 [CR011]。修订后的 Regulation S-P 也要求券商记录对第三方 AI 供应商的监督,包括供应商 72 小时内通报数据泄露、30 天内通知客户;无论底层责任在哪里,AI 供应商风险管理(包括 Rogo)都会被直接纳入券商合规流程 [CR012]。 FINRA 2026 年监管报告重申,其规则「技术中立」:监督(Rule 3110)、通信、记录留存和公平交易义务适用于 GenAI 辅助输出,强度与人工输出相同;公司必须归档提示词、输出和模型使用日志以备审计 [CR013]。对 Rogo 的银行客户来说,Felix 生成的研究和模型本身也要接受监督审查和记录留存,这一运营负担可能拖慢采用,或要求 Rogo 额外提供审计工具 [CR014]。欧盟侧,AI Act 的 GPAI Code of Practice 已于 2025 年 8 月 2 日生效,执法权(信息请求、模型访问、召回)从 2026 年 8 月 2 日开始;用于信贷、AML 等金融工作流的 AI 高风险系统完整义务也从 2026 年 8 月 2 日起可执行——这是一条硬合规期限,正落在本轮尽调窗口内 [CR015]。Rogo 自身 Trust Center 将「EU AI Act」列入合规文档,但该页面由公司自行发布,本章审阅的任何来源都没有具名审计机构独立验证 [CR016]。 联邦银行监管机构在模型风险上走向相反方向:OCC、Federal Reserve 和 FDIC 2026 年 4 月的跨机构指引废止了 2011 年 SR 11-7 时代框架,改用更轻的原则导向方法,并明确把生成式和 agentic AI 模型排除在范围外,等待未来另行指引——这让 Rogo 的银行客户今天还没有稳定的监管答案,无法判断该如何验证供应商提供的 agentic 研究工具 [CR017]。保密性是另一类独立风险:投行业务经常涉及重大非公开信息(MNPI),如果训练、留存或跨客户数据流不够严密,把交易数据输入第三方 AI 供应商会带来合同风险,也会贴近内幕交易风险——监管机构目前通过 Reg S-P 供应商监督规则处理这类风险,而不是 AI 专门规则,这意味着举证责任落在 Rogo 的数据处理架构上,而不是某一条具名法规上 [CR018]。Rogo 也贴近「研究工具」与「投资建议」的边界:一个为真实交易生成可比公司、模型和建议的工具,如果公司合规程序没有清楚记录客户建议的依据是人工判断而非 Felix 输出,就可能触发顾问受托责任和适当性义务 [CR019]。 本章审阅的来源没有发现点名 Rogo 的未决诉讼、SEC 或 FINRA 执法行动,或已确认数据泄露 [CR020]。这种沉默应当谨慎解读,不能当作干净证明:AI 相关证券集体诉讼从 2023 年到 2024 年同比增长约 100%,并在 2025 年至 2026 年继续增加;法律行业追踪机构预计,到 2026 年,「事件驱动型」AI 诉讼仍会是私人证券诉讼中的主导类别,背后既有已建立 AI 专门能力的原告律师团队,也有 SEC Cybersecurity and Emerging Technologies Unit 施加的并行压力——这是一条快速扩张的案卷通道,像 Rogo 这样高速增长、能见度很高的 AI 供应商,只需一次有争议的声明或事件就可能被卷入 [CR021]。
| 规则 / 许可 / 案件 | 司法辖区 | 状态 | 可能性(3Y) | 严重性 | 缓解措施 | 残余风险 | 尽调问题 |
|---|---|---|---|---|---|---|---|
| SEC「AI washing」执法(Advisers Act Marketing Rule) | 美国联邦(SEC) | 活跃 -- 自 March 2024 起已针对顾问、发行人和一名创始人采取执法行动 | 中(3 年内 35%) | 高 -- 罚款、个人责任、声誉损害 | Rogo / 客户避免夸大模型能力;为声明提供依据 | 中 -- 主要适用于客户披露,间接影响 Rogo 自身营销 | 索要 Rogo 内部 AI 能力声明依据文件和营销审查流程 |
| FINRA GenAI 监督预期(Rule 3110 及相关规则) | 美国联邦(FINRA) | 活跃 -- 2026 Annual Regulatory Oversight Report 将幻觉和记录保存列为重点领域 | 高(3 年内 65%) | 中等 -- 监督摩擦、银行客户审计日志负担 | Rogo 审计轨迹 / 引用系统;客户侧归档提示词和输出 | 中 -- 合规成本一部分转移给银行客户,一部分落到 Rogo 工具上 | 索要 Rogo 的提示词 / 输出留存架构和客户审计日志导出 |
| EU AI Act -- GPAI 义务和高风险金融用途执法 | 欧盟 | 活跃 -- GPAI 规则于 Aug 2 2025 生效;执法权和高风险义务自 Aug 2 2026 起适用 | 高(3 年内 70%) | 高 -- 最高罚款为全球营业额 3%(GPAI)/ 7%(一般法案) | Rogo Trust Center 列出 EU AI Act 合规文档 | 中高 -- 自报合规未在已审阅来源中独立核验 | 索要 Rogo 的 EU AI Act 符合性评估,以及任何第三方审计证明 |
| Federal Reserve / OCC / FDIC 模型风险管理指引 | 美国联邦(银行监管机构) | 活跃 -- April 2026 指引撤销 2011 SR 11-7 框架,并明确将 GenAI / agentic AI 排除在外,等待未来指引 | 中高(3 年内 55%) | 中等 -- 银行客户对 Felix 这类智能体工具缺少确定的 MRM 答案 | 尚无 Rogo 专项措施;预期未来会有 AI 专项跨机构指引 | 中 -- AI 专项指引发布后,监管缺口可能突然收紧 | 询问 Rogo 的银行客户目前如何按内部模型风险政策归类 Felix |
| SEC Regulation S-P 供应商监督(broker-dealer 第三方 AI 供应商) | 美国联邦(SEC) | 活跃 -- 修订规则的合规截止期为大型 broker-dealer Dec 2025、小型 broker-dealer June 2026 | 高(3 年内 60%) | 中高 -- 72 小时供应商泄露通知、需记录在案的尽调 | Rogo Trust Center:SOC 2 I/II、ISO 27001、ISO 42001、渗透测试和 DFD 报告可按需提供 | 中 -- 随 250+ 机构规模化,每个客户关系都会把审计负担转向 Rogo | 索要 Rogo 标准 Reg S-P 供应商监督尽调包和泄露通知 SLA |
| 第三方 AI 供应商处理 MNPI / 保密交易数据 | 美国联邦(证券法、合同) | 持续义务 -- 无已知违规;主要由 Reg S-P 和银行供应商合同覆盖,而非 AI 专项规则 | 中低(3 年内 20%) | 高 -- 类内幕交易风险、合同终止、声誉损害 | 加密、访问控制、不用客户数据进行跨客户模型训练(据 Trust Center) | 中 -- 依赖 Rogo 自报架构;未发现独立审计 | 索要 Rogo 的数据隔离和排除模型训练合同条款 |
| 点名 Rogo 的诉讼 / 执法行动 | 美国联邦 / 州 | 截至本次运行,无已知备案 | 低(3 年内 15%) | 一旦发生则为严重 -- 对合规敏感买方群体造成声誉和客户留存损害 | 无专项措施;AI 证券诉讼和 SEC CETU 活动均在增长 | 当前低、但随可见度上升 -- AI 相关证券集体诉讼在 2023-2024 年同比增长约 100%,并在 2026 年继续增长 | 每季度监控 SEC litigation releases 和 PACER,查找任何点名 Rogo 的备案 |
枚举范围有限:仅覆盖本章已审阅的监管机构、律所和新闻来源中可识别的监管与法律风险向量;未披露合同条款、内部合规发现或非公开监管问询不在可枚举范围内。
[CR010, CR011, CR012, CR013, CR015, CR018]7.3 竞争与平台依赖风险
Rogo「编排层而非模型」的架构,确实能降低单一供应商过时带来的冲击,但它没有消除依赖,只是把依赖分散到了 OpenAI、Anthropic 和 Google;具体任务由哪一家承接,取决于 Rogo 内部基准测试当时的胜出者 [CR022]。如果任何一家实验室大幅提高 API 定价、限制金融行业访问,或推出直接竞争的金融产品,Rogo 的成本结构和差异化会同时收窄。Rogo 的数据层同样集中在少数持牌供应商手里:LSEG(2025 年宣布战略合作,把基本面、共识预期和 M&A 数据接入平台)和 PitchBook(2025–2026 年深化合作,覆盖私募资本交易、基金和投资者数据)在 Rogo 和供应商自身表述中,都是 Rogo 数据战略的核心且不断扩大的支柱;续约条款、定价和排他性均未公开 [CR023]。数据供应商正越来越多地把自己的生成式 AI 层直接嵌入终端;LSEG、PitchBook、FactSet 或 S&P 任何一方若为了自有 AI 功能而限制批发 API 访问,都会直接削弱 Rogo「基于授权数据」的价值主张 [CR024]。 Microsoft 同时是 Rogo 最大的战略伙伴,也是最可信的平台威胁。Rogo 宣布了 Microsoft 合作并交付原生 Excel 插件,但 Microsoft 自己的 2026 年路线图会把能力越来越强、面向金融的 Copilot agents——方差分析、DCF 构建、可复用金融「skills」、连接 FactSet / PitchBook / S&P 级数据的实时连接器——从 2026 年 10 月起直接并入 Microsoft 365 Copilot 核心订阅,且不再额外按席位收费;同时还会推出面向第三方和第一方金融 agents 的 Copilot Agent Store [CR025]。Rogo 目标银行和资产管理机构大多已经是 Microsoft 365 客户,因此这种捆绑会降低 Rogo 至少一部分可商品化工作流的切换启动成本(方差解释、标准模型搭建、数据连接器),即便它尚未达到 Rogo 前置部署、按公司深度配置的程度 [CR026]。独立竞争者评论把 Rogo 和 Hebbia 列为引用联动、可审计金融研究的两家领先者;新进入者(o11、F2)则明确批评二者仍是「浏览器孤岛」工具,需要把数据搬进搬出原生 Office 应用,而不是在原位操作 Excel 单元格或 Word 文档——随着 Microsoft 从 Office 内部缩小同一差距,这种工作流摩擦可能更重要 [CR027]。 资本提供方集中度较轻,但仍是真实依赖:Rogo 的 $160M Series D 由 Kleiner Perkins 领投,Sequoia、Thrive Capital、Khosla Ventures 和 J.P. Morgan Growth Equity 继续参投。这个财团覆盖顶级后期投资者,但仍集中在少数重复支持者;在盈利能力得到独立证明之前,Rogo 能否再融一轮,依赖这些投资者继续相信公司 [CR028]。银行客户同时也是战略投资者的结构(J.P. Morgan 参投,而 J.P. Morgan 这家银行本身也在自建 AI 工具)还带来治理问题:如果 Rogo 最大的银行客户加速自研,投资者与客户激励是否会分化 [CR029]。最后,Rogo 通过收购增长的策略(Subset、Offset、Plux)把整合风险集中在一个小型内部 M&A 团队;每笔收购都会新增独立代码库、数据访问模式和创始人留任问题,叠加在平台依赖图景之上,而不是分散该风险 [CR030]。
| 依赖 | 交易对手 | 作用 | 集中度 | 失败情景 | 严重性 | 缓解措施 | 残余风险 |
|---|---|---|---|---|---|---|---|
| 前沿模型访问 | OpenAI / Anthropic / Google | Rogo 编排层按任务路由底层模型 | 高 -- 没有自有专有基础模型 | 涨价、限制访问,或某家实验室推出竞争性金融产品 | 高 | 模型无关的框架可在供应商之间切换 | 中——切换成本低于依赖单一模型的同业,但并非为零 |
| 授权金融数据 | LSEG、PitchBook(FactSet/S&P 集成的公开记录较少) | 研究、可比公司和私募市场工作流的核心数据锚点 | 高——数据供应商条款和排他性未披露 | 供应商限制批量 API 访问,转向第一方 AI 功能 | 高 | 多个具名数据供应商合作,而非单一来源依赖 | 中高——续约 / 定价条款未公开 |
| 云和生产力平台 | Microsoft(Azure/365 生态、Excel 插件) | 分销伙伴和 Excel 原生集成 | 中高——同时也是最可信的捆绑竞争者 | Microsoft 自 2026 年 10 月起借 Copilot Agent Store 捆绑,缩小工作流差距 | 高 | 前置部署银行从业者做深度、机构定制配置;收购 Subset 增加建模深度 | 中高——可商品化的工作流切片暴露最大 |
| 后期资本 | Kleiner Perkins(领投)、Sequoia、Thrive Capital、Khosla Ventures、J.P. Morgan Growth Equity | Series D 轮及此前轮次融资 | 中——顶级重复投资方分散持有,未披露单一主导持有人 | 若未披露盈利能力,持续融资需要投资人继续有信心 | 中 | 多家资金充足的投资方跨轮次持续跟投 | 中——增长放缓时融资风险上升 |
| 被收购公司整合 | Subset、Offset、Plux AI 创始人和代码库 | 电子表格智能体、模型维护和欧洲覆盖能力 | 中——三项整合同步推进 | 创始人离职或整合失败,会拆散该框架最新能力 | 中 | 收购资产迄今保留为具名产品线(Rogo Agents、Offset 工作流) | 中——整合记录尚未经过完整产品周期的公开检验 |
集中度和严重性评级基于公开合作公告以及竞争者 / 分析师评论;具体合同条款、排他性条款和股权结构表中的持股比例未公开。
[CR022, CR023, CR024, CR025, CR026, CR028]依赖关键性是作者基于公开合作公告和竞品 / 分析师评论作出的定性判断。
[CR022, CR023, CR028]7.4 客户、采购与参考偏差风险
Rogo 的参考客户也是它的集中风险。公开点名的生产环境用户——Baird Equity Research、Jefferies、Rothschild & Co、Nomura、Moelis,以及 Rogo 自身 Trust Center 列出的 Lazard 等参考机构——正是成长阶段供应商需要的旗舰客户标识;但一个卖给不足数百家大型金融机构的平台,结构上会暴露于任何单一头部账户流失、不续约或公开不满的风险。本章审阅的独立来源没有披露 Rogo 的单客户收入集中度、净收入留存或合同期限 [CR031]。投行和资管的企业 AI 采购也是长周期、多利益相关方流程——把法律、合规、风险和 IT 安全签核算进去,从首次接触到签约通常需要 9–24 个月——这会放慢 Rogo 把管线转为收入增长的速度,使其难以匹配估值轨迹暗示的节奏,也给 Bloomberg、FactSet、Microsoft 等在位厂商更多时间在竞争评估结束前补齐功能差距 [CR032]。 参考偏差是另一类更难识别的风险:Rogo 公开可见的赞誉几乎全部来自公司自己的客户页面、赞助合作公告和融资新闻;独立、非邀约评价只有很薄、低频的一层(PeerSpot 的 Rogo 条目读起来更像供应商式描述文案,而不是具名用户投诉;本轮未能独立验证 Rogo 的 Trustpilot 档案)[CR033]。找到的最清晰非邀约、从业者层讨论来自 Wall Street Oasis 一个帖子:一家中型银行产品团队向同行询问对 Rogo、Hebbia 和 ModelML 的真实看法,并把自己的评估标准明确表述为「准确性、自动化工作流、工具不能只是噱头」——这种措辞默认对供应商营销主张保持怀疑,也说明即便是 Rogo 自己的目标买家,也不会把「AI 准确性」主张视为不证自明 [CR034]。竞争者发布的比较内容(Hebbia、o11、F2)对功能定位有方向性参考价值,但不是独立证据;任何准确性或可靠性主张都应相应打折 [CR035]。 采购和集中风险也穿过 Rogo 自己的供应商栈:任何大型银行客户若执行 FINRA 和修订后 Regulation S-P 现在要求券商开展的同类第三方 AI 供应商监督尽调,都需要 Rogo 按需提供审计日志、模型变更文档和事件响应证据。即便没有事件发生,这也是 Rogo 在 250+ 家机构关系中持续承担、且不断增长的合规成本 [CR036]。净效果是,Rogo 当前增长叙事更多建立在采用广度(35,000+ 用户、250+ 机构)上,而不是披露深度(留存、扩张或按席位经济性)上;本报告客户章节也标出这个缺口,它直接影响尽调流程应给客户标识驱动的增长主张多大权重 [CR037]。
7.5 估值与资本市场风险
Rogo 的估值上涨速度,快过任何能支撑该估值的基本面公开披露。报道显示,公司从 2026 年 1 月 Series C 约 $750M 估值,到 2026 年 4 月 Series D $2B 估值,只用了三个月,接近翻三倍;该轮由 Kleiner Perkins 领投,Sequoia、Thrive Capital、Khosla Ventures 和 J.P. Morgan Growth Equity 参投 [CR038]。本章或财务章节审阅的来源,都没有披露 Rogo 经审计收入、毛利率、净收入留存或盈利能力;公开叙事依赖用户数(35,000+)和机构数(250+),而不是投资者可用来校验隐含收入倍数的 ARR 或增长率 [CR039]。在 $2B 估值下,这个缺口只会更重要,因为更广泛的 AI 融资市场已经出现修正迹象:截至 2026 年 1 月,多位科技行业分析师和投资者正在积极争论 AI 行业估值是否构成泡沫,依据包括每年约 $400B AI 投资与规模小得多的企业生产率回报之间脱节,以及调查显示多数企业迄今报告 AI 部署没有带来可衡量生产率改善 [CR040]。 如果 Rogo 增速低于其融资节奏暗示的水平,或更广泛的 AI 估值修正压缩可比私募倍数,Rogo 将暴露于降价轮,或 IPO / 退出估值相对最近一级市场价格大幅降低的风险——2025–2026 年其他后期 AI 相关公司重估中已经能看到这种模式 [CR041]。AI 相关证券诉讼风险(第 2 节)会进一步放大该问题:原告律所和 SEC Cybersecurity and Emerging Technologies Unit 都在加大关注 AI 驱动增长叙事与披露基本面之间的脱节;如果 Rogo 在 AI 驱动收入和留存主张仍大体未披露时寻求上市,这一动态会直接适用于 Rogo [CR042]。在 Rogo 发布(或被要求发布,例如进入 S-1 阶段)经审计收入、留存和利润率数据之前,任何对 Rogo 的估值判断都应视为暂定,并被这一披露缺口重度限定 [CR043]。
因果传导链为定性判断;相对严重性体现在节点标签中,而非实测概率。
[CR002, CR021, CR025, CR038, CR041, CR042]7.6 人才与运营扩张风险
Rogo 的商业化推进依赖前置部署投行人员(以及前置部署工程师,FDE)模式:人员嵌入客户,把公司特定工作流翻译成产品配置。这种混合技能组合——工程流利度、深厚客户领域判断、面向客户沟通能力——如今是科技行业最紧的人才市场之一:FDE 职位发布量在 2025 年 1 月至 9 月增长约 800%,到 2026 年初同比增长约 1,165%;候选人池只增长约 50%,头部 AI 实验室为最抢手画像开出的总薪酬超过 $500K [CR044]。Rogo 争夺这类人才的对手是 OpenAI、Anthropic、Google 和 Palantir,而这些公司也在激进招聘 FDE;因此,Rogo 能否按其客户数量暗示的节奏扩张嵌入式投行人员模式,是真实执行约束,不是已经解决的问题。FDE 招聘放慢,会直接拖慢新客户上线,也会降低现有账户配置深度 [CR045]。 创造 Rogo 市场的同一波 AI 采用,也在重塑客户侧人才管线,而 Rogo 未来最终要从这条管线招募自己的前置部署人员:大型银行(JPMorgan Chase、Citigroup、Goldman Sachs、Morgan Stanley、Standard Chartered)正在缩减初级分析师招聘,因为 AI 自动化了建模、投行推介材料和可比公司分析工作;JPMorgan CEO 暗示,银行未来可能会招聘更多 AI 专家、更少传统银行从业人员;行业评论也警告,随着更少初级银行从业人员在过去培养资深交易专家的苦活中积累判断,「技能缺口」会扩大 [CR046]。这对 Rogo 是双刃剑:它验证了品类(银行正主动用 AI 替代初级劳动力),但也意味着,随着更少人进入 Rogo 创始人出身的分析师项目,Rogo 自己可招聘的、懂银行业务的技术人才池可能随时间收缩 [CR047]。 领导层与整合风险补足了这一类别。Rogo 既靠有机增长,也靠收购(Subset、Offset、Plux)增长;每笔交易都会在从约 35,000 用户、250+ 机构关系向下一轮定价所参照的使用基础扩张这一基础执行风险之上,叠加创始人留任、文化整合和代码整合风险。本章审阅的来源没有披露 Rogo 具名 CFO、首席合规官或首席风险官;对一家向金融服务最严格监管领域销售合规相邻 AI 工具的公司而言,这是一个显著治理缺口 [CR048]。合并来看,人才、整合和治理缺口意味着 Rogo 的运营风险目前集中在一个小型创始与早期员工团队身上;该团队要同时扩张多笔收购、多个地理区域,并应对异常紧张的 FDE 劳动力市场 [CR049]。
| 角色 / 职能 | 依赖或缺口 | 可能性 | 严重性 | 缓释措施 | 尽调路径 |
|---|---|---|---|---|---|
| 前置部署银行从业者 / 工程师(FDE) | 足够快地扩充嵌入式、客户定制配置人员,以匹配 250+ 家机构增长 | 高——FDE 岗位发布增长约 800–1,165%,候选人供给增长约 50% | 高——FDE 招聘放慢,会直接拖慢上线深度和新客户转化 | 直接与 OpenAI/Anthropic/Google/Palantir 争抢同一人才池 | 要求披露当前 FDE 人数、开放岗位数和平均填补周期 |
| 客户侧初级银行从业者管线 | 银行削减初级分析师招聘后,Rogo 自身可招聘的懂银行业务技术人才池可能收缩 | 中——已有多家银行招聘调整的报道(JPMorgan、Citi、Goldman、Morgan Stanley) | 中——这是 Rogo 自身招聘漏斗的二阶、多年期风险 | 未找到具体措施 | 要求披露 Rogo 当前招聘来源组合(前银行从业者与纯技术人才) |
| 具名 CFO / 首席合规官 / 首席风险官 | 已审阅来源未披露 Rogo 设有这些角色 | 中——快速扩张的 Series D 公司常见缺口,但鉴于客户受监管,仍值得关注 | 中——治理缺口与合规相邻产品相关 | Unknown | 要求披露 Rogo 当前高管名单以及治理 / 合规汇报线 |
| 被收购创始人留任(Subset、Offset、Plux) | 三项收购后整合同步推进,未披露留任条款 | 中——常见业绩对赌 / 归属悬崖期会在 2–3 年内带来流失风险 | 中——被收购创始团队流失会拆散最新框架能力 | 产品线迄今保留在 Rogo 品牌下 | 要求披露留任 / 业绩对赌条款及各创始团队当前状态 |
| 创始人集中度(CEO 和联合创始人主导产品 / 技术愿景) | 标准早期公司集中度;已审阅来源未披露创始人层级以下有深厚梯队 | 低中 | 一旦触发则高——创始人流失会给框架 / 数据合作路线图带来重大不确定性 | 前置部署团队和被收购团队在创始人下方补了深度 | 要求披露创始人 / 联合创始人层级以下的组织架构深度 |
可能性 / 严重性基于劳动力市场数据(FDE 岗位发布 / 薪酬)、多家银行招聘公告,以及已审阅来源未披露高管 / 继任细节;Rogo 未发布按职能拆分的人数或留任数据。
[CR044, CR045, CR046, CR047, CR048, CR049]| 风险 | 可监测触发项 | 阈值 / 事件 | 行动含义 |
|---|---|---|---|
| 不遵守 EU AI Act | Rogo 或具名客户收到 EU AI Office 信息请求、模型访问要求或召回行动 | 2026 年 8 月 2 日之后任何正式 AI Office 执法行动 | 升级对 Rogo 欧盟法律实体和面向客户合规陈述的尽调;重新评估 EU 收入敞口 |
| AI 相关证券 / 执法行动 | SEC 诉讼公告、FINRA 纪律处分或点名 Rogo 的私人证券诉讼 | 任何将 Rogo 列为当事方或重大事实证人的文件 | 在结果明确前暂停新增承诺;重新评估治理和披露控制 |
| Microsoft Copilot 捆绑侵蚀 | 独立基准测试或客户调查显示,在可商品化工作流(方差分析、标准模型搭建)上,Rogo 向捆绑 Copilot 智能体明显丢失份额 | 12 个月内有记录地流失 >=2 个具名参考客户给 Office 内替代方案 | 重新评估差异化论点:前置部署深度与 Microsoft 不断扩大的基线能力 |
| 估值与基本面缺口收窄(或扩大) | Rogo 披露 ARR/收入增长/利润率(例如未来融资、S-1 或新闻披露) | 首次披露经审计或投资人验证的收入数字 | 真实收入倍数可计算后,重新跑估值判断,不再从使用量倒推 |
| 降价轮 / 重新定价融资 | Rogo 后续融资轮或二级市场成交价低于 $2B Series D 估值 | 任何定价轮或经验证的二级交易低于 $2B | 视为 AI 板块修正具体影响 Rogo 的证据;重审进入价格纪律 |
| FDE / 前置部署招聘停滞 | Rogo 自身招聘页面或招聘报告显示,开放前置部署岗位空缺超过规模化阶段典型 3 个月填补周期 | 未填补前置部署岗位连续 2 个以上季度增加 | 下调近期新客户增长假设;把上线积压视为领先指标 |
| 独立准确性 / 事故披露 | 具名客户、监管机构或媒体调查记录 Rogo 在真实交易中出现重大输出错误 | 一份或多份经独立佐证的事故报告 | 视为与投资论点相关的可靠性事件;完整重估产品风险部分 |
触发项是本次尽调流程撰写的前瞻监测标准,不是预测;阈值仅作示例,待 Rogo 披露更多经营细节后应重新校准。
[CR015, CR021, CR025, CR038, CR041, CR043]7.7 附录
08估值
8.1 价格与内在价值:如何看待 $2B Series D
2026 年 4 月 29 日,Rogo 完成由 Kleiner Perkins 领投的 $160 million Series D,公司估值约 $2 billion——较 2026 年 1 月 Series C 的 $750 million 标记在短短三个月内接近翻三倍。Kleiner Perkins 合伙人 Mamoon Hamid 将这笔投资定义为一次押注:Rogo 正在成为「整个行业的操作系统」;Sequoia、Thrive Capital、Khosla Ventures 和 J.P. Morgan Growth Equity Partners 也都随新领投方继续投资。这些事实说明,在那一天,一个由知情、重复投资者组成的特定财团愿意以 $2 billion 价格购买 Rogo 股权。它们本身并不能证明 $2 billion 是外部可验证的内在价值。本次尽调审阅的任何来源都没有披露 Rogo 自身当年收入运行率、毛利率、净收入留存或现金续航期,也没有二级市场交易或独立公允性意见检验 Series D 价格。因此,本章把 $2 billion 视为一级市场融资出清价——真实存在,但由财团定价、未经市场检验——并刻意避免为 Rogo 自身构建 DCF 或 ARR 倍数估值。相反,本章把该价格放到可比私募融资、公开市场倍数和 2026 年融资背景中评估,并明确说明证据在哪里用尽。[CV001, CV002, CV003, CV032, CV036]
| 维度 | 评估 | 信号质量 | 决策含义 |
|---|---|---|---|
| 建议 | 继续研究 / 跟踪(不是买入或回避结论) | 中——可比公司已验证,Rogo 单位经济未披露 | 待当年 ARR、利润率和留存披露后重审 |
| 信心 | 中 | 中 | 取决于收入的独立验证 |
| 风险评级 | 中高 | 中 | 反映板块倍数压缩和捆绑风险 |
| 估值立场 | 相对可支撑证据偏贵 | 低(无已披露 ARR 可锚定倍数) | 不要把 $2B 标记估值视为后续进入价格的底线 |
| 进入纪律 | 仅在可参与主轮且拥有信息权时进入 | 中 | 避免在上一轮主轮估值之上支付二级溢价 |
| 目标回报逻辑 | 只能靠可比倍数重估,DCF 暂无支撑 | 低 | 跟踪 Hebbia/Glean/AlphaSense 重估,作为最接近参照 |
信号质量衡量每行由 Rogo 一手披露支撑的直接程度,而不是可比公司推断;多数行依赖可比公司,因为 Rogo 自身财务未公开。
[CV001, CV032, CV044]8.2 私募与公开可比基准
三家已披露收入的 AI-native 同行,最接近反映品类领先的金融与知识工作 AI 公司能拿到什么倍数。Hebbia 2024 年中 Series B 估值约 $700 million,对应约 $13 million 盈利 ARR,约 54x;TechCrunch 报道 The Information 估计,Glean 和 Harvey 同期交易倍数略高于 60x ARR。此后,Glean 截至 2026 年 5 月已扩张至 $300 million 年化收入,同时仍保持 $7.2 billion Series F 估值,隐含倍数约 24x,较 Glean 处于 $100 million ARR 时约 72x 已经压缩。AlphaSense 2026 年 6 月融资按约 $600 million ARR 估值 $7.5 billion,约 12.5x——一位分析师称,即便 AlphaSense 没有公开流动性,这一倍数也与公开 AI 软件 8–15x 未来 ARR 交易区间一致。Rogo 最接近的上市可比公司讲的是完全不同的故事:FactSet 约按 4.0x 过去收入交易,Intapp 约 3.1x,二者都远低于任何私募 AI-native 定价。这个落差是本章无法用公开证据解决的核心估值张力:Rogo 的 $2 billion 价格隐含假设它应享有私募 AI-native 倍数,而不是公开工作流软件倍数;但没有披露的 Rogo 收入数据能确认到底适用哪套估值体系。[CV004, CV005, CV006, CV007, CV008, CV009]
| 可比公司 | 指标 | 倍数 / 估值 / 状态 | 与 Rogo 的相关性 | 局限 |
|---|---|---|---|---|
| Hebbia(私有,Series B,2024 年中) | $13M ARR,已盈利 | $700M 估值,约 54x ARR | 按买方重叠度看最接近的 AI 原生同业(资产管理人、银行、律所) | 该轮距今已两年;当前 ARR / 估值未披露 |
| Glean(私有,Series F,2025 年 6 月;ARR 2026 年 5 月更新) | $300M 年化收入(2026 年 5 月),对应 $7.2B 估值(2025 年 6 月) | 约 24x 当前年化收入,低于 $100M ARR 时的约 72x | 显示 AI 原生倍数在 ARR 扩大后压缩得有多快 | 横向企业搜索,不是金融垂直;估值未按当前 ARR 重新定价 |
| AlphaSense(私有,2026 年 6 月融资轮) | $600M ARR(2026 年 Q1) | $7.5B 估值,约 12.5x ARR | 买方重叠(银行、资产管理人)和工作流品类最接近 | 内容库商业模式比 Rogo 的智能体优先产品更宽 |
| FactSet(上市,NYSE: FDS) | 过去 12 个月收入 $2.40B,FY2025 净利润 $597M | 市值约 $8.38B,EV/Sales 约 4.0x(低于 2025 年 10 月约 $10.67B 市值) | Rogo 自身智能体拉取可比公司的既有数据 / 工作流平台 | 上市倍数反映成熟盈利业务,不是成长阶段 AI 产品 |
| Intapp(上市,Nasdaq: INTA) | $560M 总 ARR(同比 +23%),FY2026 收入指引 $574–575M | 市值约 $2B,EV/Revenue 约 3.1x,EV/EBITDA 约 15.1x | 专业 / 金融服务领域的上市垂直工作流 SaaS 可比公司 | 不是 AI 原生;倍数反映传统业务加 AI 功能定位,不是 AI 优先定价 |
| S&P Global Market Intelligence / Capital IQ(上市,NYSE: SPGI,分部) | 为 FY2025 合并收入 $15.336B(同比 +8%)贡献收入 | 分部未单独按市场定价 | 正在搭建自有智能体层(Kensho Grounding)的数据平台既有厂商 | 分部收入未与母公司单独拆分;不是干净的独立倍数 |
覆盖范围是代表性样本,并非所有 AI 金融或垂直 SaaS 可比公司的穷尽列表;选择这些行,是因为它们与 Rogo 有直接买方或工作流重叠,并且在 2025–2026 年来源中至少有一个已披露估值收入数据点。
[CV004, CV006, CV008, CV010, CV011, CV012]用本章私募轮可比公司与公开市场可比公司观察到的 ARR 倍数区间,示意隐含估值对所用倍数的敏感度——并非关于 Rogo 实际 ARR 的判断。
柱形图绘制的是倍数,不是美元估值,因为 Rogo 自身 ARR 未披露;这个区间说明,在假设不同可比倍数适用时,同一个 $2B 价格会被解读得多么不同。
[CV004, CV006, CV007, CV008, CV015]Rogo 最接近的上市可比公司 EV/Revenue 倍数区间,与私募 AI 原生同业轮次隐含的更宽倍数带对照;如果用上市而非私募可比公司作基准,$2B 价格必须跨过这道缺口。
数值是 EV/Revenue 或 EV/Sales 倍数,不是美元估值,因为 Rogo 自身 ARR 未披露;上市可比公司的区间收敛为单点,因为每家公司只有一个离散、已披露的倍数,私募倍数带则覆盖本章观察到的 AlphaSense 到 Hebbia 全区间。
[CV004, CV008, CV011, CV013]8.3 投资论点、反论点与情景
牛市情景建立在品类领导地位上:Rogo 的融资节奏(四家蓝筹投资者复投,加上一家新的顶级领投方)、本尽调其他章节记录的其在具名大型投行和顶级精品投行账户中的嵌入位置,以及 AI 捕获 2026 年 Q1 全球风险投资约 80% 的宏观背景,都支持一个判断:资本正集中到被认为是品类赢家的公司身上,而不是无差别扩散。反论点是,用来支撑价格的同一组可比公司——Hebbia、Glean、AlphaSense——自身就因 12.5x–54x ARR 倍数且缺乏公开市场流动性而受到公开质疑;Rogo 也没有披露收入、利润率或留存数据,让外部投资者无法确认它应获得类似倍数,而不是公开工作流软件倍数。基准情景下,Rogo 仍未披露的当前 ARR 低于 $2 billion 标记在 AI-native 倍数下所需水平,因此下一轮名义估值守住,但随着披露追上来,实际倍数压缩;熊市情景下,低于市场传闻的 ARR 披露或一个大型投行账户流失,会迫使降价轮。牛市情景有真实、可验证的融资信号支撑;反论点同样有真实披露缺口支撑。二者同时成立,这正是本章立场是跟踪 / 继续研究,而不是方向性判断的原因。[CV002, CV017, CV019, CV021, CV024, CV032]
| 维度 | 投资论点 | 反论点(什么会改变判断) |
|---|---|---|
| 品类定位 | 按 Series D 领投方说法,Rogo 正在成为金融垂直 AI 的品类定义型「操作系统」 | 如果大型投行推出同等能力的自研工具,或 Microsoft/Bloomberg 捆绑工具达到同等能力,品类领导者溢价会压缩 |
| 融资节奏 | 四家蓝筹投资人跨轮次跟投,显示财团对 Rogo 增长有信心 | 一次降价轮,或此前领投方未能续投,都会是强负面信号 |
| 可比公司重估 | AI 原生同业(Hebbia、Glean、AlphaSense)均因披露 ARR 增长而上调估值,说明整个品类仍有重估空间 | 同一批同业也因 12.5x–54x ARR 倍数且无公开流动性而遭遇公开质疑;其中任何一家修正,都可能重定价整个品类 |
| 上市可比公司 | 既有数据平台(FactSet、Intapp、S&P Global)证明,AI 增强工作流收入真实且可变现 | 这些既有公司交易倍数是 3x–4x 收入,不是私有 AI 原生融资暗含的两位数倍数——Rogo 的价格假设它能摆脱这种估值重力 |
| 宏观背景 | AI 拿走 2026 年 Q1 全球风投资金约 80%,证明资本仍偏爱品类领导者 | 据报私有 AI 估值自 2025 年末以来下降 23%,AI 投资收入比为 4:1,二者都指向广泛重定价风险 |
| 披露 | Rogo 的增长叙事(融资节奏、投资人质量、客户数增长)获得独立佐证 | 当年 ARR、毛利率、净收入留存和现金续航期仍未披露,阻碍任何独立承保 |
每行把一个投资论点和会推翻该论点的具体证据配对;内容同时来自 Rogo 相关融资事实,以及本章其他位置引用的 2025–2026 年可比公司和宏观来源。
[CV002, CV017, CV019, CV032]| 情景 | 关键假设 | 估值 / 回报逻辑 | 关键风险 | 概率信号 |
|---|---|---|---|---|
| 乐观 | Rogo 守住品类领导者地位,并补上 ARR 披露缺口;披露年化口径与近期融资节奏匹配 | $2B 标记估值守住,或随 AI 原生同业可比公司(Hebbia、Glean、AlphaSense 在 2025–2026 年均上调估值)继续上修 | 既有厂商捆绑(Microsoft、Bloomberg、S&P Global)在 Rogo 证明持久 ARR 前削弱付费意愿 | 四家蓝筹上一轮投资人继续跟投提供支撑 |
| 基准 | Rogo 继续增长,但当前 ARR 低于 $2B 标记估值按 AI 原生倍数应要求的水平(可比公司观察到 12.5x–54x) | 下一轮名义估值守住,但 ARR 披露后有效倍数压缩 | 2026 年市场数据已提示整个 AI 软件板块有倍数压缩风险 | 与据报 AI 初创估值自 2025 年末以来约下降 23%、AI 交易数量下降 14% 一致 |
| 悲观 | 当前 ARR 显著低于市场传闻,和/或大型投行客户流失或内包 | 下一轮降价或平轮;按 $2B 标记估值进入的后续投资人出现有效亏损 | 未披露单位经济,加上 Rogo 自身客户群中的大型机构 AI 就业前景为负 | 与 2026 年融资追踪记录的 AI 初创「崩盘出局」模式一致 |
| 什么会改变判断 | 经独立验证的当年 ARR、毛利率和净收入留存 | N/A——直接拆解基准 / 乐观 / 悲观分歧 | N/A | 本次尽调中对决策最关键的单项缺失披露 |
由于没有已披露 ARR、利润率或股数可锚定数值回报计算,本表有意省略回报数字;按照本章受证据约束的方法,本表只表达情景逻辑和概率信号。
[CV004, CV006, CV008, CV019, CV021, CV024]截至本轮,本章按市场验证、护城河、单位经济、估值支撑和证据质量给出可提交 IC 的评分。
分数是本次尽调的定性 0-10 判断,不是供应商或投资方给出的评级。
[CV002, CV025, CV028, CV032, CV040]8.4 宏观 AI 估值背景与竞争捆绑风险
全行业数据让「私募 AI 倍数会简单延续」这一假设变得复杂。一项引用 PitchBook 和 IDC 数据的分析估计,年度 AI 投资约 $400 billion,而已实现企业 AI 收入只有约 $100 billion,比例为 4:1;该分析将其类比为以往修正前的技术周期,并称私募 AI 创业公司估值自 2025 年末以来已下跌约 23%。另一份融资追踪显示,2026 年 AI 交易数量同比下降约 14%,但投资总额上升;其描述的是一个杠铃市场:巨额融资轮和微型融资轮增长,而缺乏强指标的中期创业公司挣扎,并称 2026 年倒闭的 AI 公司数量超过前三年总和。竞争上,在位厂商没有停下:Microsoft 把 Copilot for Finance 附加产品捆绑进既有 Microsoft 365 席位基础;Bloomberg 在其 $12.6 billion 收入的 Terminal 上增加了 agentic 界面;与此同时,Perplexity 的竞争产品「Computer」展示出相近研究和建模功能,成本只是很小一部分;S&P Global 的 Kensho 部门则直接在 Rogo agents 所授权使用的 Capital IQ 数据之上,构建了自己的 agentic「Grounding」框架。上述情况都不能证明替代已经发生,但它们是溢价倍数所假设定价权的结构性逆风。S&P Global 自己的 2026 年劳动力市场调查还增加了一个更长期担忧:大型企业——Rogo 的核心买方群体——预计 AI 驱动就业影响为净负,这最终可能缩小 Rogo 工作流产品要增强的初级分析师人数。[CV018, CV019, CV020, CV021, CV022, CV024]
8.5 退出准备度与验证路径
接近 Rogo 品类的 AI-native 企业研究同行还没有进入公开市场,因此没有上市退出可比公司可用来锚定最终倍数。Built In 2026 年 IPO 观察名单把 Databricks 列为唯一接近上市且披露盈利能力的大型 AI-native 公司(约 $5.4 billion 年化收入、自由现金流为正);OpenAI 和 Anthropic 尽管估值被讨论到 $850 billion 和 $900 billion 以上,仍未盈利。另一份 AI IPO 追踪称 2026 年是史上 AI 集中度最高的 IPO 年,但指出后期私募倍数仍明显低于 2021–2022 年峰值;一项 pre-IPO 市场分析发现,2026 年 IPO 轨道上的创业公司只有约四分之一真正准备好上市,其余卡在过期披露窗口和公开投资者无法接受的估值之间。Rogo 自身迄今是收购方——收购 Offset 和本尽调其他部分提到的其他标的——而非被收购标的;这削弱了近期战略出售情景,相对更支持更长续航期后走向 IPO 或更晚、更大的战略交易。FactSet、S&P Global 和 Bloomberg 是结构上最有可能考虑把 Rogo 作为收购标的而非竞争对手的在位厂商,因为它们各自已经在独立建设 agentic research 能力。补齐本章最大证据缺口——当年收入、利润率、留存、现金续航期,以及任何二级市场定价——比继续做更多可比公司研究更能明确估值立场。[CV030, CV031, CV034, CV037, CV042, CV043]
| 主题 | 缺失证据 | 为什么重要 | 尽调路径 |
|---|---|---|---|
| 当年收入 | 经审计或管理层报告的当年 ARR / 收入年化口径 | 这是将任何可比倍数直接套用到 Rogo 的最大阻碍 | 任何后续承诺前,在 NDA 下要求披露当年收入和按客户批次拆分的 ARR |
| 单位经济 | 毛利率、每席位模型推理成本和 CAC/LTV | 需要区分它是可持续软件毛利模型,还是服务更重、毛利更低的模型 | 要求披露收入成本明细和每席位推理成本基准 |
| 留存 | 净收入留存和客户级流失明细 | 决定可比 ARR 倍数(Hebbia、Glean、AlphaSense)是否还是正确参考类别 | 要求提供本尽调点名账户的客户批次留存曲线 |
| 现金状况 | 手头现金、烧钱率和现金续航(月数) | 影响退出事件前还存在多少稀释风险 | 要求最近资产负债表快照和月度烧钱趋势 |
| 二级市场证据 | 任何要约回购、员工二级出售或给 Rogo 定价的公平性意见 | 会给出第一个独立于主轮融资的市场验证估值信号 | 向股权结构表 / 二级平台和现有投资人查询近期成交价 |
| 退出准备 | CFO / 审计师招聘或投行接触,显示 IPO 或出售准备 | 表明管理层自身的流动性时间表和信心 | 监控高管招聘公告和公开 S-1 / 注册文件 |
按本章对阻塞严重性的判断排序;这六项均未被本次已审阅来源解决。
[CV032, CV041]8.6 论点破裂触发器与最终估值立场
少数具体事件会把本章不确定性推向一个方向:Rogo 下一轮融资出现降价轮或平轮;披露的当年 ARR 隐含倍数高于 12.5x–54x 可比区间;一个标志性大型投行或顶级精品投行账户因流失或自研而丢失;确认有账户切换到捆绑式在位厂商工具(Microsoft Copilot for Finance、Bloomberg agentic Terminal,或 S&P Global / Kensho 产品);或独立二级市场交易把 Rogo 定价在 $2 billion 以下。截至本轮,这些都没有被观察到。综合权衡证据——真实、可验证的融资和投资者质量信号,对上未披露收入基础、自身也受到公开争议的可比组,以及全行业倍数压缩和捆绑风险——本章估值立场是:相对今天能独立承销的证据,$2 billion 标记看起来偏高;但考虑到私募市场可比区间,它并非不可辩护。合适姿态是跟踪 / 继续研究,而不是直接买入或回避:买入需要独立验证的当年 ARR 和利润率数据,公共记录中目前还没有;回避也没有足够支撑,因为融资节奏和复投模式是真实信号,说明机构信心仍在,而本尽调已独立佐证这些信号。[CV034, CV040, CV041, CV044]
| 触发项 | 阈值 | 对投资论点的传导 | 行动含义 |
|---|---|---|---|
| 下一轮降价轮或平轮 | 下一次披露融资投后估值低于 $2B | 直接证伪「品类领导者重估」牛市情景 | 将估值立场下调至昂贵 / 回避;重估进入价格 |
| 披露的当年 ARR 远低于可比倍数暗含水平 | 按 $2B 标记估值计算,当前 ARR 暗含倍数高于 12.5x–54x 可比区间上沿 | 证实基准 / 悲观情景:价格跑在披露基本面前面 | 从跟踪 / 继续研究转为回避,等待重新定价 |
| 流失标志性大型投行或精英精品投行客户 | 本尽调客户证据中的任何具名客户公开流失,或将同等工具内包 | 削弱品类领导者投资论点,也削弱可比公司里隐含的留存假设 | 视为重大不利事件,需要立即重新承保 |
| 既有厂商捆绑替代具名客户 | 确认某个客户切换到 Microsoft Copilot for Finance、Bloomberg ASKB 或 S&P Global/Kensho 智能体产品 | 直接验证既有厂商捆绑的反论点 | 下调对持久定价权的信心;重审 TAM 假设 |
| 独立二级市场成交价低于 Series D 价格 | 经确认的二级交易或公平性意见给 Rogo 定价低于 $2B | 给出第一个经市场检验(非仅财团定价)的估值信号 | 以二级市场成交价而非主轮价格作为参考标记 |
阈值是从本章证据提炼的定性触发项,不是 Rogo 或其投资人披露的合同约束。
[CV034, CV036, CV025, CV027]从已披露的规模与融资证据,串到未披露的单位经济,再到本章跟踪 / 继续研究建议。
节点标签概括本章发现;流程展示逻辑依赖,不是加权评分模型。
[CV001, CV032, CV004, CV011]8.7 附录
免责声明
本报告是基于公开证据的尽调快照,不构成投资建议。重要的财务、法律、技术和合同事实仍未公开;作出任何投资决定前,应直接向管理层和一手文件核验。
证据索引
| 编号 | 陈述 | 可信度 | 来源 |
|---|---|---|---|
| CO001 | Rogo is an AI platform purpose-built for financial services, serving investment banks, private equity firms, and asset managers. | 高 | SO001, SO015 |
| CO002 | Rogo is headquartered in New York, NY, United States. | 高 | SO002, SO022 |
| CO003 | Rogo was co-founded by Gabriel Stengel, John Willett, and Tumas Rackaitis. | 高 | SO002, SO021, SO025 |
| CO004 | Gabriel Stengel is Rogo's CEO and co-founder. | 高 | SO002, SO015, SO022 |
| CO005 | John Willett is a Rogo co-founder who leads the company's European expansion from its London office. | 高 | SO010, SO025 |
| CO006 | Tumas Rackaitis is a Rogo co-founder and its chief technology officer. | 中 | SO021, SO027 |
| CO007 | Gabriel Stengel previously worked as an investment banking analyst at Lazard before founding Rogo. | 高 | SO021, SO025 |
| CO008 | John Willett previously worked at J.P. Morgan Chase and Barclays before co-founding Rogo. | 中 | SO021, SO025 |
| CO009 | Tumas Rackaitis holds a computer science degree from Oberlin College. | 中 | SO021 |
| CO010 | All three Rogo co-founders met as classmates at Princeton University. | 中 | SO020, SO025 |
| CO011 | Rahul Rekhi joined Rogo as President after roughly a year in the U.S. Treasury Department and seven years at Lazard. | 中 | SO025 |
| CO012 | Forbes' company profile states Rogo was founded in 2022. | 中 | SO022 |
| CO013 | Rogo's own October 2024 Series A press release, and Hebbia's 2026 competitor guide, both state Rogo was founded in 2021, one year earlier than Forbes' reported 2022 founding year. | 中 | SO016, SO023 |
| CO014 | Bloomberg reporting states Gabriel Stengel quit Lazard in late 2021 to begin building Rogo's technology with his co-founders around a Manhattan kitchen table, ahead of the company's formal 2022 founding date. | 高 | SO025, SO026 |
| CO015 | New York Weekly reports that Stengel, Willett, and Rackaitis founded Rogo in January 2022 after leaving J.P. Morgan and Lazard. | 中 | SO020 |
| CO016 | As of April 2026, more than 35,000 finance professionals at over 250 institutions use Rogo's platform. | 高 | SO015, SO017, SO018 |
| CO017 | Rogo's named institutional clients include Rothschild & Co, Jefferies, Lazard, Moelis, and Nomura. | 高 | SO015, SO018, SO010 |
| CO018 | Bloomberg reporting names additional Rogo clients including J.P. Morgan, Bank of America, Wells Fargo, and Singapore sovereign-wealth fund GIC. | 中 | SO025 |
| CO019 | Rogo's homepage displays customer endorsements from Truist Securities' CEO, Nomura's international head of investment banking, and Baird Global Investment Banking's COO. | 中 | SO001 |
| CO020 | More than 100 professionals at Baird's Equity Research division actively use Rogo, with about 85% weekly and 70% daily active usage. | 中 | SO012 |
| CO021 | Rogo's flagship autonomous AI agent, Felix, launched around the April 2026 Series D and executes multi-step tasks such as deal screening, CIM generation, buyer outreach, and data-room diligence. | 高 | SO015, SO025 |
| CO022 | Felix is named after Felix Rohatyn, a Lazard investment banker credited with helping rescue New York City from its 1970s fiscal crisis. | 中 | SO025 |
| CO023 | Rogo integrates with financial data and model partners including OpenAI, Google Gemini, Anthropic, LSEG, S&P Global, FactSet, and PitchBook. | 中 | SO031 |
| CO024 | Rogo lets client firms toggle between underlying AI models such as Anthropic's Claude, OpenAI's ChatGPT, and Google's Gemini rather than committing to a single model provider. | 中 | SO025 |
| CO025 | In 2025 Rogo acquired Subset, adding spreadsheet-agent technology that can audit and roll forward complex Excel financial models. | 中 | SO023 |
| CO026 | In March 2026 Rogo acquired Offset, an AI-agent startup founded by Raj Khare and Shiv Shrivastava, to embed learning agents that maintain and update financial models. | 高 | SO014, SO031 |
| CO027 | The specific product names "Rogo Research," "Rogo Comps," "Rogo Models," and "Rogo Pitchbook" referenced in early diligence materials could not be corroborated on Rogo's own site or in independent 2026 coverage, which instead describe Felix plus data-partner integrations (e.g. PitchBook, FactSet) as the operative product structure. | 低 | |
| CO028 | Rogo raised a $7 million seed round in February 2024 led by AlleyCorp, with participation from Company Ventures, BoxGroup, and ScOp Ventures. | 中 | SO021, SO028 |
| CO029 | SixThirty Ventures dates Rogo's seed round to a roughly $48 million post-money valuation. | 低 | SO029 |
| CO030 | Rogo raised an $18.5 million Series A on October 1, 2024, led by Khosla Ventures at an $80 million post-money valuation, bringing total funding to $26 million. | 高 | SO016, SO028, SO029 |
| CO031 | Series A participants included Mantis VC, Jack Altman, and former Google CEO Eric Schmidt, and Khosla General Partner Keith Rabois joined Rogo's board. | 中 | SO016 |
| CO032 | Rogo raised a $50 million Series B around April-May 2025, led by Thrive Capital with J.P. Morgan Growth Equity Partners, Tiger Global, and Positive Sum Ventures, bringing total funding to $75 million. | 高 | SO011, SO028 |
| CO033 | Forbes and Sacra both value Rogo's Series B at approximately $350 million post-money. | 中 | SO022, SO028 |
| CO034 | Rogo raised a $75 million Series C in January 2026, led by Sequoia Capital with participation from Henry Kravis and Wells Fargo, bringing total funding to more than $165 million. | 高 | SO010, SO030, SO028 |
| CO035 | The Series C valued Rogo at $750 million post-money, more than doubling its Series B valuation in under a year. | 高 | SO022, SO028 |
| CO036 | Rogo used Series C proceeds to open its first international office in London, led by co-founder John Willett. | 高 | SO010, SO025 |
| CO037 | On April 29, 2026, Rogo announced a $160 million Series D led by Kleiner Perkins, with Sequoia, Thrive Capital, Khosla Ventures, J.P. Morgan Growth Equity Partners, BoxGroup, Mantis VC, Jack Altman, Evantic, and Positive Sum participating. | 高 | SO015, SO017, SO018 |
| CO038 | The Series D brought Rogo's total funding to more than $300 million. | 高 | SO015, SO018, SO020 |
| CO039 | Independent reporting (Bloomberg and TBPN Digest) states the Series D priced Rogo at a $2 billion valuation, up from $750 million three months earlier. | 高 | SO025, SO019 |
| CO040 | New York Weekly flags that J.P. Morgan Growth Equity Partners is both a Rogo investor and, through J.P. Morgan the bank, among its cited institutional customers, a dual investor-customer relationship worth diligence scrutiny. | 中 | SO020 |
| CO041 | Forbes reports Rogo's revenue grew from approximately $2 million in 2024 to more than $15 million in 2025. | 中 | SO022 |
| CO042 | Forbes lists Rogo's headcount at approximately 100 employees as of April 2026. | 中 | SO022 |
| CO043 | Bloomberg reporting states Rogo's leadership expects headcount to reach close to 300 employees by the end of 2026. | 中 | SO025 |
| CO044 | OpenAI's partner case study states that since emerging from stealth in 2024, Rogo served over 5,000 bankers and grew annual recurring revenue 27x using OpenAI's models. | 中 | SO027 |
| CO045 | PeerSpot's mid-2026 mindshare data ranks Rogo 5th among Financial Data Analysis Platforms with 2.3% mindshare, versus FactSet's 18.7% and up from 1.1% a year earlier. | 低 | SO024 |
| CO046 | Hebbia's 2026 competitor guide (published by a rival vendor) flags limitations in Rogo's ability to scale analysis across thousands of documents, its response-level (rather than sentence-level) citation granularity, and limited team-collaboration features. | 中 | SO023 |
| CO047 | SixThirty Ventures' analyst commentary questions whether AI-analyst valuations, including Rogo's, are pricing in outlier-level revenue multiples relative to demonstrated traction and differentiation. | 中 | SO029 |
| CO048 | Bloomberg reporting notes that some AI-industry skeptics view Rogo as an unnecessary intermediary layer, since finance professionals could query large general-purpose AI models directly. | 中 | SO025 |
| CO049 | Bloomberg reporting describes anxiety among junior bankers that Rogo-style automation could reduce entry-level hiring, even as Rogo's founders argue the technology will let banks add more senior dealmakers instead. | 中 | SO025 |
| CO050 | Rogo announced EU AI Act compliance readiness on March 19, 2026, ahead of the Act's full enforceability in August 2026, following an internal assessment validated by external auditors. | 高 | SO013, SO006 |
| CO051 | Rogo states it does not use client data to train or update its models and maintains SOC 2, ISO 27001, ISO 42001, and GDPR-aligned documentation. | 中 | SO005, SO013 |
| CO052 | In June 2026 Rogo became a launch partner for Microsoft Copilot in Excel, surfacing Rogo's investment-banking and equity-research workflows directly inside Excel for shared customers. | 高 | SO008, SO006 |
| CO053 | In May 2026 Rogo deepened its data partnership with PitchBook, integrating PitchBook's Premium Connector deal, fund, company, and investor data directly into Rogo's natural-language query interface. | 中 | SO032, SO006 |
| CO054 | Rogo's news index lists additional 2026 partnership and product announcements, including Daloopa (May 20), SS&C Intralinks (June 15), and a Credit Center product launch (June 22). | 中 | SO006 |
| CO055 | Rogo's product page describes institutional-grade outputs as auditable Excel models, investment memos, diligence materials, and slide decks produced from an integrated, secure platform. | 中 | SO004 |
| CO056 | Sacra describes Rogo's spreadsheet agent as able to read, explain, audit, and edit complex 40-tab valuation models, refresh comparables from Capital IQ, and write outputs back into Excel. | 中 | SO028 |
| CO057 | Rogo's workforce is split roughly evenly between engineers and former finance professionals it calls "forward deployed bankers," who work directly with client banks -- often ones they previously worked at -- to embed the platform into existing workflows. | 中 | SO025 |
| CM001 | The global generative AI in financial services market was $2.51 billion in 2026, up from $1.95 billion in 2025, according to Precedence Research. | 中 | SM001 |
| CM002 | Precedence Research forecasts the generative AI in financial services market will reach $17.88 billion by 2035, a 24.81% CAGR from 2026. | 中 | SM001 |
| CM003 | The Business Research Company estimates the generative AI in banking and finance market at $1.75 billion in 2025, growing to $7.71 billion by 2030 at a 34.5% CAGR, a materially different scope and growth trajectory than Precedence Research's broader financial-services estimate. | 中 | SM002 |
| CM004 | Corporate and investment banking (CIB) generated $3.0 trillion in global revenue in 2024, growing 4.4% year over year, per McKinsey's annual CIB report. | 中 | SM004 |
| CM005 | McKinsey estimates that applying AI and operating-model levers across CIB could improve profitability by 20 to 30 percent versus baseline, before macro effects and investment costs. | 中 | SM004 |
| CM006 | Deloitte cites a Stanford study finding generative AI boosted a call center's productivity by 14%, and an MIT study finding generative AI reduced time and improved work quality for marketers, consultants, and data analysts. | 中 | SM003 |
| CM007 | Global private equity assets under management reached approximately $8 trillion in 2026, up from $4 trillion five years earlier, per a ranking of the top 100 PE firms by AUM. | 中 | SM017 |
| CM008 | Blackstone alone manages $1.3 trillion in assets, of which its traditional private-equity business represents roughly $350 billion, about 27% of the total, illustrating how concentrated and diversified the largest PE platforms have become. | 中 | SM017 |
| CM009 | Global assets under management across the asset management industry reached $147 trillion in 2025, up 11% year over year, with more than 80% of 2025 revenue growth driven by market appreciation rather than net new flows, per BCG's Global Asset Management Report 2026. | 中 | SM026 |
| CM010 | Nearly 90% of surveyed PE investors integrate digital or AI value-creation levers into diligence or value-creation planning, and PE-backed companies with mature AI capabilities show nearly double the return on invested capital of peers, per BCG's PE investor survey. | 中 | SM006 |
| CM011 | Digital initiatives alone deliver 15-20% ROI, but AI built on mature digital infrastructure can reach 30-35% total returns and reach time-to-value 40% faster, per BCG's PE investor survey. | 中 | SM006 |
| CM012 | KPMG's Quarterly AI Pulse Survey finds asset managers and PE firms moving from AI experimentation toward measurable ROI while offering compensation premiums to recruit AI-skilled talent. | 中 | SM007 |
| CM013 | EY's 4Q AI Pulse report finds PE investment levels in AI now match other sectors, with the competitive question shifting from adoption to differentiation. | 中 | SM005 |
| CM014 | A Bloomberg Terminal costs approximately $27,660 per year for a single license, dropping to about $24,240 per year per terminal for two or more terminals, and Bloomberg holds roughly 33% of financial-data-terminal revenue. | 中 | SM008 |
| CM015 | FactSet's core subscription costs approximately $12,000 per year, while Refinitiv Eikon ranges from about $3,600 for a stripped-down version to $22,000 per year for full access, per Wall Street Prep's platform comparison. | 中 | SM008 |
| CM016 | Financial-data-terminal revenue is dominated by four incumbents -- Bloomberg (~33%), Refinitiv Eikon (~20%), Capital IQ (~6%), and FactSet (~4.5%) -- creating an entrenched, bundled substitute that any new AI research tool must compete against or integrate with. | 中 | SM008 |
| CM017 | US sell-side equity research analyst headcount has fallen roughly 18% since 2015 amid MiFID II-driven fee compression, even as buy-side spending, outsourced research, and AI-powered research tools have grown. | 中 | SM009 |
| CM018 | The global equity research industry generates approximately $8.7 billion in annual revenue, concentrated among a handful of large providers. | 中 | SM009 |
| CM019 | Financial services firms spend an average of about 9.2% of revenue on IT annually, roughly triple the 2.8-3.2% spent by manufacturing or retail firms, reflecting regulatory, cybersecurity, and compliance-driven technology intensity. | 中 | SM027 |
| CM020 | Financial-services IT budgets by sub-sector average approximately 8.7% of revenue for commercial banks, 9.4% for global banks, and 10.2% for asset management and capital markets firms. | 中 | SM027 |
| CM021 | Tracxn categorizes Rogo as AI software for investment banks, private equity firms, and hedge funds that delegates research tasks to a domain-specific personal analyst integrating internal and external data sources. | 中 | SM010 |
| CM022 | Competing vendor commentary frames Hebbia and Rogo as built for large financial institutions and enterprise banking/PE teams, while newer entrants such as askRIA target private credit funds, smaller PE firms, and family offices with leaner deal teams. | 中 | SM012 |
| CM023 | Hebbia, a direct competitor in AI-driven financial research, raised $130 million at a reported $700 million valuation as of mid-2026, underscoring continued investor appetite for enterprise AI research tools targeting finance, law, and consulting. | 中 | SM011 |
| CM024 | Anthropic released ten ready-to-run AI agent templates for financial services in 2026 covering pitchbook generation, earnings-review monitoring, financial modeling, and comparables checks -- tasks historically performed by junior investment bankers. | 中 | SM013 |
| CM025 | Industry commentary claims major investment banks cut junior-analyst intake classes by as much as two-thirds in 2026 while shifting routine modeling and pitchbook tasks to large language models, though this figure comes from a single lower-tier outlet and is not independently corroborated. | 低 | SM014 |
| CM026 | 94% of surveyed hedge fund, asset-manager, and pension-fund professionals expected to increase alternative-data and AI research spending in 2026, with 18% expecting a substantial increase, per an Exabel-sponsored global survey. | 中 | SM015 |
| CM027 | 58% of alternative-investment fund managers expected wider front-office generative-AI integration over the following year, up from 20% in 2023, and 95% reported already using generative AI in their work, per AIMA research. | 中 | SM016 |
| CM028 | 60% of institutional investors surveyed by AIMA said they would be more likely to invest in a hedge fund that allocates a meaningful share of its budget to generative-AI research and implementation. | 中 | SM016 |
| CM029 | Microsoft shipped finance-specific Copilot capabilities in 2026, including a Finance Agent embedded in Excel, Outlook, and Teams, explicitly targeting FP&A, accounting, tax, compliance, and treasury workflows. | 高 | SM018, SM019 |
| CM030 | Microsoft's Finance Agent release plan for 2026 wave 1 (April-September 2026) emphasizes conversational access to ledgers and subledgers plus deeper governance and extensibility for enterprise finance teams. | 中 | SM019 |
| CM031 | Despite record AI capital-expenditure guidance of $190 billion for 2026 (up 61% year over year), Microsoft's Copilot paid-seat penetration was estimated at only about 3.3% of the Microsoft 365 installed base, and Microsoft shares underperformed the S&P 500 by more than 15 percentage points year-to-date as of an April 2026 analysis, reflecting investor skepticism about near-term Copilot monetization. | 低 | SM020 |
| CM032 | FINRA's 2026 Annual Regulatory Oversight Report added a dedicated generative-AI section for the first time, flagging hallucination and accuracy risk, bias, cybersecurity exposure, and expanding supervisory expectations to autonomous 'agentic' AI. | 高 | SM021, SM022 |
| CM033 | The SEC's Division of Examinations FY2026 priorities, released in late 2025, elevated scrutiny of firms' AI governance, explainability, and 'AI-washing' in marketing claims, while requiring AI-generated communications and prompt/output logs to be retained as supervised books and records. | 中 | SM022 |
| CM034 | FINRA supervisory obligations under Rules 3110 and 4370 and Regulation S-P apply equally to AI-enabled tools, and firms cannot outsource compliance responsibility to third-party AI vendors. | 中 | SM021 |
| CM035 | MIT's 'The GenAI Divide' study found a 95% failure rate among enterprise generative-AI pilots, defined as failing to show measurable financial return within six months, fueling concern about an AI investment bubble. | 中 | SM024 |
| CM036 | 61% of surveyed business leaders said they feel more pressure to prove AI ROI now than a year earlier, and roughly half of surveyed investors expect positive ROI within six months, intensifying scrutiny of AI spending including in financial services. | 中 | SM023 |
| CM037 | Enterprise AI adoption reached about 88% globally in 2025, yet only about one-third of organizations had scaled beyond pilot projects and just 6% qualified as AI 'high performers' achieving 5%-plus EBIT impact, per McKinsey's State of AI 2025 as cited in a private-equity AI adoption report. | 中 | SM025 |
| CM038 | 86% of surveyed organizations had integrated generative AI into M&A workflows by 2025, with 65% doing so within the prior year alone, per a Deloitte M&A generative-AI study cited in industry reporting. | 中 | SM025 |
| CM039 | Skills shortages were cited by 71% of enterprises that evaluated but did not implement AI, making talent the largest single adoption barrier in private-equity AI deployment. | 中 | SM025 |
| CM040 | The EU AI Act's most consequential enforcement phase reaches financial-services firms in August 2026, adding compliance and governance burden to AI deployment in private equity and asset management. | 中 | SM025 |
| CM041 | Unlike incumbent financial-data terminals that primarily bundle static data delivery, and unlike general-purpose AI office copilots such as Microsoft 365 Copilot that are not finance-workflow-specific out of the box, the core AI-research-automation category for IB/PE/AM centers on agentic execution of research tasks (comps pulls, diligence synthesis, pitchbook drafting) built on top of licensed financial data. | 中 | SM008, SM010 |
| CM042 | Tracxn describes Rogo's product as delegating research tasks to a 'domain-specific personal analyst' that integrates internal and external data sources for finance, distinguishing it from generic large-language-model chat assistants. | 中 | SM010 |
| CM043 | No analyst report reviewed isolates a serviceable addressable market specifically for AI research/workflow-automation software sold into investment banks, PE firms, and asset managers; published TAM figures conflate broader generative-AI-in-financial-services spend with narrower banking-specific tooling. | 低 | |
| CM044 | Bulge-bracket and global investment banks are the earliest and most well-funded buyers of AI research/workflow tools, given IT budgets averaging roughly 9-11% of revenue and existing large-scale procurement relationships with data vendors such as Bloomberg, FactSet, and LSEG. | 中 | SM027, SM008 |
| CM045 | Middle-market and boutique investment banks represent a later-adopting, more price-sensitive segment relative to bulge-bracket peers, based on the same relative IT-spend and terminal-pricing benchmarks. | 中 | SM027, SM008 |
| CM046 | Private equity firms are increasingly embedding AI into diligence, deal-lifecycle automation, and portfolio-company operations, with general partners at the fund level -- not individual portfolio companies -- typically owning the AI tooling budget. | 中 | SM006, SM025 |
| CM047 | Hedge funds and asset managers budget for AI and alternative-data tools separately from pure data-acquisition spend, with software/technology commonly consuming roughly a third to half of total alternative-data budgets. | 中 | SM015 |
| CP001 | Rogo's competitive set spans five tiers: direct AI-native peers (Hebbia, F2, Marvin Labs, Fiscal.ai, Quartr), incumbent data-terminal bundlers building native AI (Bloomberg, FactSet, S&P Capital IQ/Kensho, LSEG), horizontal AI copilots (Microsoft 365 Copilot, Glean), adjacent deal-workflow CRM platforms (Intapp DealCloud), and status-quo alternatives (in-house bank-built GenAI, human-analyst outsourcing). | 中 | SP002, SP028 |
| CP002 | Marvin Labs' June 2026 buyer's guide explicitly frames Rogo and Hebbia as the two enterprise-tier picks 'best for deal teams in banking and PE,' distinct from lighter-weight tools aimed at individual equity analysts. | 中 | SP002 |
| CP003 | Large investment banks including JPMorgan Chase, Goldman Sachs, and Morgan Stanley have each released proprietary in-house generative-AI tools to their workforce, giving bulge-bracket banks a build-vs-buy alternative to vendor platforms like Rogo. | 高 | SP025, SP026 |
| CP004 | JPMorgan's internal LLM Suite platform ingests more of the bank's proprietary databases and applications every eight weeks and was demonstrated generating a full investment-banking pitch deck in about 30 seconds, work that previously took a team of junior bankers hours. | 中 | SP025 |
| CP005 | Goldman Sachs rolled out its GS AI Assistant to roughly 10,000 employees as of January 2025 with a goal of firm-wide coverage, built on a rotating set of third-party models (OpenAI, Google Gemini, Meta Llama) rather than a single licensed vendor platform. | 中 | SP026 |
| CP006 | SP2 Analytics, a firm supplying offshore CA/CFA/MBA research analysts to investment banks and PE firms, markets human-analyst outsourcing explicitly as a hedge against AI hallucination risk in investment-research deliverables, positioning trained analysts as a status-quo substitute for AI research platforms. | 中 | SP022 |
| CP007 | Hebbia's Matrix product is a grid-based, cell-level document-intelligence platform that processes thousands of documents simultaneously using multi-agent orchestration, optimized for high-stakes diligence and bulk data-room analysis rather than standardized deal-workflow generation. | 中 | SP023, SP002 |
| CP008 | Hebbia has raised a cumulative $160 million across two funding rounds, reaching a $700 million valuation as of May 2026, including a $130 million Series B led by a16z in 2024; it was founded in 2020 by CEO George Sivulka and is headquartered in New York. | 中 | SP027, SP028 |
| CP009 | o11, an Excel-native AI layer vendor, characterizes Rogo and Hebbia together as 'Search and Synthesis Engines' whose browser-based architecture creates a 'Last Mile' problem: users must leave their spreadsheet, upload documents to a separate platform, and copy results back into their financial model or memo. | 低 | SP001 |
| CP010 | F2, a competing AI underwriting platform, reports that its native Excel formula engine scores 95.25% on the independently verified SpreadsheetBench Verified benchmark, while Hebbia's Matrix does not offer in-platform Excel formula evaluation and instead generates new models exportable to Excel via Financial Modeling Agents introduced in September 2025. | 中 | SP023 |
| CP011 | F2 provides a three-layer audit trail (claim to formula to source) for institutional deal teams, which it presents as a structural advantage over Hebbia's two-layer trail that traces claims to source documents but does not expose the underlying computation. | 低 | SP023 |
| CP012 | Bloomberg's BloombergGPT is a 50-billion-parameter large language model trained on a 363-billion-token dataset built primarily from Bloomberg's proprietary financial data (augmented with 345 billion general-purpose tokens), giving Bloomberg a multi-year, large-scale domain-specific training foundation that predates most AI-native finance startups. | 高 | SP006, SP005 |
| CP013 | Bloomberg's ASKB conversational AI interface, built directly into the Terminal, coordinates a network of AI agents that ground responses in Bloomberg data, provide the underlying Bloomberg Query Language code so answers can be extended in Excel or BQuant, and draw on proprietary research from Bloomberg Intelligence, BloombergNEF, and Bloomberg Economics alongside sell-side research from over 800 providers. | 中 | SP005 |
| CP014 | FactSet operates an industry-first Model Context Protocol server giving AI applications and agents direct, secure access to FactSet market data without custom integrations, and has launched 'FactSet AI for Banking,' a workflow-automation ecosystem built with Finster AI specifically for investment banking teams. | 高 | SP007, SP012 |
| CP015 | FactSet's Chief AI Officer Kate Stepp states the firm's AI strategy is 'anchored in building open, flexible, and secure solutions' and that FactSet is building on Anthropic, Google, and OpenAI models to enable natural-language access to FactSet data across its base of more than 9,000 clients and 241,000 individual users. | 中 | SP007 |
| CP016 | S&P Global's Kensho LLM-ready API supplies structured financial data (public and now private-company financials for over 12 million companies, plus Capital IQ estimates) directly to customer AI applications with source-document links for auditability, positioning S&P Global as a data-layer competitor rather than a workflow competitor to Rogo. | 中 | SP008 |
| CP017 | S&P Capital IQ Pro's ChatIQ, a generative-AI assistant co-developed with Kensho, is 'specifically tailored to support the needs of banking and buyside analysts' and enables company, industry, and sector research with full source traceability. | 中 | SP009 |
| CP018 | Rogo announced a strategic data partnership with LSEG under which LSEG's company fundamentals, estimates, and M&A database covering more than 1.5 million global transactions are integrated directly into Rogo's platform, with interoperability between LSEG Workspace and Rogo for customers holding a Workspace license. | 中 | SP018 |
| CP019 | LSEG simultaneously sells its own AI research agent, Deep Research, natively inside Workspace and Microsoft Teams, meaning LSEG functions as both a licensed data supplier to Rogo and a direct competitor offering an overlapping natural-language research capability on the same underlying content. | 中 | SP017, SP018 |
| CP020 | Rogo's own strategic-partnership announcement names Moelis, Nomura, and Tiger Global as customers trusting Rogo to 'work smarter, move faster, and outpace competitors,' and lists Thrive Capital, Khosla Ventures, J.P. Morgan Growth Equity Partners, and Tiger Global among its investors. | 中 | SP018 |
| CP021 | Microsoft Copilot had reached 15 million paid commercial seats with 160% year-over-year seat growth by mid-2026, and Forrester's Total Economic Impact study calculated a 353% ROI for small and medium business deployments, with Lloyds Banking Group measuring 46 minutes saved per employee per day. | 中 | SP010 |
| CP022 | Microsoft's Agent 365 orchestration platform for governing AI agents at scale reached general availability in May 2026 at $15 per user per month, while base Microsoft 365 license prices are set to increase across every tier in July 2026. | 中 | SP010 |
| CP023 | A Copilot data-loss-prevention bypass active from January 21 to February 3, 2026 allowed Copilot to process and summarize confidential emails in Sent Items and Drafts while ignoring sensitivity labels and DLP policies, a governance failure directly relevant to regulated financial institutions handling customer account and wire data. | 中 | SP010 |
| CP024 | Roughly 40% of organizations delayed their Microsoft Copilot rollout by three or more months over data-exposure concerns, according to Gartner 2025 research cited in a 2026 financial-institution deployment guide, indicating governance friction slows horizontal-copilot adoption in regulated finance relative to purpose-built vendors. | 中 | SP010 |
| CP025 | Glean announced an expanded financial-services MCP ecosystem in June 2026 integrating CB Insights, Crunchbase, Daloopa, FactSet, and S&P Global directly into its permission-aware enterprise search and agent platform, positioning Glean as a horizontal substitute that layers third-party financial data into a bank's existing context rather than shipping proprietary finance-specific agents. | 中 | SP012 |
| CP026 | An independent June 2026 analyst survey of 820 AI-platform decision makers found reliability and hallucination management (55%) and data privacy (53%) are the top two AI adoption challenges in regulated industries including financial services, the exact trust gap that both horizontal copilots and finance-specific vendors like Rogo must close before winning enterprise budget. | 中 | SP011 |
| CP027 | Harvey, an adjacent legal-AI platform, raised $200 million at an $11 billion valuation in March 2026 (up from an $8 billion valuation just months earlier), co-led by repeat investors Sequoia Capital and GIC, bringing its total funding to more than $1 billion. | 高 | SP013, SP014 |
| CP028 | Harvey's customer base includes more than 50 asset-management firms alongside the majority of the AmLaw 100 and 500+ in-house legal teams across 60 countries, and its funding will expand 'long-horizon agents' for complex, multi-step workflows including fund formation -- a banking-adjacent use case that could extend Harvey's agent infrastructure toward Rogo's buyer base. | 中 | SP013 |
| CP029 | Harvey reached $190 million in annual recurring revenue in January 2026, up from $100 million just months earlier, illustrating that adjacent vertical-AI entrants can scale revenue fast enough to fund expansion into neighboring professional-services categories such as banking and asset management. | 中 | SP014 |
| CP030 | Intapp DealCloud's Celeste agentic AI continuously analyzes relationships, communications, deal flow, and engagements across investment banking, private capital, legal, and consulting firms, automatically logging Outlook interactions (zero-entry activity capture) -- a deal-origination and relationship-management focus distinct from Rogo's research-and-analysis-generation focus. | 中 | SP015 |
| CP031 | Ensis Partners, a New York City restructuring-focused investment bank founded in February 2026 by veterans of PJT Partners, Perella Weinberg, Blackstone, and Citigroup, selected Intapp DealCloud with Celeste without evaluating any competing CRM platform, because both founders had prior hands-on experience deploying DealCloud at previous firms. | 中 | SP016 |
| CP032 | The Ensis Partners case illustrates a distribution-power dynamic relevant to Rogo's own category: dealmakers' prior personal familiarity with an incumbent platform from previous employers can eliminate a competitive evaluation entirely, favoring whichever vendor has the broadest existing footprint among banker alumni networks. | 低 | SP016 |
| CP033 | Independent analyst firm CB Insights lists Rogo among Daloopa's top alternatives and competitors alongside Fintool and Metal, indicating that even in the narrower financial-data-extraction niche, market analysts treat Rogo's broader platform as a substitute for point-solution data-extraction tools. | 中 | SP019 |
| CP034 | Marvin Labs' June 2026 comparison names Daloopa as 'best for financial model automation' with a free tier available, positioning it as a lower-cost, narrower-scope alternative to Rogo's enterprise agentic workflow platform for teams that need automated data extraction but not full deal-workflow generation. | 中 | SP002 |
| CP035 | AlphaSense prices across five quote-based tiers ranging from $10,000-$15,000 per user per year for core AI search up to $50,000-$100,000+ per year for Enterprise Intelligence team licenses, with a distinct $25,000-$50,000+ per user per year tier for its Expert Transcript Library (Tegus) aimed at hedge funds and PE firms doing primary research. | 中 | SP004, SP003 |
| CP036 | Marvin Labs' 2026 buyer's guide positions AlphaSense as 'best for broker research and expert calls' at enterprise pricing, a different core workflow from Rogo's positioning as best for 'deal teams in banking and PE' doing agentic CIM, comps, and memo work. | 中 | SP002 |
| CP037 | AlphaSense users in independent reviews consistently highlight the platform's accuracy and speed at delivering critical information versus comparable solutions, though pricing is not published and requires a custom sales quote. | 中 | SP024 |
| CP038 | The BankerToolBench benchmark, built by Handshake AI and McGill University with 502 active and former investment bankers from Goldman Sachs, JPMorgan, Morgan Stanley, and Evercore, found that none of nine tested AI models produced client-ready output without revision on standard junior-banker tasks, with the best model (GPT-5.4) achieving only 16% acceptable results and 27% of all outputs judged completely unusable. | 中 | SP020 |
| CP039 | A separate AA-Omniscience benchmark cited alongside BankerToolBench found GPT-5.5 carries an 86% hallucination rate despite leading performance rankings, compared with a 36% hallucination rate for Claude Opus 4.7, showing wide variance in reliability among the frontier models that AI research platforms in this category, including Rogo, may depend on. | 中 | SP020 |
| CP040 | JurisTech's 2026 LLM hallucination benchmark found models vary widely in their willingness to refuse an answer when presented with insufficient or contradictory financial data, concluding that model choice, prompt design, and workflow oversight all materially affect whether an AI tool fabricates a plausible-but-wrong answer rather than flagging the gap. | 中 | SP021 |
| CP041 | In July 2025, Deloitte Australia delivered a roughly A$440,000 generative-AI-assisted report to the Australian government that was later found to contain fabricated academic citations, forcing Deloitte to refund the final installment -- a documented real-world instance of AI fabrication reaching a paid professional-services deliverable. | 中 | SP022 |
| CP042 | Research cited by SP2 Analytics from Stanford and Anthropic found leading AI models affirm a user's stated view roughly 49% more often than a human would, meaning an analyst who challenges a correct AI answer with 'are you sure?' can cause the model to reverse itself into an incorrect answer -- a sycophancy risk relevant to any AI research tool used in an adversarial investment-committee setting. | 中 | SP022 |
| CP043 | Inline, source-linked citations are described across independent commentary as the single most effective mitigation for AI hallucination in investment research because they make errors auditable at the point of use rather than preventing errors outright -- a design principle Rogo, Bloomberg ASKB, LSEG Deep Research, and Kensho's API all claim to implement via source traceability. | 中 | SP022, SP008, SP017 |
| CP044 | Rogo's most durable moat elements are workflow-native integration into a bank's own systems (SharePoint, CRM, data rooms) plus licensed external data partnerships (LSEG, PitchBook, S&P Global), which create switching costs that a horizontal copilot cannot easily replicate without the same finance-specific integration work. | 中 | SP018, SP017 |
| CP045 | Rogo's moat is more exposed at the model layer than at the workflow layer: if frontier model providers or well-capitalized incumbents (Bloomberg, FactSet, S&P Global) close the finance-specific reasoning gap, Rogo's differentiation increasingly rests on distribution, data integration breadth, and customer trust rather than unique underlying model capability. | 中 | SP007, SP006 |
| CP046 | No public source reviewed discloses Rogo's per-seat or per-tier pricing, unlike AlphaSense, Marvin Labs, Microsoft's Agent 365, or Daloopa's free-tier offering, leaving pricing competitiveness as an open diligence question rather than a verifiable comparison point. | 低 | |
| CP047 | Across the seven buying criteria most relevant to enterprise finance buyers -- deal-workflow generation, document-intelligence depth, native Excel computation, licensed market-data breadth, horizontal office-suite integration, source-citation traceability, and banking-specific compliance posture -- no single vendor reviewed leads on every dimension: Rogo and Hebbia lead on deal-workflow and document depth respectively, F2 leads on native Excel computation, Bloomberg/FactSet/S&P Global lead on licensed data breadth, and Microsoft/Glean lead on horizontal office integration. | 中 | SP002, SP023, SP007, SP010 |
| CI001 | Forbes' company profile reports Rogo's revenue grew from approximately $2 million in 2024 to more than $15 million in 2025, a roughly 7x increase in disclosed historical revenue. | 中 | SI001 |
| CI002 | CB Insights separately lists Rogo's 2024 revenue at approximately $1 million, roughly half of the ~$2 million figure Forbes reports for the same year, indicating third-party revenue trackers do not agree even on Rogo's disclosed historical base. | 中 | SI008 |
| CI003 | OpenAI's official partner case study states that since emerging from stealth in 2024, Rogo has served more than 5,000 bankers and grown annual recurring revenue 27x. | 中 | SI002 |
| CI004 | An independent ZenML LLMOps technical review of Rogo's OpenAI case study cautions that the 27x ARR growth and 10+ hours saved per week figures are self-reported and promotional, and should be viewed with appropriate skepticism absent independent verification. | 中 | SI014 |
| CI005 | No source reviewed in this chapter discloses a per-seat or per-contract list price for Rogo's platform, consistent with the pricing evidence gap already identified for Rogo in the Competitors chapter. | 中 | SI012 |
| CI006 | A competitor comparison page from Novis frames Rogo as an 'enterprise custom pricing' product, contrasting it with Novis's own $25-per-month entry tier and positioning Rogo as a premium, negotiated-only price point. | 中 | SI012 |
| CI007 | Comparable enterprise research platform AlphaSense discloses subscription pricing of $10,000 to $20,000 per seat annually, with average enterprise deal sizes of $50,000 to $100,000 or more. | 中 | SI016 |
| CI008 | Comparable enterprise-knowledge platform Glean is estimated to price seats at $45 to $50-plus per user per month with a minimum annual commitment near $50,000 to $60,000, versus Microsoft 365 Copilot's $30 per user per month. | 中 | SI023 |
| CI009 | Industry analysis of AI-first B2B SaaS pricing argues flat per-seat pricing misaligns cost and value for AI products because a heavy user can generate roughly 100x the inference cost of a light user on an identical subscription fee, pushing vendors toward hybrid platform-fee-plus-consumption pricing models. | 中 | SI025 |
| CI010 | Rogo's own April 2026 Series D announcement states the platform is used by more than 35,000 financial professionals across 250-plus institutions, including Rothschild & Co, Jefferies, Lazard, Moelis, and Nomura. | 中 | SI004 |
| CI011 | Independent coverage of Rogo's January 2026 Series C reported the platform was then used by more than 25,000 professionals across 50-plus tier-one financial firms. | 中 | SI022 |
| CI012 | Comparing Rogo's disclosed user counts at the January 2026 Series C (~25,000 professionals, 50-plus firms) against the April 2026 Series D (35,000-plus professionals, 250-plus institutions) implies roughly 40% growth in named users but a much larger jump in institution count over three months, which likely reflects a broadened definition of institutional usage rather than only new paying enterprise accounts. | 低 | SI004, SI022 |
| CI013 | Growjo's algorithmic estimate lists Rogo at approximately 292 employees with 143% year-over-year headcount growth and estimated revenue per employee of $148,000, though Growjo's estimation methodology is not independently disclosed. | 低 | SI007 |
| CI014 | Rogo's live Ashby careers board listed 56 open roles as of the access date for this chapter, spanning engineering, sales, product, security, customer success, marketing, and finance functions across New York, London, and Singapore offices. | 中 | SI011 |
| CI015 | New York State's Empire State Development agency confirmed it is offering Rogo up to $6.5 million in performance-based Excelsior Jobs Program tax credits in exchange for 422 new full-time positions and nearly $14 million invested in Rogo's Manhattan headquarters, alongside roughly $40 million in planned R&D activity. | 中 | SI010 |
| CI016 | The Excelsior Jobs Program tax credits are structured so that public funds disburse only if Rogo meets the specified hiring and investment milestones, meaning the announced 422-role hiring target is an incentive-linked commitment rather than a guaranteed outcome. | 中 | SI010 |
| CI017 | A cross-vendor 2026 AI-agent productivity benchmark compiled from McKinsey, Gartner, Forrester, Bain, Deloitte, BCG, and MIT Sloan research, plus Q1 2026 vendor telemetry, finds a median of 6.4 hours saved per week, a 6.7-month average payback period, and only 41% of deployments hitting year-one ROI targets across production AI-agent deployments. | 中 | SI024 |
| CI018 | An independent LLMOps technical review describes Rogo's layered OpenAI model architecture -- GPT-4o for user-facing Q&A, o1-mini for data contextualization, and o1 reserved for high-stakes evaluation and reasoning -- as a deliberate cost-optimization strategy that limits use of the most expensive model tier. | 中 | SI014 |
| CI019 | 2026 GPU FinOps benchmarking estimates inference now consumes 55-80% of enterprise AI GPU spend, with cost-per-million-tokens ranging from roughly $1.67 on A100 GPUs to $4.54-plus on H200 GPUs, and a single 70-billion-parameter model serving typical enterprise traffic can incur roughly $347,000 per year in compute alone. | 中 | SI020 |
| CI020 | Industry analysis of AI-first B2B SaaS economics estimates gross margins of 55-70% for AI-native software companies, versus 78-85% for traditional SaaS, driven by variable inference COGS of 20-40% of revenue against under 5% of revenue for classic SaaS. | 中 | SI025 |
| CI021 | FactSet Research Systems' fiscal 2025 10-K, filed with the SEC on October 22, 2025 for the year ended August 31, 2025, together with aggregated third-party financial data drawn from that filing, show FactSet's trailing-twelve-month gross margin at approximately 52.7% on $2.32 billion of revenue. | 高 | SI017, SI018 |
| CI022 | No source reviewed in this chapter discloses Rogo's own gross margin, COGS composition, or model-inference spend, leaving the AI-first SaaS peer benchmark (55-70%) and the legacy data-platform benchmark (FactSet, ~52.7%) as the only available bounding references. | 低 | SI025, SI017 |
| CI023 | A competitor comparison page describes Rogo's implementation as 'white-glove banker-led,' with direct integrations into firms' SharePoint, CRM, and proprietary data rooms, implying materially higher service-delivery cost per enterprise account than a lighter-weight, self-serve competitor. | 中 | SI012 |
| CI024 | Two independent trackers put Rogo's cumulative funding raised through its April 2026 Series D at approximately $310-314 million across six rounds: CB Insights reports $310.5 million and Growjo reports $314 million, both consistent with the 'more than $300 million' total already established in Company Overview. | 高 | SI007, SI008 |
| CI025 | An SEC Form D filed July 25, 2024 by 'Rogo Series A ScOp SPV Jul 2024, a Series of CGF2021 LLC,' a Delaware entity claiming a Section 3(c)(1) private-fund exemption, confirms that at least one Series A investor routed its investment through a special-purpose vehicle rather than investing directly in Rogo. | 中 | SI009 |
| CI026 | The SEC Form D filing for the Rogo Series A SPV does not itself disclose Rogo's Series A valuation or total round size, since Form D reporting for a feeder special-purpose vehicle covers the feeder's own securities offering rather than the underlying portfolio company's terms. | 中 | SI009 |
| CI027 | Rogo's official April 2026 Series D announcement states the capital will be used to deepen institutional partnerships, scale the Felix agentic platform, and accelerate global expansion, without disclosing a specific dollar allocation across those uses. | 中 | SI004 |
| CI028 | Coverage of Rogo's January 2026 Series C states the proceeds were directed at accelerating growth in Europe, expanding R&D capacity, and supporting North American partners with cross-border operations, at a time when Rogo employed just over 100 people. | 中 | SI027 |
| CI029 | No source reviewed discloses Rogo's cash on hand, monthly burn rate, or runway in months, making capital-adequacy assessment dependent on inference from total funds raised and disclosed hiring and expansion commitments rather than direct financial statements. | 低 | |
| CI030 | Rogo's disclosed hiring and facilities commitment tied to its Series D period -- 422 new roles, roughly $14 million of headquarters capital expenditure, and roughly $40 million of planned R&D spend -- implies a capital-intensive near-term growth phase that will consume a material share of the roughly $310 million raised to date well before any visible profitability inflection. | 中 | SI010 |
| CI031 | Axios Pro's January 28, 2026 exclusive independently corroborates that Sequoia led Rogo's Series C at a $750 million valuation, matching the figure already established for that round in Company Overview. | 中 | SI026 |
| CI032 | SixThirty Ventures' analyst commentary argues that AI-analyst-category fundraises, including Rogo's, were pricing in 'outlier-level revenue multiple premia' relative to the still-developing recurring-revenue traction most companies in the category had shown as of 2024. | 中 | SI005 |
| CI033 | Finro's Q1 2026 AI valuation database finds investors are no longer pricing AI companies as a single category, instead rewarding monetization clarity, margin quality, and durability with premium multiples while repricing companies still selling narrative over demonstrated unit economics. | 中 | SI019 |
| CI034 | Qubit Capital's 2026 valuation-multiple analysis finds most AI startups trade at 10x to 50x revenue with a median around 20x-30x, that late-stage rounds specifically averaged roughly 25.8x, and that category-defining infrastructure leaders can clear 40x-100x. | 中 | SI021 |
| CI035 | Comparable AI-native enterprise research platforms show a wide valuation-multiple band: AlphaSense trades at roughly 12.5x its disclosed $600 million 2026 ARR at a $7.5 billion valuation, while Glean trades at roughly 36x its disclosed $200 million 2026 ARR at a $7.2 billion valuation. | 高 | SI015, SI016, SI023 |
| CI036 | Because Rogo has not disclosed current ARR, applying the 12.5x-36x comparable multiple band to Growjo's low-confidence $43.2 million ARR estimate would imply a Rogo multiple near 46x-93x on its $2 billion valuation -- above every comparable cited in this chapter -- illustrating how sensitive any multiple conclusion is to an unverified third-party estimate rather than a company-confirmed figure. | 低 | SI007 |
| CI037 | Growjo's estimate of $43.2 million in annualized revenue for Rogo as of mid-2026 is an algorithmically generated third-party figure without disclosed methodology transparency and should not be treated as company-confirmed ARR. | 低 | SI007 |
| CI038 | Rogo's disclosed historical revenue growth (Forbes: roughly $2 million in 2024 rising to more than $15 million in 2025) confirms strong percentage growth off a small base but does not by itself establish a current 2026 annual recurring revenue figure sufficient to underwrite the Series D valuation without additional company disclosure. | 中 | SI001 |
| CI039 | Rogo's layered use of OpenAI's GPT-4o, o1-mini, and o1 models mirrors the broader industry pattern of tiered model routing to manage inference costs described in 2026 GPU FinOps benchmarking, suggesting Rogo's cost-management approach is directionally consistent with industry best practice even though its actual COGS remain undisclosed. | 中 | SI014, SI020 |
| CI040 | A 2026 cross-vendor AI-agent productivity benchmark reports a median of 6.4 hours saved per week across production deployments, meaningfully below the 10-plus hours per week that Rogo and its OpenAI case study claim for its own users, a gap that warrants independent verification rather than acceptance at face value. | 中 | SI024, SI002 |
| CI041 | Rogo's product mechanics -- a core research/workflow subscription combined with the Felix agentic add-on and forward-deployed, banker-led implementation support -- imply at least three inferred revenue streams (platform licensing, agentic workflow expansion, and services/implementation), though no source discloses the revenue mix across them. | 低 | SI012 |
| CI042 | Across this chapter's review, the categories of Rogo financial data that remain undisclosed and block full underwriting are: audited GAAP revenue/ARR, gross margin and COGS breakdown, cash balance and burn rate, per-seat/contract pricing, customer concentration and net revenue retention, and formal revenue-recognition policy. | 低 | |
| CE001 | Felix is Rogo's AI agent for finance that turns a single prompt into client-ready PowerPoint decks, Excel models, Word documents, dashboards, and sourced research. | 中 | SE001 |
| CE002 | Felix operates as a model-agnostic orchestration harness that routes tasks across multiple frontier LLM vendors, including OpenAI, Anthropic, and Google models, rather than running on a single fine-tuned model. | 高 | SE001, SE014 |
| CE003 | Rogo Agents let firms encode proprietary templates, methodologies, formatting standards, and recurring workflows as reusable custom automations that run on demand across any deal, company, or dataset. | 中 | SE001 |
| CE004 | Felix supports asynchronous, email-initiated workflows: users can delegate research, monitoring, and document-production tasks to Felix by email and receive scheduled or on-demand deliverables. | 高 | SE001, SE015 |
| CE005 | Felix's concrete finance workflows include deal screening, Confidential Information Memorandum generation, buyer outreach, and data-room diligence for M&A processes. | 中 | SE014, SE024 |
| CE006 | Rogo's Excel Plug-in is a native Microsoft Excel add-in that lets bankers query, build, audit, and stress-test models directly inside a workbook, grounded in Rogo's data sources and the firm's own templates. | 中 | SE001 |
| CE007 | Rogo's May 2026 release added a Custom MCP capability that lets firms connect any Model Context Protocol server, including self-built ones, as an extension of Rogo's tool layer. | 中 | SE001 |
| CE008 | Rogo's May 2026 release added a Slides Annotator that lets users mark up a deck directly in Rogo and have Felix execute the revisions, with every iteration versioned and trackable. | 中 | SE001 |
| CE009 | Rogo's Memory feature retains a user's role, formatting conventions, and long-running preferences across chats, and is presented to users as auditable and user-editable. | 中 | SE001 |
| CE010 | Rogo's Library feature centralizes every artifact Felix has produced for a user, including presentations, PDFs, HTML, and Excel files, in one place. | 中 | SE001 |
| CE011 | Rogo's platform integrates external financial-data providers, including LSEG, S&P Capital IQ, FactSet, and PitchBook, alongside internal CRM, communications, and file-repository connectors. | 高 | SE001, SE006 |
| CE012 | Rogo's strategic partnership with LSEG, announced in 2025, gives Workspace-licensed customers real-time access to LSEG company fundamentals, estimates, and an M&A database spanning more than 1.5 million global transactions. | 高 | SE002, SE008, SE018 |
| CE013 | Rogo integrated S&P Capital IQ data, including earnings transcripts, fundamentals, consensus estimates, and real-time market data, into its AI-powered workflows in December 2024. | 高 | SE011, SE019 |
| CE014 | Rogo deepened its PitchBook integration in 2026 with PitchBook Premium, giving in-platform access to full company profiles, financing histories, cap tables, and investor portfolios. | 高 | SE001, SE003 |
| CE015 | Rogo's May 2026 release added connectors for Affinity, Microsoft Teams, Moody's, Daloopa, Dropbox, Granola, and Slack, bringing relationship, communications, credit-ratings, and meeting-notes data into the platform. | 中 | SE001 |
| CE016 | Rogo's partnership with Daloopa brings structured, audit-ready financial data into Rogo's workflows, with every datapoint linked back to its original public source. | 中 | SE026 |
| CE017 | An independent review lists SharePoint document ingestion, Salesforce CRM data, and in-cell citations in generated spreadsheets among Rogo's differentiating integrations. | 中 | SE016 |
| CE018 | Rogo's public GitHub organization publishes infrastructure tooling, including a Terraform provider and a Google Cloud on-premises deployment scaffold, indicating an infrastructure-as-code, multi-environment deployment posture. | 中 | SE010 |
| CE019 | In June 2026, Microsoft named Rogo a launch partner for partner-built skills in Copilot for Excel, alongside LSEG, Ramp, and Vena, to be distributed through Microsoft Marketplace starting in the third quarter of 2026. | 中 | SE009 |
| CE020 | Microsoft's June 2026 Copilot-in-Excel update also added its own live data connectors for CB Insights, Daloopa, FactSet, Morningstar, PitchBook, and S&P Global, meaning several of Rogo's own data partners are simultaneously integrating directly with Microsoft. | 中 | SE009 |
| CE021 | Rogo publishes an open reference harness for its Big Finance benchmark on GitHub, including a ReAct agent scaffold, four public tools (web search, SEC EDGAR search, URL fetch, sandboxed Python execution), and a 50-item public data subset. | 中 | SE013 |
| CE022 | The companion research paper BigFinanceBench documents 928 expert-authored, workflow-grounded financial-research tasks scored against more than 36,000 rubric points that check each step of the derivation rather than only the final answer. | 高 | SE023, SE013 |
| CE023 | Evaluating ten frontier and open-weight AI agents on BigFinanceBench, the best-performing system reached only 58.8 percent of available rubric points, showing substantial headroom in auditable financial-derivation quality even for leading models. | 高 | SE023, SE013 |
| CE024 | The BigFinanceBench authors find that final-answer accuracy is a lossy proxy for derivation quality and that model capability varies non-uniformly across the stages of a financial-research workflow. | 中 | SE023 |
| CE025 | Rogo's own published benchmark harness deliberately excludes vector-store retrieval and premium financial-data sources such as FactSet, Capital IQ, and Bloomberg so the evaluation measures the underlying model rather than the tool scaffold. | 中 | SE013 |
| CE026 | An independent technical case study finds Rogo runs a tiered multi-model architecture: a primary model for chat-based analysis, a smaller model for data contextualization and search structuring, and a top-tier model reserved for evaluation and synthetic data generation. | 中 | SE007 |
| CE027 | The same independent case study reports that Rogo's platform searches and analyzes more than 50 million financial documents and uses a human-in-the-loop labeling process staffed by former bankers and investors to improve model output quality. | 中 | SE007 |
| CE028 | That same analysis cautions that Rogo's headline growth and productivity metrics, including bankers served, hours saved weekly, and annual recurring revenue growth multiples, are self-reported by the company and were not independently verified in the source material it reviewed. | 中 | SE007 |
| CE029 | The specifics of Rogo's fine-tuning methodology, including dataset composition, training approach, and how model updates from upstream vendors are absorbed without disrupting production behavior, are not detailed in any public source reviewed for this chapter. | 低 | |
| CE030 | Rogo operates a second internal agent, Sisyphus, which runs automated offensive-security campaigns against Rogo's own infrastructure roughly once or twice a day, chaining findings across authentication abuse, authorization bypass, injection, SSRF, and LLM-specific exploit classes. | 中 | SE014 |
| CE031 | Rogo holds SOC 2 Type II certification, alongside SOC 2 Type I, validating security, availability, processing integrity, and confidentiality controls over customer data. | 高 | SE012, SE014 |
| CE032 | Rogo achieved ISO/IEC 42001:2023 certification, the first international standard for AI management systems, covering enterprise copilots, model training and orchestration pipelines, data integrations, and internal AI-governance structures. | 高 | SE004, SE012 |
| CE033 | Rogo holds ISO/IEC 27001 certification for its information security management system, in addition to its SOC 2 and ISO 42001 certifications. | 高 | SE012, SE014 |
| CE034 | Rogo's Trust Center lists CCPA alignment, a VPAT accessibility conformance report, and EU AI Act compliance documentation alongside its SOC 2 and ISO certifications. | 中 | SE012 |
| CE035 | Rogo's Trust Center names Jefferies, Truist, Rothschild & Co, Raymond James, Nomura, Tiger Global, Moelis, and Lazard as institutions that reviewed and trust Rogo's security posture. | 中 | SE012 |
| CE036 | From August 2, 2026, the EU AI Act's obligations for high-risk AI systems become fully enforceable across the EU, requiring documented risk management, data governance, technical documentation, human oversight, and post-market monitoring. | 中 | SE022 |
| CE037 | Rogo's public status page reports 100 percent API uptime and 99.97 percent application uptime for the April-to-June 2026 window, with one partial outage of 502 errors on specific dedicated tenants resolved within roughly 18 minutes on April 30, 2026. | 中 | SE025 |
| CE038 | Rogo's Sisyphus security agent identified 18 additional exploitable vulnerabilities in a single afternoon within a week of a third-party penetration test, with high-confidence findings later calibrated to a greater than 95 percent true-positive rate against Rogo's own human security team. | 中 | SE014 |
| CE039 | Rogo acquired Subset in 2025; Subset's spreadsheet agent understands complex financial-model formulas and ranges and connects to Capital IQ, FactSet, PitchBook, LSEG, and firm-private data to build, roll forward, and audit models. | 中 | SE005 |
| CE040 | Subset was founded by Jason Chan, a former investment banker at Bank of America and investor at Providence Equity Partners, and AJ Nandi, a former investor at Insight Partners, and was backed by Index Ventures before its acquisition. | 中 | SE005 |
| CE041 | Rogo acquired Offset, announced in March 2026; Offset's agentic systems build memory of how a specific financial model's assumptions, formulas, and outputs evolve over time so the model can be maintained automatically as new information arrives. | 高 | SE006, SE024 |
| CE042 | Offset was founded by Raj Khare and Shiv Shrivastava; its technology is being integrated into the Rogo platform, which served more than 25,000 finance professionals as of the March 2026 acquisition announcement. | 中 | SE006 |
| CE043 | Rogo acquired UK-based Plux AI in early 2026; Plux's systems monitor filings, lender updates, court documents, and company disclosures to surface long-form market signals, expanding Rogo's coverage and engineering presence across UK and European markets. | 高 | SE017, SE024 |
| CE044 | Plux AI was founded by Deepak Guneja, formerly of D.E. Shaw and Morgan Stanley, and Pratyush Chaudhary, formerly of Google and D.E. Shaw, who joined Rogo to help build out its European engineering presence. | 中 | SE017 |
| CE045 | Rogo's go-to-market model relies on Forward Deployed Bankers, ex-finance professionals embedded inside client institutions who onboard teams from analyst to managing director and translate firm-specific workflow requirements into product configuration. | 中 | SE014, SE024 |
| CE046 | Rogo's founder has stated publicly that Rogo sits at an unusual intersection as simultaneously a customer, a distribution partner, and a potential competitive target of OpenAI and Anthropic, since both model vendors are separately expanding into finance-facing AI products. | 高 | SE021, SE014 |
| CE047 | An independent review of Rogo notes that deployment typically requires significant setup, including data-source integration, permissions mapping, and template configuration, often with help from Rogo's own team, and that the platform is priced for large enterprise clients rather than smaller firms. | 中 | SE016 |
| CE048 | Commentary on agentic research tools for finance, discussing Felix specifically, argues that AI agents cannot fully replace financial analysts because the cost of hallucination or data inaccuracy in finance is extreme, and that firms should verify an agent can cite the specific page, document, or data point behind each generated insight before adoption. | 中 | SE020 |
| CE049 | Kevin Buehler, formerly of McKinsey where he led the investment-banking practice and co-founded its global risk practice, joined Rogo as Chief Innovation Officer in 2026 to help embed Felix into institutional client workflows. | 中 | SE015 |
| CE050 | No public source reviewed for this chapter independently corroborates the greater than 95 percent true-positive rate Rogo reports for its Sisyphus security-testing agent; the figure comes only from a single independent technical analysis restating Rogo's own account. | 低 | |
| CU001 | Rogo's official customers page displays a "Trusted by the world's leading financial institutions" banner with individually labeled logos for Jefferies, Lazard, Moelis, Nomura, Rothschild & Co, and Truist, and an independent 2026 funding writeup names the same five bulge-bracket/global names as active users. | 高 | SU002, SU008 |
| CU002 | Rogo's April 2026 Series D press release states the platform is trusted by professionals at investment banks, private equity firms, and asset managers, naming three primary buyer segments. | 中 | SU017 |
| CU003 | Forbes' company profile separately describes Rogo's buyers as investment banks, private equity firms, and hedge funds, substituting hedge funds for asset managers relative to Rogo's own materials. | 中 | SU020 |
| CU004 | Dakota's February 2026 investment-bank league table shows deal volume concentrated among both bulge-bracket-adjacent consortium advisors and independent boutique or mid-market firms, the two investment-banking sub-segments spanned by Rogo's named logos. | 中 | SU007 |
| CU005 | Jefferies' own published interview with Rogo's CEO frames the core deployment challenge as integrating AI onto "live real deal data" inside Chinese-wall- and MNPI-controlled environments, distinguishing bulge-bracket procurement from a generic software sale. | 中 | SU022 |
| CU006 | FeaturedCustomers aggregates a boutique-advisory-bank analyst testimonial about trusting Rogo's information and a $300B-AUM asset-management senior managing director testimonial about time savings, indicating usage beyond the six named logo accounts. | 中 | SU004 |
| CU007 | 2026 Series C coverage names GTCR, a private-equity firm, alongside Rothschild, Jefferies, Lazard, Moelis, Nomura, and Truist Securities as a Rogo client, giving the private-equity segment a named account distinct from the investment-banking logo wall. | 中 | SU011 |
| CU008 | No fetched source discloses a segment-by-segment breakdown of Rogo's 35,000+ users (e.g., what share are investment bankers versus private-equity, hedge-fund, or asset-management professionals). | 低 | |
| CU009 | OpenAI's 2024 partner case study states Rogo had served "over 5,000 bankers" and grown annual recurring revenue 27x since emerging from stealth in 2024, establishing an early adoption baseline. | 中 | SU018 |
| CU010 | Forbes' company profile, dated to the April 2026 Series D window, states Rogo has "more than 25,000 users across 150 firms" on a per-seat subscription basis. | 中 | SU020 |
| CU011 | Rogo's own April 29, 2026 Series D press release and an independent funding writeup both state more than 35,000 financial professionals across 250-plus institutions use the platform, a higher figure than the 25,000/150 figure Forbes' profile still displays. | 高 | SU017, SU008 |
| CU012 | Series C coverage from January 2026 states Rogo served roughly 25,000 professionals processing 50,000 daily queries and estimates the platform saved the equivalent of 500 years of human work. | 中 | SU011 |
| CU013 | Forbes' April 2026 profile still displays the ~25,000-user/150-institution figure associated with Rogo's January 2026 Series C window even though Rogo's own April 29, 2026 Series D release already reports 35,000-plus users at 250-plus institutions, indicating at least one public profile had not refreshed to the newer figure at the time of this run. | 高 | SU020, SU017 |
| CU014 | An independent Series D recap states more than 100 analysts at Baird run in excess of 10,000 workflows weekly with 95% platform engagement, a different metric set than the weekly/daily active-usage percentages Rogo's own Baird case study discloses. | 中 | SU016 |
| CU015 | Rogo's own published Baird Equity Research case study states more than 100 professionals are active on the platform, with approximately 85% weekly active usage and nearly 70% daily active usage, without disclosing a weekly workflow-volume figure. | 中 | SU001 |
| CU016 | FeaturedCustomers separately cites "Baird executes 250K+ AI workflows with Rogo" as its case-study headline metric, a third figure for the same deployment that does not cleanly reconcile with either Rogo's or the independent Series D recap's stated numbers. | 中 | SU004 |
| CU017 | Raw page markup on Rogo's official customers page contains discrete, individually named logo placements for Jefferies, Lazard, Moelis, Nomura, Rothschild&Co, and Truist, and an independent 2026 funding article separately lists the same five non-Truist names as active users, corroborating each as a genuine reference logo. | 高 | SU002, SU008 |
| CU018 | Jefferies' own corporate website -- not Rogo's marketing surface -- published a leadership-spotlight interview with Rogo's CEO describing a live deployment on real deal data, an independent-domain corroboration that Jefferies is an active customer rather than a marketing-only logo placement. | 高 | SU022, SU002 |
| CU019 | Truist Securities CEO Tom Hackett is quoted on the record describing "successful integration, boosted productivity, reduced risk, and increased capacity for bankers to focus on relationships and growth" from deploying Rogo. | 中 | SU011 |
| CU020 | 2026 Series C funding coverage names GTCR alongside Rothschild, Jefferies, Lazard, Moelis, Nomura, and Truist Securities as Rogo clients, the only fetched source to name a private-equity-segment account. | 中 | SU011 |
| CU021 | Truist Ventures appears among Rogo's Series C institutional investors in the same funding coverage that names Truist Securities as a client with an on-the-record CEO quote, repeating the dual investor-customer relationship pattern already identified for J.P. Morgan in this company's earlier chapters. | 高 | SU011, SU002 |
| CU022 | Anthropic's own customer story quotes Rogo's Head of Product, Strib Walker, describing a requirement to produce "structured PowerPoint and Excel output at institutional quality, not a generic approximation of it," functioning as partner-level rather than end-client proof of institutional-grade deployment. | 中 | SU014 |
| CU023 | Google Cloud's case study states that many of Rogo's own clients have confided that they trust Google more than other AI vendors for security features and regulatory preparedness, an indirect signal of end-customer vendor-trust sensitivity relayed through a partner rather than a named client. | 中 | SU015 |
| CU024 | Bloomberg reporting notes that skeptics characterize Rogo as an "unnecessary layer" because finance professionals could use large AI models directly, a competitive-necessity critique surfacing from within the same finance-professional customer base Rogo sells to. | 中 | SU021 |
| CU025 | None of the fetched sources disclose a signed, quantified reference account beyond Baird's Equity Research division; the six-to-eight named or quoted logos identified across all fetched sources represent a small fraction of the 250-plus institutions Rogo claims as customers. | 低 | |
| CU026 | No fetched source discloses Rogo's net revenue retention, gross revenue retention, logo churn rate, renewal rate, or average contract length. | 低 | |
| CU027 | An anonymous poster on the Wall Street Oasis finance forum states their firm "churned from it after ~1 year," calling the platform "the most overhyped piece of shit I've ever used," the only concrete public churn account found for Rogo. | 低 | SU005 |
| CU028 | Other participants in the same Wall Street Oasis thread describe Rogo as "much better now" since the Felix launch and report colleagues who "love it," directly contradicting the churn account within the same discussion. | 低 | SU005 |
| CU029 | A self-identified director-level poster in the same thread states Rogo is "utterly useless for most things" from vice-president level upward, while other posters describe daily reliance at analyst and associate levels, suggesting satisfaction may vary by seniority. | 低 | SU005 |
| CU030 | Baird's 85% weekly active usage and 70% daily active usage figures are the only quantified repeat-usage proxy found across fetched sources for any named Rogo account. | 中 | SU001 |
| CU031 | AlphaSense's own competitor-comparison page claims superior collaboration tooling and a larger dedicated support team than Rogo without disclosing any comparable Rogo usage or satisfaction data, illustrating that public retention signal for Rogo comes mostly from competitor or anecdotal sources rather than the company itself. | 低 | SU006 |
| CU032 | A Series C funding writeup frames Rogo's dependency on third-party foundation-model providers (OpenAI, Google) as a company-level risk that could expose customer workflows to model pricing volatility or access disruption. | 中 | SU011 |
| CU033 | An independent Series D recap states that heavy reliance on a single AI vendor for critical deal workflows creates concentration risk for the institutions adopting Rogo, a customer-side vendor-lock-in framing distinct from Rogo's own model-provider dependency. | 中 | SU016 |
| CU034 | LSEG's press release states the partnership gives "Rogo customers globally with a Workspace license" real-time access to LSEG's flagship data and M&A database, making part of Rogo's customer value proposition contingent on this data-licensing relationship. | 中 | SU023 |
| CU035 | Rogo's own announcement of a deepened PitchBook partnership states the integration serves "shared customers" jointly licensed by both companies, a second named data-partner dependency behind Rogo's customer-facing workflows. | 中 | SU024 |
| CU036 | Daloopa's partnership announcement adds a third named data-provider dependency layered into Rogo's customer-facing workflows alongside LSEG and PitchBook. | 中 | SU025 |
| CU037 | Rogo's customer value proposition is contingent on maintaining simultaneous data-licensing relationships with at least LSEG, PitchBook, and Daloopa, a multi-partner supply-side dependency that sits alongside, not instead of, its own model-provider concentration risk. | 中 | SU023, SU024, SU025 |
| CU038 | An analyst recap states an eight-month period from initial conversation to production deployment is industry-standard for institutional banks adopting Rogo, indicating meaningful enterprise sales-cycle friction ahead of any expansion or renewal decision. | 中 | SU009 |
| CU039 | The same analyst recap warns that data incumbents (Bloomberg, S&P Capital IQ, FactSet, Refinitiv) could close the competitive gap by 2027 if Rogo does not cement its deal-room position, a retention risk tied to competitive substitution rather than product failure. | 中 | SU009 |
| CU040 | Deloitte's 2026 banking-agent risk guidance states banks must build agent registries, immutable audit trails, and disclosure layers before scaling agentic deployments, describing the governance buildout that gates a Rogo-style procurement inside a regulated bank. | 中 | SU013 |
| CU041 | Jefferies' CEO interview states that deploying AI on live deal data requires jointly building compliance infrastructure covering Chinese walls, MNPI controls, and role-based access, describing procurement friction specific to bulge-bracket customers. | 中 | SU022 |
| CU042 | An independent Series D recap states that smaller and mid-market banks gain access to analytical capabilities "previously the exclusive advantage of bulge-bracket firms," describing an expansion vector into the boutique/middle-market segment. | 中 | SU016 |
| CU043 | Rogo's Trust Center publishes formal Acceptable Use, Access Control, and Asset Management policies as customer-facing procurement collateral, indicating a formal trust package built to reduce security-review friction in enterprise sales cycles. | 中 | SU026 |
| CU044 | No fetched source discloses Rogo's revenue concentration among its largest customer or top accounts. | 低 | |
| CU045 | No fetched source confirms whether any named logo customer -- Jefferies, Lazard, Moelis, Nomura, Rothschild & Co, Truist Securities, GTCR, or Baird -- has expanded seats, renewed, or reduced usage since its initial deployment. | 低 | |
| CU046 | Rogo's Series D announcement states the new capital funds "more forward-deployed bankers and engineers embedded within the firms we serve" as the core post-sale deployment model. | 中 | SU003 |
| CU047 | An independent Series D recap describes Rogo's "Forward Deployed Banker" model explicitly: former finance professionals who physically embed inside partner institutions to drive adoption, drawing an explicit parallel to Palantir's Forward Deployed Engineers. | 中 | SU016 |
| CU048 | Rogo's own Baird case study states its team conducted in-person one-on-one sessions at Baird's Milwaukee headquarters alongside office hours and direct outreach to individual users, describing the concrete on-site deployment motion behind the Baird account. | 中 | SU001 |
| CU049 | OpenAI's partner case study attributes Rogo's "deployment team of ex-bankers and investors" working with customers to refine features in real time as central to the company's onboarding motion. | 中 | SU018 |
| CU050 | Anthropic's customer story quotes Rogo's Head of Product describing a shift from "skepticism to excitement" among users once a workable output is validated, describing a change-management adoption pattern rather than an instant-adoption one. | 中 | SU014 |
| CR001 | BigFinanceBench, a 928-item benchmark scoring 36,241 derivation-level rubric points across ten frontier and open-weight financial-research agents, finds the best-performing system reaches only 58.8% of available rubric score, co-published by Rogo's own research team. | 中 | SR025 |
| CR002 | Final-answer accuracy is a lossy proxy for derivation quality, meaning a Rogo/Felix output can look complete and well-cited while embedding a wrong period, accounting definition, or unstated assumption. | 中 | SR025 |
| CR003 | No independent, Rogo-specific benchmark of citation-to-source granularity (page-level vs. paragraph-level vs. dataset-level precision) was found among sources reviewed for this chapter. | 低 | |
| CR004 | FINRA's 2026 Annual Regulatory Oversight Report names hallucination as a top-line GenAI risk and states that firms relying on GenAI within a supervisory system must weigh the model's integrity, reliability, and accuracy. | 高 | SR001, SR002 |
| CR005 | Trade press covering the 2026 FINRA report frames the regulatory message as brokerage regulators urging firms to be vigilant for hallucination risk in day-to-day GenAI use. | 中 | SR002 |
| CR006 | Independent technical coverage describes Rogo's Felix product as an orchestration harness that routes tasks across multiple frontier models (OpenAI, Anthropic, Google) rather than a single fine-tuned model, pushing reliability engineering onto Rogo's internal evaluation harness. | 中 | SR025 |
| CR007 | Deloitte Australia had to partially refund a AU$440,000 government contract in late 2025 after outside academics found fabricated references and an invented Federal Court quotation generated by an undisclosed GPT-4o workflow that was not caught before delivery. | 高 | SR013, SR034 |
| CR008 | The Deloitte Australia incident is a category precedent showing that a reputable professional-services firm's own internal review process can fail to catch AI-fabricated content in a paid client deliverable -- the same failure mode a forward-deployed human-in-the-loop model is meant to prevent. | 中 | SR013 |
| CR009 | US courts imposed at least $145,000 in sanctions in the first quarter of 2026 alone for AI-hallucinated case citations in legal filings, part of a broader 2026 escalation in court sanctions for fabricated AI-generated content. | 中 | SR012, SR028 |
| CR010 | The SEC has brought enforcement actions against investment advisers, public issuers, and at least one startup founder since March 2024 for overstated or misleading 'AI washing' claims, and expects AI-related statements to be substantiated with the same rigor as performance or risk-factor disclosures. | 高 | SR004, SR024 |
| CR011 | In December 2025 the SEC's Investor Advisory Committee voted to recommend that issuers be required to define 'artificial intelligence,' disclose board oversight of AI deployment, and report separately on internal and consumer-facing AI effects. | 高 | SR005, SR026, SR031 |
| CR012 | Amended SEC Regulation S-P requires broker-dealers to document oversight of third-party AI vendors, including 72-hour vendor breach notification and 30-day customer notification, with compliance deadlines of December 2025 (large firms) and June 2026 (smaller firms). | 中 | SR023 |
| CR013 | FINRA's 2026 report reiterates that its rules are technology-neutral: supervision (Rule 3110), communications, recordkeeping, and fair-dealing obligations apply to GenAI-assisted output exactly as they would to human-produced output, and firms must archive prompts, outputs, and model-usage logs. | 高 | SR001, SR002 |
| CR014 | Felix-generated research and models used by Rogo's bank customers are themselves subject to FINRA supervisory review and recordkeeping obligations, an operational burden that could slow adoption or require additional Rogo-side audit tooling. | 中 | SR001 |
| CR015 | The EU AI Act's GPAI Code of Practice took effect August 2, 2025 with AI Office enforcement powers (information requests, model access, recalls) beginning August 2, 2026, and full high-risk-system obligations for AI used in credit, AML, and similar financial workflows also become enforceable from August 2, 2026. | 高 | SR006, SR030 |
| CR016 | Rogo's own Trust Center lists 'EU AI Act' among its compliance documentation, but no source reviewed for this chapter shows independent third-party verification of that claim by a named auditor or the EU AI Office. | 中 | SR022 |
| CR017 | The April 2026 interagency model risk management guidance from the OCC, Federal Reserve, and FDIC rescinds the 2011 SR 11-7-era framework in favour of a lighter, principles-based approach, and explicitly excludes generative and agentic AI models from its scope pending separate future guidance. | 高 | SR003, SR008 |
| CR018 | Investment-bank use of third-party AI vendors for confidential deal data (potential MNPI) is currently governed through Reg S-P vendor-oversight rules and bank-vendor contract terms rather than an AI-specific rule, placing the burden of proof on vendor data-segregation and model-training-exclusion architecture. | 中 | SR023, SR024 |
| CR019 | A tool that surfaces comps, models, and recommendations for real transactions sits close to the boundary between 'research tool' and 'investment advice,' implicating adviser fiduciary and suitability obligations if a firm's compliance program does not clearly document that human judgment, not AI output, is the basis for client-facing advice. | 中 | SR027 |
| CR020 | No source reviewed for this chapter identifies a pending lawsuit, SEC or FINRA enforcement action, or confirmed data breach naming Rogo specifically as of the 2026-07-01 run date. | 中 | SR018, SR019 |
| CR021 | AI-related securities class actions grew roughly 100% year-over-year from 2023 to 2024 and continued growing through 2025 into 2026, and legal-industry trackers expect event-driven AI litigation, reinforced by the SEC's Cybersecurity and Emerging Technologies Unit, to remain the dominant private-securities-litigation category through 2026. | 高 | SR018, SR019 |
| CR022 | Rogo's Felix harness routes tasks across OpenAI, Anthropic, and Google models based on internal benchmark performance rather than relying on a single proprietary foundation model. | 中 | SR025 |
| CR023 | Rogo's data layer depends on a small number of licensed providers, principally LSEG (fundamentals, consensus estimates, M&A data, partnership announced 2025) and PitchBook (private-capital deal, fund, and investor data, partnership deepened through 2025-2026), with contract renewal, pricing, and exclusivity terms not publicly disclosed. | 中 | SR033 |
| CR024 | Financial data providers are increasingly building first-party generative-AI features directly into their own terminals, which could lead LSEG, PitchBook, FactSet, or S&P to restrict wholesale API access in ways that would impair Rogo's data-grounded value proposition. | 中 | SR033 |
| CR025 | Microsoft's 2026 roadmap folds finance-specific Copilot agents (variance analysis, DCF construction, reusable finance skills, live data connectors) directly into the core Microsoft 365 Copilot subscription at no additional per-seat cost starting October 2026, alongside a new Copilot Agent Store. | 中 | SR016 |
| CR026 | Because most of Rogo's target banks and asset managers are already Microsoft 365 customers, Microsoft's Copilot bundling lowers the switching-cost barrier for the commoditizable slice of Rogo's workflow even though it does not yet match Rogo's forward-deployed, firm-specific configuration depth. | 中 | SR016 |
| CR027 | Independent competitor commentary frames Rogo and Hebbia as leaders on citation-linked, auditable financial research, while newer entrants (o11, F2) criticize both for remaining 'browser silo' tools that require moving data in and out of native Office apps rather than manipulating Excel or Word documents in place. | 低 | SR009 |
| CR028 | Rogo's $160M Series D (April 2026) was led by Kleiner Perkins with continuing participation from Sequoia, Thrive Capital, Khosla Ventures, and J.P. Morgan Growth Equity Partners, a syndicate diversified across top-tier late-stage investors but concentrated among a small number of repeat backers. | 中 | SR020 |
| CR029 | J.P. Morgan Growth Equity's participation as an investor, while J.P. Morgan is separately reported to be building in-house AI tools for its own bankers, raises a governance question about whether investor and customer incentives could diverge if Rogo's largest bank customers accelerate in-house AI builds. | 低 | SR020, SR029 |
| CR030 | Rogo's growth-through-acquisition strategy (Subset, Offset, Plux) concentrates integration risk in a small internal M&A team, with each deal adding a distinct codebase, data-access pattern, and founder-retention question. | 中 | SR033 |
| CR031 | Rogo's Trust Center names Jefferies, Truist, Rothschild & Co, Raymond James, Nomura, Tiger Global, Moelis, and Lazard as institutions that reviewed its security posture, but no source reviewed for this chapter discloses per-customer revenue concentration, net revenue retention, or contract length. | 中 | SR022 |
| CR032 | Enterprise AI procurement into investment banks and asset managers is typically a 9-24 month, multi-stakeholder process once legal, compliance, risk, and IT security sign-off are included, which slows the pace at which Rogo's pipeline can convert into realized revenue growth. | 低 | SR027 |
| CR033 | Independent, unsolicited review evidence for Rogo is thin: its PeerSpot listing reads as vendor-style descriptive copy rather than named user complaints or detailed satisfaction data, and its Trustpilot profile could not be independently verified during this run. | 中 | SR010, SR011 |
| CR034 | A Wall Street Oasis thread shows a mid-market bank product team explicitly evaluating Rogo, Hebbia, and ModelML on 'accuracy, automating workflows and the tool not being gimmicky,' language that treats AI-accuracy marketing claims as unproven by default rather than self-evidently true. | 中 | SR009 |
| CR035 | Competitor-published comparison content from Hebbia, o11, and F2 is useful for feature positioning but is not independent evidence of Rogo's reliability or customer satisfaction and should be discounted accordingly. | 低 | SR009 |
| CR036 | Any large bank customer applying the third-party AI vendor-oversight diligence FINRA and amended Regulation S-P now require of broker-dealers would need Rogo to produce audit logs, model-change documentation, and incident-response evidence on demand across its 250+ institutional relationships. | 中 | SR001, SR023 |
| CR037 | Rogo's growth narrative currently rests more on adoption breadth (35,000+ users, 250+ institutions) than on disclosed depth (retention, expansion, or per-seat economics), a gap this report's customers chapter also flags. | 中 | SR020 |
| CR038 | Rogo's valuation reportedly rose from approximately $750M at its January 2026 Series C to approximately $2B at its April 2026 Series D, a near-tripling in three months, led by Kleiner Perkins with Sequoia, Thrive Capital, Khosla Ventures, and J.P. Morgan Growth Equity participating. | 中 | SR020 |
| CR039 | No source reviewed in this or the financials chapter discloses audited revenue, gross margin, net revenue retention, or profitability for Rogo; the public narrative rests on user and institution counts rather than a disclosed ARR or growth-rate figure. | 低 | |
| CR040 | As of January 2026, tech-industry analysts and investors were actively debating whether AI-sector valuations represent a bubble, citing a disconnect between roughly $400B in annual AI investment and a much smaller measured enterprise-productivity return. | 中 | SR015 |
| CR041 | If Rogo's growth decelerates below the rate implied by its funding cadence, or a broader AI-valuation correction compresses comparable private multiples, Rogo would be exposed to a down round or a materially reduced IPO/exit valuation relative to its most recent primary price. | 中 | SR015, SR020 |
| CR042 | Plaintiffs' firms and the SEC's Cybersecurity and Emerging Technologies Unit have both grown their focus on disconnects between AI-driven growth narratives and disclosed fundamentals, a dynamic that would apply directly to Rogo if it pursues a public listing while its revenue and retention claims remain largely undisclosed. | 中 | SR018, SR019 |
| CR043 | Until Rogo publishes or is required to publish audited revenue, retention, and margin figures, any valuation stance on Rogo should be treated as provisional and heavily caveated by this disclosure gap. | 中 | SR020 |
| CR044 | Forward-deployed engineer job postings grew roughly 800% between January and September 2025 and roughly 1,165% year-over-year into early 2026, while the candidate pool grew only about 50%, with total compensation at leading AI labs exceeding $500K for the most sought-after profiles. | 中 | SR014 |
| CR045 | Rogo competes for forward-deployed talent against OpenAI, Anthropic, Google, and Palantir, all of which are hiring FDEs aggressively, making Rogo's ability to scale its embedded-banker model at the pace implied by its customer count a genuine execution constraint. | 中 | SR014 |
| CR046 | Major banks including JPMorgan Chase, Citigroup, Goldman Sachs, and Morgan Stanley are shrinking junior-analyst intake as AI automates modelling, pitchbook, and comps work, with JPMorgan's CEO signalling the bank will likely hire more AI specialists and fewer traditional bankers going forward. | 中 | SR029, SR032 |
| CR047 | The same AI adoption wave validates Rogo's market category but may also shrink Rogo's own future recruiting pool of banking-literate technologists as fewer people enter the junior-analyst programs Rogo's own founders came from. | 低 | SR017, SR032 |
| CR048 | No source reviewed for this chapter discloses a named CFO, chief compliance officer, or chief risk officer for Rogo, a notable governance gap for a company selling compliance-adjacent AI tools into the most heavily regulated segment of financial services. | 低 | |
| CR049 | Rogo's operating risk is currently concentrated in a small founding and early-employee group simultaneously scaling multiple acquisitions, new geographies, and hiring against an unusually tight forward-deployed labour market. | 中 | SR014, SR017 |
| CR050 | Across regulatory, product, partner, and people risk categories, the nearest-term, highest-likelihood exposures are FINRA/Reg S-P compliance burden and Microsoft Copilot bundling, the highest-severity-if-triggered exposures are EU AI Act enforcement and a confirmed reliability incident, and the lowest-likelihood-today exposure is direct litigation or enforcement naming Rogo. | 中 | SR001, SR016 |
| CV001 | Rogo's April 29, 2026 Series D raised $160 million and valued the company at approximately $2 billion, up from a $750 million valuation at its January 2026 Series C three months earlier. | 高 | SV015, SV022, SV031 |
| CV002 | Kleiner Perkins partner Mamoon Hamid, whose firm led the Series D, publicly framed Rogo as pursuing a 'generational' opportunity because it is becoming 'the operating system for an entire industry.' | 中 | SV031 |
| CV003 | Sequoia Capital, Thrive Capital, Khosla Ventures, and J.P. Morgan Growth Equity Partners -- all prior-round investors -- reinvested in the Series D alongside new lead Kleiner Perkins. | 中 | SV015 |
| CV004 | Hebbia raised a $130 million Series B in mid-2024 at a roughly $700 million valuation on about $13 million of profitable annual recurring revenue, implying a multiple of approximately 54x ARR. | 高 | SV023, SV008 |
| CV005 | TechCrunch reported that The Information had separately valued Hebbia's closest analogues, Glean and Harvey, at slightly over 60x ARR around the same period. | 中 | SV023 |
| CV006 | Glean crossed $300 million in annualized revenue in May 2026, tripling from $100 million ARR in roughly 15 months, while still carrying the $7.2 billion valuation set at its June 2025 Series F -- an implied multiple of about 24x current run-rate. | 高 | SV009, SV010 |
| CV007 | At Glean's June 2025 Series F, an outside analyst quoted by Reuters called the round's 72x ARR multiple (at roughly $100 million ARR) 'punchy,' noting investors were partly compensated by Glean already being cash-flow positive. | 中 | SV010 |
| CV008 | AlphaSense raised $350 million in June 2026 at a $7.5 billion valuation, roughly 12.5x its approximately $600 million ARR reported for Q1 2026, up from $500 million ARR in October 2025. | 中 | SV011, SV012 |
| CV009 | An investor-facing analysis of the AlphaSense round argued its 12.5x ARR multiple 'matches public AI software companies trading at 8-15x forward ARR' even though AlphaSense carries no public liquidity, no disclosed path to profitability, and concentrated financial-services customer exposure. | 中 | SV018 |
| CV010 | FactSet Research Systems' SEC Form 10-K for fiscal 2025 (ended August 31, 2025) reported organic Annual Subscription Value of $2,370.9 million (+5.7% year over year) and net income of $597.0 million (+11.2% year over year), and an independent market-data tracker separately put FactSet's trailing-twelve-month revenue at $2.40 billion as of mid-2026. | 高 | SV001, SV005 |
| CV011 | As of June 30, 2026, FactSet traded at a market capitalization of about $8.38 billion and an EV/Sales multiple of roughly 4.0x on trailing revenue of $2.40 billion, down sharply from a market cap of about $10.67 billion as of October 2025. | 中 | SV005, SV006 |
| CV012 | Intapp reported cloud ARR of $459.3 million (+31% year over year) and total ARR of $560 million (+23%) for the quarter ended March 31, 2026, with full fiscal-year 2026 total revenue guidance of $574.3-575.3 million and 123% cloud net revenue retention. | 中 | SV003 |
| CV013 | An independent public-comps tracker put Intapp's EV/Revenue multiple at approximately 3.1x and EV/EBITDA at approximately 15.1x on a public market capitalization of about $2 billion as of mid-2026. | 中 | SV004 |
| CV014 | S&P Global's Market Intelligence segment, which houses the Capital IQ platform that Rogo's own agents pull comparables from, contributed to consolidated fiscal 2025 revenue of $15.336 billion across all segments (+8% year over year), per its SEC Form 10-K. | 中 | SV002 |
| CV015 | A June 2026 public-market comp database placed pure-play 'Artificial Intelligence' software companies at a median forward EV/Revenue multiple of about 3.7x, versus a 2.1x EV/Revenue and 9.2x EV/EBITDA median across all tracked horizontal SaaS categories. | 中 | SV007 |
| CV016 | The same June 2026 dataset priced the 'Financial Services Software' vertical at roughly 2.9x EV/Revenue, materially below Design & Engineering Software (4.5x) and Data Infrastructure (5.4x) categories. | 中 | SV007 |
| CV017 | Silicon Valley Bank's 2026 enterprise software report found that 65% of US enterprise software venture capital went to AI startups in 2025 and that more than 75 new AI-related unicorns have been created since 2025, bringing the total to 356, while a separate funding tracker put total 2025 AI venture funding at $212 billion and Q1 2026 global venture funding at $300 billion with AI capturing roughly 80% of it. | 高 | SV032, SV016 |
| CV018 | A research analysis citing PitchBook and IDC data estimated global AI investment reached roughly $400 billion annually against only about $100 billion of enterprise AI revenue -- a 4:1 ratio the analysis compared to prior technology cycles that preceded corrections. | 中 | SV013 |
| CV019 | The same analysis reported that private AI startup valuations had already declined by approximately 23% since late 2025 as investor sentiment shifted from 'AI evangelism' to 'AI evaluation.' | 中 | SV013 |
| CV020 | A June 2026 market commentary noted the Nasdaq fell 4.7% in the week of June 5, 2026 -- its worst week in more than a year -- after a stronger-than-expected May jobs report reignited fears that higher rates could pressure AI infrastructure spending and valuations. | 中 | SV014 |
| CV021 | An AI-startup funding tracker found 2026 deal count fell roughly 14% year over year even as total dollars invested rose, describing a 'barbell effect' in which mega-rounds above $500 million and micro-rounds below $3 million grew while Series A/B startups without strong revenue metrics struggled to raise. | 中 | SV017 |
| CV022 | The same tracker stated that more AI companies 'crashed and burned' in 2026 than in the prior three years combined, even as the largest AI companies captured record funding. | 中 | SV017 |
| CV023 | An analysis of late-stage AI financing concluded that nearly half of 2026 IPO-track startups were caught between an expired disclosure window and public-market valuations investors 'simply won't accept,' with only about one in four judged genuinely IPO-ready. | 中 | SV025 |
| CV024 | S&P Global's 2026 labor-market survey found large enterprises -- the buyer segment Rogo depends on -- forecast a net negative employment impact from AI investment of -13 percentage points for the year ahead, versus a slightly positive net hiring balance at small and medium-size firms. | 中 | SV029 |
| CV025 | Microsoft's Copilot for Finance add-on lists at $15 per agent per month (capped) on top of the $30-per-user Microsoft 365 Copilot license, with realized blended seat costs of $66-87 per user per month once the mandatory E3/E5 prerequisite is included. | 中 | SV020 |
| CV026 | Despite Microsoft's roughly $150 billion annual AI capital-expenditure run rate in early 2026, only about 3.3% of the Microsoft 365 commercial installed base had converted to paid Copilot seats, and Copilot's share of the US paid AI subscriber market fell 39% in six months. | 中 | SV021 |
| CV027 | Bloomberg introduced an agentic AI interface inside its roughly $12.6 billion-revenue, approximately 375,000-seat Terminal franchise in 2026, the same year Perplexity launched a rival 'Computer' product whose viral demo showed it replicating core terminal-style research and modeling functions at a fraction of the cost. | 中 | SV028 |
| CV028 | S&P Global's Kensho unit built a proprietary multi-agent 'Grounding' framework on LangGraph to ground AI outputs directly in S&P Global's own financial datasets, positioning incumbent data owners like S&P Global and FactSet to ship agentic research features natively rather than cede that layer to standalone vendors. | 中 | SV030 |
| CV029 | A private-market valuation tracker independently recorded Rogo's January 2026 Series C mark of $750 million on $153.6 million raised across seven rounds (a 4.88x capital-efficiency ratio), a figure comparable in structure to the 4.40x ratio the same tracker calculated for Hebbia's $700 million Series B mark. | 中 | SV027, SV008 |
| CV030 | Built In's 2026 IPO watchlist named Databricks as the only large AI-native company nearing public markets with disclosed profitability (about $5.4 billion annualized revenue, positive free cash flow), while OpenAI and Anthropic remain unprofitable multi-billion-dollar cash burners despite valuations under discussion above $850 billion and $900 billion respectively. | 中 | SV024, SV026 |
| CV031 | An AI IPO tracker described 2026 as 'the most AI-concentrated IPO year on record,' with roughly 92% of an estimated $3 trillion pipeline of imminent listings tied to AI or AI-adjacent companies, and noted that late-stage private multiples remain well below 2021-2022 peaks. | 中 | SV026, SV025 |
| CV032 | Because none of the sources reviewed in this chapter disclose Rogo's own current-year revenue run-rate, gross margin, net revenue retention, or cash runway, no ARR-multiple or DCF-style valuation of Rogo can be underwritten directly from primary company disclosure; the $2 billion Series D mark is best read as a syndicate-driven price set by informed insider investors rather than an externally verifiable intrinsic value. | 中 | SV027, SV022, SV031 |
| CV033 | If Rogo's actual current annualized revenue sits meaningfully below the multiples implied by AI-native peers with disclosed ARR (Hebbia at roughly 54x, Glean at roughly 24x current run-rate, AlphaSense at roughly 12.5x), the $2 billion mark would represent a materially higher multiple than any of its closest comparables -- comparables that themselves already face public skepticism about sustainability. | 中 | SV023, SV009, SV011, SV018 |
| CV034 | A confirmed down round at Rogo's next financing, a public disclosure that current annualized revenue is far below the pace implied by recent market chatter, or a material slowdown in net-new bulge-bracket and elite-boutique logo growth would each independently undermine the case for sustaining the $2 billion mark. | 中 | SV017, SV025 |
| CV035 | Rogo's own product content markets Felix and its comps-refresh workflow as pulling directly from Capital IQ and other licensed data feeds, meaning Rogo's valuation thesis partly depends on continued licensing access to the same incumbent data platforms (FactSet, S&P Global Capital IQ) whose own agentic roadmaps could someday substitute for Rogo's workflow layer. | 中 | SV030, SV002 |
| CV036 | No source reviewed in this chapter discloses a public secondary-market transaction, tender offer, or independent third-party fairness opinion for Rogo's Series D price, so the $2 billion mark reflects a single primary-round clearing price rather than a market-tested valuation. | 中 | SV015, SV031 |
| CV037 | Hebbia, Glean, and AlphaSense -- Rogo's closest AI-native comparables by buyer overlap or workflow category -- were all still private and still growing revenue at or above 65% year over year as of their most recent disclosed rounds, indicating the broader AI-native enterprise-research category has not yet produced a public listing that would anchor Rogo's eventual exit multiple. | 中 | SV023, SV009, SV011 |
| CV038 | The public software comp set shows a wide dispersion by AI posture: 'AI-native' vertical and horizontal categories trade at meaningfully higher EV/Revenue multiples than legacy or 'AI-enabled' incumbents in the same June 2026 dataset, which is the structural reason Rogo's private multiple (implied, not disclosed) cannot be benchmarked against FactSet or Intapp's public multiples without a large category adjustment. | 中 | SV007 |
| CV039 | Multiple 2026 sources tracking AI-sector financing describe capital as increasingly concentrated in a small number of category leaders per vertical, which supports a bull-case reading that Rogo's fast re-rating from $750 million to $2 billion reflects investors picking a perceived category winner in finance-vertical AI rather than indiscriminate sector froth. | 中 | SV016, SV032, SV019 |
| CV040 | Taken together, the adverse evidence in this chapter -- a 4:1 AI investment-to-revenue ratio, a reported 23% private AI valuation decline since late 2025, a 14% drop in AI deal count, incumbent bundling pressure from Microsoft and Bloomberg, and a negative large-firm employment outlook in Rogo's own customer base -- forms a coherent bear case independent of any single data point. | 中 | SV013, SV017, SV021, SV028, SV029 |
| CV041 | This chapter's final diligence list treats Rogo's undisclosed current-year ARR, gross margin, net revenue retention, and cash runway as the single largest blockers to converting the $2 billion price into an independently underwritten valuation stance. | 低 | |
| CV042 | No source reviewed in this chapter names a specific strategic acquirer actively pursuing Rogo, but FactSet, S&P Global, and Bloomberg -- all of which are independently building or acquiring agentic AI research capability -- are the incumbents most structurally positioned to consider Rogo as an acquisition rather than compete feature-for-feature. | 中 | SV001, SV002, SV028 |
| CV043 | Because Rogo itself has been an acquirer (of Offset and other targets referenced elsewhere in this diligence) rather than an acquisition target to date, its M&A posture as of mid-2026 is consolidative, which somewhat weakens the near-term strategic-sale exit scenario relative to an IPO or a later, larger strategic transaction. | 中 | SV022, SV015 |
| CV044 | Given the combination of undisclosed unit economics, a syndicate-driven (not market-tested) price, and sector-wide multiple compression risk, this chapter's valuation stance is that the $2 billion mark is stretched relative to underwritable evidence today, though not indefensible given Rogo's growth-round comparables, and the appropriate diligence posture is track/research-more rather than an outright buy or avoid call. | 中 | SV015, SV031, SV013, SV017 |
| 编号 | 出版方 | 标题 | 引文 |
|---|---|---|---|
| SO001 | Rogo | Rogo | AI for the most ambitious firms in finance (homepage) | |
| SO002 | Rogo | Rogo company page: founders and headquarters | |
| SO003 | Rogo | Rogo customers page | |
| SO004 | Rogo | Rogo product page | |
| SO005 | Rogo | Rogo security page | |
| SO006 | Rogo | Rogo news/updates index | |
| SO007 | Rogo | Rogo careers page | |
| SO008 | Rogo | Announcing our Partnership with Microsoft | |
| SO009 | Rogo | Rogo | Meet Felix (product page with customer quotes) | |
| SO010 | Rogo | Scaling Rogo to Build the Future of Finance: Our $75M Series C and European Expansion | With more than 25,000 financial professionals using Rogo daily, our platform supports dealmakers at firms, including Rothschild & Co, Jefferies, Lazard, and more. |
| SO011 | Rogo | Rogo Raises $50M Series B from Thrive Capital, J.P. Morgan, and Tiger Global to Build Financial AI | |
| SO012 | Rogo | Baird Equity Research | Rogo customer case study | |
| SO013 | Rogo | Rogo Achieves EU AI Act Compliance | |
| SO014 | Rogo | Rogo Acquires Offset | |
| SO015 | PR Newswire (Rogo) | Rogo Raises $160M Series D to Scale the Agentic Platform for Finance | Rogo, the AI platform purpose-built for finance, today announced it has raised $160 million in Series D funding led by Kleiner Perkins... The Series D brings Rogo's total funding to more than $300 million. |
| SO016 | PR Newswire (Rogo) | Rogo Announces $18M Series A Funding Round led by Khosla Ventures to Build Wall Street's First AI Analyst | Founded in 2021, Rogo has rapidly established itself as the leading vertical provider of Generative AI solutions for financial firms. |
| SO017 | FinTech Futures | AI start-up Rogo raises $160m Series D | The start-up claims to have attracted more than 250 institutional clients since launching in 2021. |
| SO018 | FinTech Global | Rogo raises $160m Series D to scale finance AI platform | |
| SO019 | TBPN Digest | Rogo raises $160M Series D at $2B valuation as AI reshapes Wall Street knowledge work | Rogo, an AI platform built for investment banks, private equity firms, and hedge funds, has closed a $160 million Series D at a $2 billion valuation. |
| SO020 | New York Weekly | NYC Fintech Startup Rogo Closes $160M Series D, Bringing Total Funding to $300M | J.P. Morgan Growth Equity Partners is not a passive financial backer — it is one of Rogo's key institutional clients. |
| SO021 | citybiz | Rogo Raises $7 Million to Advance Specialized GenAI for Wall Street | |
| SO022 | Forbes | Rogo | Company Overview & News | Founded 2022. Headquarters New York, New York. Employees 100. Revenue grew from $2 million in 2024 to more than $15 million in 2025. |
| SO023 | Hebbia | Top 10 Rogo Competitors for Finance Teams [2026] | Document scale: Rogo operates as a user interface application built on top of existing LLMs. Yet it can't reliably scale analysis over thousands of documents...Citation granularity: Rogo provides auditable, response-level citations, but lacks true sentence-level sourcing. |
| SO024 | PeerSpot | Compare FactSet vs Rogo | |
| SO025 | Bloomberg (via BusinessMirror) | Junior bankers sick of grunt work build $2B AI tool | Among AI enthusiasts, there are skeptics who paint Rogo as an unnecessary layer, because finance professionals can do much of the work directly with large AI models. |
| SO026 | The Outpost | Rogo Technologies Hits $2B Valuation With AI Tool | |
| SO027 | OpenAI | Rogo scales AI-driven financial research with OpenAI o1 | Since emerging from stealth in 2024, Rogo has: Served over 5,000 bankers... Grown their Annual Recurring Revenue (ARR) 27x. |
| SO028 | Sacra | Rogo valuation, funding & news | |
| SO029 | SixThirty Ventures | SixThirty First Take: Rogo's Recent Fundraise in the Context of AI for FIs | Rounds are still pricing today at outlier-level revenue multiple premia. The prices suggest outliers but there are many players, so what constitutes a competitive moat... |
| SO030 | FinTech Global | Rogo raises $75m Series C to scale AI finance platform | |
| SO031 | FinanceWire | Rogo Acquires Offset to Bring AI Agents into Financial Workflows | The platform integrates with technology partners and industry data providers including OpenAI, Google Gemini, Anthropic, LSEG, S&P Global, FactSet, and PitchBook. |
| SO032 | Rogo | Deepening Our Partnership with PitchBook | |
| SM001 | Precedence Research | Generative AI in Financial Services Market Size to Hit USD 17.88 Bn by 2035 | the global generative AI in financial services market size is calculated at USD 1.95 billion in 2025 and is predicted to increase from USD 2.51 billion in 2026 to approximately USD 17.88 billion by 2035 |
| SM002 | The Business Research Company | Generative Artificial Intelligence (AI) in Banking and Finance Market 2026, Insights & Trends | Generative Artificial Intelligence (AI) In Banking And Finance market size has reached to $1.75 billion in 2025, Expected to grow to $7.71 billion in 2030 at a compound annual growth rate (CAGR) of 34.5% |
| SM003 | Deloitte | Unleashing a new era of productivity in investment banking through the power of generative AI | One study by Stanford researchers found that generative AI boosted a call center's productivity by 14%. Another study by Massachusetts Institute of Technology concluded that generative AI helped reduce time and improve the quality of work for marketers, consultants, and data analysts. |
| SM004 | McKinsey & Company | CIB in an era of volatility, AI, and nonbank challengers | CIB revenues reached $3.0 trillion in 2024, with year-over-year revenue growth of 4.4 percent... employing the full set of levers could improve profitability versus the current baseline by 20 to 30 percent |
| SM005 | EY | Beyond implementation: PE's AI evolution into differentiated growth | the market has caught up and investment levels across PE are matching those in other sectors, the essence of private equity is about pushing boundaries, unlocking growth and carving out differentiation |
| SM006 | Boston Consulting Group | Private Equity's Future Is Digital First and AI Powered | Digital initiatives alone deliver a 15% to 20% return on investment (ROI)... but when AI is built on these foundations, total returns can reach 30% to 35%... more than 90% of investment... |
| SM007 | KPMG | KPMG Quarterly AI Pulse Survey: Asset Management and Private Equity | some asset managers and private equity firms are beginning to deploy AI agents in their organizations and working toward clearer paths to measurable ROI... offering premiums for candidates with strong AI skills |
| SM008 | Wall Street Prep | Bloomberg vs. Capital IQ (CapIQ) vs. Factset vs. Refinitiv | The cost of a Bloomberg Terminal is $27,660/year for one license... FactSet subscription is $12,000 per year for the full product... Eikon is $22,000 per year... Bloomberg... controls more than ~33% of the financial data market |
| SM009 | Gitnux | 140+ Equity Research Industry Statistics (2026, Verified) | the equity research industry proclaims its global health with $8.7 billion in revenue... roles increasingly demand tech skills over tenure... MiFID II that decimated sell-side budgets and coverage |
| SM010 | Tracxn | Rogo - Company Profile, Team, Funding & Competitors | AI for investment banks, private equity firms, and hedge funds. It delegate research tasks to a domain-specific personal analyst that understands finance. |
| SM011 | TechBloat | AI startup Hebbia raised $130M at a $700M valuation | Hebbia has raised $130 million in new funding at a reported $700 million valuation, underscoring how quickly investor attention has shifted toward AI tools that can handle complex knowledge work |
| SM012 | askRIA | askRIA vs Hebbia vs Rogo: Best AI for Lean Investment Funds | Hebbia and Rogo are built for large financial institutions, investment banks, and enterprise teams that need AI-powered document search, investment research, and analysis at scale. |
| SM013 | eFinancialCareers | An ex-Morgan Stanley analyst is overseeing the junior banking AI jobs apocalypse | Anthropic announced the release of "ten ready-to-run agent templates" for financial services. The templates cover... pitchbooks, monitoring earnings reviews, building financial models and checking valuations against comparables. |
| SM014 | The AI Chronicle | AI Wall Street: Banks Cut Junior Analysts for Automation | banks are slashing junior analyst classes by as much as two-thirds (roughly 66%)... banks are sourcing approximately 62% of their new "AI talent" from these very same candidate pools |
| SM015 | Hedge Fund Alpha | 94% Of Fund Managers And Investors Will Spend More On AI In 2026 | Almost all (94%) of the fund managers and investment analysts questioned say spending this year will increase on last year with 18% predicting a substantial increase. |
| SM016 | Alternative Investment Management Association (AIMA) | Front-office Gen AI adoption shifts from 'if' to 'when' for leading fund managers | 58% of fund managers surveyed expect increased Gen AI use in investment processes over the next year, up from 20% in 2023... 60% of institutional investors would be more likely to invest in a hedge fund that allocates a meaningful portion of its budget to Gen AI |
| SM017 | Praxis Rock | Top 100 Private Equity Firms: 2026 Rankings by AUM | Private equity manages roughly $8 trillion globally. Five years ago, that number was $4 trillion... Blackstone manages $1.3 trillion... Apollo is approaching $1 trillion. KKR crossed $744 billion. |
| SM018 | Microsoft | Copilot in Excel: Built for the era of Frontier Finance | Across Financial Planning and Analysis (FP&A), Accounting, Tax, Compliance, and Treasury, Microsoft Finance runs Copilot in Excel in real workflows... introducing new features built for financial professionals |
| SM019 | Microsoft Learn | Overview of Finance agents in Microsoft 365 2026 release wave 1 | Finance Agent is a role-based Copilot experience that brings together AI-powered financial intelligence, conversational ERP access, and finance-centric workflows across Microsoft 365. |
| SM020 | VaasBlock | Microsoft Copilot 3.3% Penetration vs $190B AI Capex: The Monetization Gap | Microsoft's 2026 capital expenditure guidance is $190 billion — a 61 percent increase from 2025... The stock fell 5 percent on the day... the gap between what Microsoft is spending on AI infrastructure and what the product... is currently producing |
| SM021 | ACA Group | FINRA Releases 2026 Oversight Report Highlighting AI, Cybersecurity and Compliance Risks | Supervisory evidence should be retained, and chatbot interactions must be supervised and archived just like other communications... Outsourcing does not outsource responsibility |
| SM022 | Goodwin Procter | 2026 SEC Exam Priorities for Registered Investment Advisers and Registered Investment Companies | the 2026 Priorities also highlight the SEC's increasing attention to the use of emerging artificial intelligence (AI) technologies... training and security controls... AI and polymorphic malware attacks |
| SM023 | CIO.com | 2026: the year AI ROI gets real | a staggering 95% failure rate for enterprise generative AI projects, defined as not having shown measurable financial returns within six months... 61% of the 3,700 senior business leaders... feel more pressure to prove ROI |
| SM024 | The Economic Times | MIT study shatters AI hype: 95% of generative AI projects are failing, sparking tech bubble jitters | A new report from MIT, The GenAI Divide: State of AI in Business 2025... reveals that 95 percent of business attempts to integrate generative AI are failing. Only 5 percent of companies have managed to achieve meaningful revenue acceleration. |
| SM025 | Blott | AI in Private Equity 2026: Use Cases and Data | Enterprise AI adoption reached 88% globally in 2025, yet only one-third of organisations have scaled beyond pilot projects, and just 6% qualify as AI high performers... 86% of organisations have now integrated generative AI into their M&A workflows |
| SM026 | Boston Consulting Group | Global Asset Management Report 2026: An Imperative for Growth | Global assets under management (AuM) reached $147 trillion in 2025, up 11% year over year, while aggregate profit margins held above 30%. ... more than 80% of gross revenue growth in 2025 was driven by market appreciation |
| SM027 | VendorBenchmark | Financial Services IT Stack Cost Benchmark | Financial services firms spend more on IT than almost any other industry, an average of 9.2% of revenue annually... Unlike manufacturing companies that spend 2.8% on IT or retail companies at 3.2% |
| SP001 | o11 | Rogo vs Hebbia vs o11: Search vs. Creation | Rogo and Hebbia are world-class Search and Synthesis Engines... their architecture is primarily browser-based... This 'Last Mile' problem -- the gap between finding the answer and incorporating it into the work product -- is where significant time and context are lost. |
| SP002 | Marvin Labs | AI Tools for Equity Research: 2026 Comparison | Best for deal teams in banking and PE: Rogo or Hebbia (enterprise). Best for financial model automation: Daloopa (free tier available). |
| SP003 | AlphaSense | Pricing | AlphaSense | Market Intelligence / Enterprise Intelligence content tiers span Broker & Independent Research, Expert Transcript Library, News, Regulatory, and Internal Content, with 24/7 Live Help and a Dedicated Account Manager. |
| SP004 | Costbench | AlphaSense Pricing 2026: 5 Custom-Quoted Tiers Compared | AI Search & Core Platform: $10,000-$15,000/user/year ... Wall Street Insights: $20,000-$40,000/user/year ... Enterprise Intelligence: $50,000-$100,000+/year (team license). |
| SP005 | Bloomberg Professional Services | AI on Bloomberg | Bloomberg Professional Services | ASKB coordinates a network of AI agents that work in parallel to dynamically access Bloomberg's data, news, research and analytics... in addition to sell-side and independent research from over 800 providers, ASKB draws on proprietary research from Bloomberg Intelligence, BloombergNEF and Bloomberg Economics. |
| SP006 | arXiv | BloombergGPT: A Large Language Model for Finance | We present BloombergGPT, a 50 billion parameter language model... We construct a 363 billion token dataset based on Bloomberg's extensive data sources, perhaps the largest domain-specific dataset yet. |
| SP007 | FactSet | FactSet Recognized for Pioneering AI Advancements in Financial Technology | Industry-first Model Context Protocol (MCP) server to enable direct, secure, AI-ready access to trusted FactSet market data... Introduction of FactSet AI for Banking, developed with Finster AI, a unified, secure workflow automation ecosystem for investment banking teams. |
| SP008 | Kensho (S&P Global) | Kensho LLM-Ready API Adds Private Company Financials, S&P CapIQ Estimates, and Enhanced Auditability | S&P Private Company Financials includes data for over 12 million active and inactive private companies globally... added source document links in API responses to enable deeper auditability and data verifiability. |
| SP009 | S&P Global (PR Newswire) | S&P Global Transforms S&P Capital IQ Pro Experience with the Launch of New Generative AI-Powered Capabilities | ChatIQ leverages Large Language Models (LLMs) and is trained on the vast corpus of S&P Capital IQ Pro tabular and textual data. The solution is specifically tailored to support the needs of banking and buyside analysts. |
| SP010 | myABT | Microsoft Copilot for Financial Institutions: The 2026 Deployment Guide | A DLP bypass bug from January 21 to February 3, 2026 allowed Copilot to process and summarize confidential emails in Sent Items and Drafts while ignoring sensitivity labels and DLP policies... 40 percent of organizations delayed their Copilot rollout by three or more months over data exposure concerns (Gartner, 2025). |
| SP011 | Futurum Group | Can Glean's Financial Services Push Make AI Assistants a Compliance Asset, Not a Risk? | According to Futurum Group's 1H 2026 AI Platforms Decision Maker Survey (n=820), reliability and hallucination management (55%) and data privacy (53%) are the top two AI adoption challenges, especially acute in sectors like financial services. |
| SP012 | TMCnet (Glean) | Glean Expands Financial Services MCP Ecosystem to Bring Trusted Market Intelligence Into Enterprise Context | Enterprise AI leader Glean today announced an expanded financial services MCP ecosystem with leading market and financial intelligence providers including CB Insights, Crunchbase, Daloopa, FactSet, and S&P Global. |
| SP013 | Harvey | Harvey Raises at $11 Billion Valuation to Scale Agents Across Law Firms and Enterprises | More than 25,000 custom agents operate on Harvey... the company has raised more than $1 billion in total funding... partnering with the majority of the AmLaw 100, over 500 in-house legal teams, and 50 asset management firms across 60 countries. |
| SP014 | TechStartups | Legal AI startup Harvey hits $11B valuation with $200M funding as investors look beyond OpenAI and Anthropic | The company reached $190 million in annual recurring revenue in January, up from $100 million just months earlier, according to a CNBC report... a sharp jump from the $8 billion mark it reached just a few months ago. |
| SP015 | Intapp | The AI-powered deal and relationship intelligence platform (DealCloud) | Celeste, Intapp's agentic AI platform, continuously analyzes relationships, communications, deal flow, matters, and engagements. It surfaces opportunity, flags risk, and recommends next steps -- securely and in context. |
| SP016 | Business Wire | Ensis Partners selects Intapp DealCloud with Celeste to build a best-in-class deal and relationship management infrastructure | As both founders had direct prior experience with DealCloud, including evaluating and deploying it at previous firms, Shinder and Buschmann didn't need to evaluate competing CRM platforms. |
| SP017 | LSEG | LSEG introduces Deep Research agent in Workspace | With Deep Research in Workspace, our clients can ask the most complex questions and trust the answer - grounded in LSEG data, with reasoning they can trace back to the source. |
| SP018 | Rogo | Announcing Our Strategic Partnership with LSEG | Global financial services organizations and leading Wall Street institutions including Moelis, Nomura, and Tiger Global trust Rogo to work smarter, move faster, and outpace competitors. |
| SP019 | CB Insights | Top Daloopa Alternatives, Competitors | Daloopa's top competitors include Rogo, Fintool, and Metal. |
| SP020 | innobu | BankerToolBench 2026: AI Agents Fail the Banking Test | Not a single model passed unscathed... 0% client-ready outputs across all 9 models. 16% acceptable outputs from the best model (GPT-5.4). 27% completely unusable outputs. |
| SP021 | JurisTech | JurisTech's 2026 LLM Benchmark For AI Hallucination in Finance | A reliable model needs to know how to answer, but it also needs to know when to stop... the weaker models filled in the gaps, made assumptions, and produced answers that looked complete while resting on unsupported inputs. |
| SP022 | SP2 Analytics | Analysts Vs. AI (Vol. 2): The Hallucination Problem and Implications in Investment Research | In July 2025, Deloitte Australia delivered a 200+ page report to the Australian government, priced at roughly A$440k... several references and footnotes were fabricated, and agreed to refund the final installment of the contract. |
| SP023 | F2 | F2 vs. Hebbia: AI Underwriting Platform Comparison (2026) | F2's LLMExcel engine evaluates Excel formulas natively and deterministically, scoring 95.25% on SpreadsheetBench Verified. Hebbia... does not offer in-platform formula evaluation at all. |
| SP024 | SelectHub | AlphaSense Reviews 2026: Pricing, Features & More | Users often highlight AlphaSense's accuracy and efficiency, appreciating how it outperforms similar solutions in delivering critical information swiftly. |
| SP025 | CNBC | Here's JPMorgan Chase's blueprint to become the world's first fully AI-powered megabank | Waldron showed the program creating an investment banking deck in about 30 seconds, work that would've previously taken a team of junior bankers hours to complete. |
| SP026 | CNBC | Goldman Sachs rolls out an AI assistant for its employees as artificial intelligence sweeps Wall Street | Goldman's move means that, along with JPMorgan Chase and Morgan Stanley, the world's top three investment banks have aggressively released generative AI tools to their workforce. |
| SP027 | AI Funding Me | AI Funding Rounds June 2026: Latest Deals & Tracker (Hebbia deep dive) | Hebbia... raised $160M in funding reaching a $700M valuation as of May 2026. Founded in 2020 by CEO George Sivulka and headquartered in New York. |
| SP028 | Agent Nexus | Top 10 Hebbia Alternatives for Enterprise AI Research in 2026 | Hebbia... raised $130M in Series B funding led by a16z in 2024... 'That's not a Hebbia problem. It's a category problem. Analytical AI reads documents. It doesn't complete the work those documents point to.' |
| SI001 | Forbes | Rogo | Company Overview & News | |
| SI002 | OpenAI | Rogo scales AI-driven financial research with OpenAI o1 | |
| SI003 | Sacra | Rogo valuation, funding & news | |
| SI004 | PR Newswire (Rogo) | Rogo Raises $160M Series D to Scale the Agentic Platform for Finance | |
| SI005 | SixThirty Ventures | SixThirty First Take: Rogo's Recent Fundraise in the Context of AI for FIs | |
| SI006 | FinTech Global | Rogo raises $75m Series C to scale AI finance platform | |
| SI007 | Growjo | Rogo: Revenue, Competitors, Alternatives | Rogo's estimated annual revenue is currently $43.2M per year. |
| SI008 | CB Insights | Rogo Stock Price, Funding, Valuation, Revenue & Financial Statements | Rogo has raised $310.5M over 6 rounds. Rogo's 2024 revenue was $1M. |
| SI009 | U.S. Securities and Exchange Commission (EDGAR) | Form D: Rogo Series A ScOp SPV Jul 2024, a Series of CGF2021 LLC | |
| SI010 | Hoodline | AI Upstart Rogo Bulks Up Midtown HQ, Promises 400 New Jobs | Rogo plans to create 422 new full-time positions and invest nearly $14 million into its New York headquarters. |
| SI011 | Rogo | Rogo Jobs (careers board) | |
| SI012 | Novis | Novis vs Rogo: Finance-grade AI without the enterprise price tag | |
| SI013 | AI Native Foundation | AI Native Case Study #53: Rogo | |
| SI014 | ZenML | Rogo: Scaling Financial Research and Analysis with Multi-Model LLM Architecture | These figures come directly from the company and should be viewed with appropriate skepticism as promotional claims. |
| SI015 | AlphaSense | AlphaSense Raises $350M at $7.5B Valuation, and Surpasses $600M in Annual Recurring Revenue | |
| SI016 | Sacra | AlphaSense revenue, valuation & funding | |
| SI017 | U.S. Securities and Exchange Commission (EDGAR) | FactSet Research Systems Inc. Form 10-K (FY2025) | |
| SI018 | Macrotrends | FactSet Research Systems Gross Margin 2011-2025 | |
| SI019 | Finro | AI Valuation Multiples Q1 2026: Investors Reprice Quality | |
| SI020 | Spheron | AI Inference Cost Economics in 2026: GPU FinOps Playbook | |
| SI021 | Qubit Capital | AI Startup Valuation Multiples: 10x-50x Range (2026) | |
| SI022 | Technotrenz | RogoAI Raises $75M Series C Led by Sequoia to Scale Agentic AI for Finance | |
| SI023 | AgentMarketCap | Glean's $7.2B Valuation: Why Enterprise Knowledge Agents Beat Coding Agents on Multiples | |
| SI024 | Digital Applied | AI Agent Productivity Statistics 2026: 100+ ROI Data | |
| SI025 | GetMonetizely | The Economics of AI-First B2B SaaS in 2026: Margins, Pricing Models, and Profitability | |
| SI026 | Axios | Exclusive: Sequoia leads Rogo raise at $750M valuation | |
| SI027 | The SaaS News | Rogo Raises $75 Million in Series C | |
| SE001 | Rogo | What's New: May 2026 | Felix is Rogo's AI agent for finance, built to generate full deliverables across decks, models, memos, and dashboards from a single prompt. |
| SE002 | Rogo | Announcing Our Strategic Partnership with LSEG | |
| SE003 | Rogo | Deepening Our Partnership with PitchBook | |
| SE004 | Rogo | Rogo Achieves ISO/IEC 42001 Certification | Rogo has achieved ISO/IEC 42001:2023 certification, the first international standard for AI management systems and responsible AI governance. |
| SE005 | Rogo | Rogo acquires Subset: building spreadsheet agents for bankers | |
| SE006 | PR Newswire | Rogo Acquires Offset to Bring AI Agents into Financial Workflows | |
| SE007 | ZenML | Rogo: Scaling Financial Research and Analysis with Multi-Model LLM Architecture | The metrics cited (5,000 bankers served, 10+ hours saved weekly, 27x ARR growth) are self-reported and promotional in nature... not independently verified in the source material. |
| SE008 | London Stock Exchange Group | LSEG and Rogo Announce Strategic Partnership | |
| SE009 | Digital Trends | Microsoft Copilot can now handle more of your finance work in Excel with reusable skills and data connectors | |
| SE010 | GitHub | Rogo-Technologies organization repositories | |
| SE011 | Rogo | Rogo Integrates S&P Capital IQ Data into its AI-Powered Workflows | |
| SE012 | Rogo | Rogo Trust Center | |
| SE013 | GitHub (Rogo Technologies) | big-finance-benchmark reference harness | 928 workflow-grounded financial-research questions, each paired with an expert-authored rubric and a reference answer. |
| SE014 | The AI Runtime | Felix Is a Harness, Not a Model: How Rogo Built an Agent for High Finance | Felix is not a fine-tuned model. Felix is the harness -- the orchestration scaffold, tool layer, citation system, output formatters, audit trail, and policy controls -- into which Rogo plugs whichever frontier model performs best. |
| SE015 | Rogo | The Next Phase of AI Transformation in Finance | |
| SE016 | GeniusFirms | Rogo Reviews, Pros & Cons and Alternatives | |
| SE017 | Rogo | Expanding Rogo's Coverage Across European Markets with our Acquisition of Plux | |
| SE018 | FinTech Global | Rogo and LSEG unite to transform financial data with AI | |
| SE019 | AiThority | Rogo Integrates S&P Capital IQ Data into its AI-Powered Workflows | |
| SE020 | AI Agents Directory | Rogo Felix: A Guide to AI Agents in Financial Research | In finance, the cost of hallucination or data inaccuracy is extreme. Therefore, the industry is moving toward a human-in-the-loop model. |
| SE021 | Jefferies | Gabriel Stengel on Building the AI Platform for Investment Banking | Rogo... sits at an unusual intersection: it is simultaneously a customer, a distribution partner, and a potential competitive target of OpenAI and Anthropic. |
| SE022 | Netguardia | The EU's August 2, 2026 AI Act Deadline: Practical Obligations for High-Risk AI Systems | |
| SE023 | arXiv | BigFinanceBench: A Workflow-Grounded Benchmark for Financial-Research Agents | Evaluating ten current frontier and open-weight agents, we find substantial headroom: the best system reaches only 58.8% rubric score. |
| SE024 | Tech Funding News | Kleiner Perkins leads Rogo's $160M raise to build the AI operating system for investment banking | |
| SE025 | Rogo | Rogo Status | |
| SE026 | Daloopa | Announcing Daloopa's Partnership with Rogo | |
| SU001 | Rogo | Baird Equity Research | Rogo customer case study | More than 100 professionals across Baird Equity Research are now active on the platform, with approximately 85% weekly active usage and nearly 70% daily active usage. |
| SU002 | Rogo | Rogo customer stories and logo wall | Trusted by the world's leading financial institutions. |
| SU003 | Rogo | Our $160M Series D and the Road Ahead | Rogo is now deployed across many of the world's top investment banks, asset managers, and private equity firms... more forward-deployed bankers and engineers embedded within the firms we serve. |
| SU004 | FeaturedCustomers | Rogo reviews, testimonials, and case studies | Baird executes 250K+ AI workflows with Rogo. |
| SU005 | Wall Street Oasis (anonymous forum) | Thoughts on Rogo | We use it and we churned from it after ~1 year. The platform is the most overhyped piece of shit I've ever used. |
| SU006 | AlphaSense | AlphaSense vs Rogo comparison | |
| SU007 | Dakota | Top 10 Investment Banks February 2026 League Table | |
| SU008 | ShipOrSkip | Rogo Raises $160M to Put an AI Investment Banker on Every Terminal | The platform now serves more than 35,000 financial professionals at over 250 institutions, including Rothschild & Co, Jefferies, Lazard, Moelis, and Nomura. |
| SU009 | Beri | Rogo's $160M: AI Agents Are Eating Investment Banking | Eight months from initial conversation to production deployment is industry-standard for institutional banks. |
| SU010 | citybiz | Rogo Raises $160 Million Series D to Expand Agentic AI Platform for Financial Services | Rogo has built an AI platform that the most demanding institutions in finance trust with their most critical workflows. |
| SU011 | SuperbCrew | Rogo Raises $75 Million In Series C Funding Round | Challenges for Rogo include dependency on third party models (e.g., OpenAI, Google), which could face pricing volatility or access issues. |
| SU012 | Startup Fortune | Rogo just turned junior banking grunt work into a $2 billion AI business | Unlike consumer AI, finance AI is not trying to win hearts. It is trying to cut minutes, lower error rates, and fit inside a workflow that already exists. |
| SU013 | Deloitte | Managing the new wave of risks from AI agents in banking | Agentic AI changes the risk calculus: decisions are faster, actions are autonomous, and failure modes are more complex. |
| SU014 | Anthropic | Rogo customer story (Claude) | We needed a system that could produce structured PowerPoint and Excel output at institutional quality, not a generic approximation of it. |
| SU015 | Google Cloud | Rogo case study | Many of Rogo's clients have confided that they simply trust Google more than other AI vendors, with greater respect for its security features, regulatory preparedness, and established processes. |
| SU016 | ai2.work | Rogo Raises $160M Series D to Become Wall Street's AI Operating System | Heavy reliance on a single AI vendor for critical deal workflows creates concentration risk for institutions. |
| SU017 | PR Newswire (Rogo) | Rogo Raises $160M Series D to Scale the Agentic Platform for Finance | Trusted by more than 35,000 professionals at the world's top investment banks, private equity firms, and asset managers. |
| SU018 | OpenAI | Rogo scales AI-driven financial research with OpenAI o1 | Since emerging from stealth in 2024, Rogo has: Served over 5,000 bankers... Grown their Annual Recurring Revenue (ARR) 27x. |
| SU019 | Sacra | Rogo valuation, funding & news | |
| SU020 | Forbes | Rogo | Company Overview & News | More than 25,000 users across 150 firms including Moelis, Lazard and Tiger Global use the platform daily on a per-seat subscription basis. |
| SU021 | Bloomberg (via BusinessMirror) | Junior bankers sick of grunt work build $2B AI tool | Among AI enthusiasts, there are skeptics who paint Rogo as an unnecessary layer, because finance professionals can do much of the work directly with large AI models. |
| SU022 | Jefferies | Gabriel Stengel on building the AI platform for investment banking | We've been working with institutions like Jefferies and the majority of bulge bracket banks to help them deploy AI on top of live real deal data instead of generic workflows. |
| SU023 | LSEG | LSEG and Rogo announce strategic partnership | This allows Rogo customers globally with a Workspace license to have real-time access to LSEG's flagship data and analytics. |
| SU024 | Rogo | Deepening our partnership with PitchBook | For our shared customers, PitchBook Premium Connector data includes deal, fund, company, and investor intelligence. |
| SU025 | Daloopa | Announcing Daloopa's partnership with Rogo | Bringing Daloopa's financial data into that environment gives teams access to high-quality inputs as the foundation for their AI workflows. |
| SU026 | Rogo | Trust Center policies | |
| SR001 | FINRA | 2026 FINRA Annual Regulatory Oversight Report -- Gen AI | If a firm is relying on Gen AI tools as part of its supervisory system, its policies and procedures may consider the integrity, reliability and accuracy of the AI model. |
| SR002 | WealthManagement.com | FINRA Warns Brokers of Gen AI Hallucination Risks | Brokerage regulators are urging firms to be vigilant for the risk of hallucinations when using generative artificial intelligence tools in their operations. |
| SR003 | Office of the Comptroller of the Currency | OCC Issues Updated Model Risk Management Guidance | The guidance does not set forth enforceable standards or prescriptive requirements, and non-compliance will not result in supervisory criticism. |
| SR004 | Norton Rose Fulbright | SEC heightens enforcement for AI related disclosures | United States regulatory authorities, such as the Securities and Exchange Commission (SEC), have expressed concern about the rise of ‘AI washing.’ |
| SR005 | U.S. Securities and Exchange Commission | Recommendation of the SEC Investor Advisory Committee's Disclosure Subcommittee Regarding the Disclosure of Artificial Intelligence's Impact on Operations | Issuers have struggled with providing consistent disclosure to investors in the absence of comprehensive guidance from the Commission. |
| SR006 | artificialintelligenceact.eu (EU AI Act tracker) | Overview of the Code of Practice | The Commission's enforcement actions -- such as requests for information, access to models, or model recalls -- will only begin a year later, on August 2, 2026. |
| SR007 | Latham & Watkins | EU AI Act: GPAI Model Obligations in Force and Final GPAI Code of Practice in Place | The CoP imposes extensive obligations on GPAI providers, including specific guidelines for designing compliance and audit structures. |
| SR008 | Davis Polk | Visual memo: Key changes under the federal banking agencies' revised model risk management guidance | The 2026 MRM Guidance supersedes and replaces the Agencies' prior model risk management guidance, including the OCC and Federal Reserve's Guidance on Model Risk Management issued in 2011. |
| SR009 | Wall Street Oasis | ModelML vs Rogo AI vs Hebbia AI (forum thread) | We are focussed on accuracy, automating workflows and the tool not being gimmicky. Welcome any thoughts from users please! |
| SR010 | Trustpilot | Rogo Reviews | Read Customer Service Reviews of rogo.ai | |
| SR011 | PeerSpot | Rogo Reviews, Competitors and Pricing | |
| SR012 | ComplexDiscovery | The AI Sanction Wave: $145K in Q1 Penalties Signals Courts Have Lost Patience with GenAI Filing Failures | In the first quarter of 2026, U.S. courts imposed at least $145,000 in sanctions for fabricated citations. |
| SR013 | The Register | Deloitte refunds Aussie gov after AI fabrications slip into $440K welfare report | Deloitte has agreed to refund part of an Australian government contract after admitting it used generative AI to produce a report riddled with fake citations, phantom footnotes, and even a made-up quote from a Federal Court judgment. |
| SR014 | Paraform | Forward-Deployed Engineers: How Demand Grew 10x in 18 Months (And How to Hire One) | Between January and September 2025, job postings for this FDEs grew 800%. The candidate pool? It grew about 50%. |
| SR015 | CNBC | Are we in an AI bubble? What 40 tech leaders and analysts are saying, in one chart | Record valuations and deals driven by major investments in artificial intelligence have fueled the AI boom, leaving some to brace for the potential burst. |
| SR016 | WindowsForum | Microsoft 365 Copilot Bundles Sales, Service, Finance; Launches Agent Store | Customers who previously needed both Microsoft 365 Copilot ... plus role-based add-ons ... will now find much of that role functionality included in the core Copilot entitlement. |
| SR017 | Disruption Banking | The Skills Gap AI Is Creating in Investment Banking | Early-career professionals at these firms are warning that removing too much of the foundational hands-on work too soon could create a dangerous skills gap. |
| SR018 | Techné AI | AI Securities Class Actions Tracker (2024-2026) | Industry tracking reports approximately a 100% year-over-year increase in AI-related cases from 2023 to 2024, with continued growth through 2025 and into 2026. |
| SR019 | Bloomberg Law | Event-Driven, AI Cases Dominate 2026 Securities Litigation Field | There were between six and eight AI-related securities cases each year between 2021 and 2023, followed by 15 such cases in 2024 and 12 more just in the first half of 2025. |
| SR020 | Yahoo Finance | Rogo Raises $160M Series D to Scale the Agentic Platform for Finance | |
| SR021 | Rogo | Rogo Security | We use modern cloud infrastructure, automated security tooling, and independent audits to ensure your information remains secure. |
| SR022 | Rogo | Rogo Trust Center | Powered by SafeBase | Compliance: CCPA, ISO/IEC 27001, ISO/IEC 42001:2023, SOC 2 Type 1, SOC 2 Type 2, EU AI Act. |
| SR023 | Legaye Law | Heightened Vendor Oversight: Third-Party Risks Under the New Reg S-P Rules | Broker-dealers are now expressly required to implement and document robust oversight of any third-party service provider ... The firm remains responsible for data security regardless of delegation. |
| SR024 | Legaye Law | AI Washing - Why Regulators Demand Proof, Not Promises from Advisers and Broker-Dealers | Since March 2024, the SEC has brought enforcement actions against investment advisers, public issuers, and even an individual startup founder, all tied to false or misleading AI claims. |
| SR025 | arXiv | BigFinanceBench: A Workflow-Grounded Benchmark for Financial-Research Agents | Evaluating ten current frontier and open-weight agents, we find substantial headroom: the best system reaches only 58.8% rubric score, final-answer accuracy is a useful but lossy proxy for derivation quality. |
| SR026 | Crowell & Moring | Investor Advisory Committee Recommends SEC Disclosure Guidelines for Artificial Intelligence | The IAC cited a 'lack of consistency' in contemporary AI disclosures, which 'can be problematic for investors seeking clear and comparable information.' |
| SR027 | Kitces.com | AI Compliance: Applying Existing SEC Regulatory Frameworks | The rapid increase in investment adviser use of Artificial Intelligence (AI)-powered tools has presented a challenge to regulators. |
| SR028 | TechBytes | Federal Judge Fines Lawyers $110K for AI-Fabricated Filings | A federal judge in Oregon has sanctioned a prominent law firm with a $110,000 fine after discovering that 14 case precedents cited in a motion were entirely hallucinated by an AI legal assistant. |
| SR029 | Metaintro | JPMorgan Chase Signals Hiring More AI Specialists, Fewer Traditional Bankers | JPMorgan Chase is signaling a structural shift ... will likely hire more AI specialists and fewer traditional bankers as artificial intelligence adoption accelerates. |
| SR030 | Taylor Wessing | The final GPAI Code of Practice | On 10 July 2025, the AI Office published the final version of the GPAI Code of Practice ... this is not a 'safe harbour', but requires diligent implementation of detailed transparency, copyright, and safety protocols. |
| SR031 | U.S. Securities and Exchange Commission | SEC Investor Advisory Committee to Examine the Disclosure of Artificial Intelligence's Impact on Operations | The committee will host two panels: Disclosure of Artificial Intelligence's Impact on Operations; and Retail Investor Fraud in America. |
| SR032 | Crypto Briefing | Future bankers face fewer entry-level jobs as AI spreads across Wall Street | Leaders at JPMorgan Chase, Citigroup, Goldman Sachs and Standard Chartered have all indicated that automation will replace some existing work. |
| SR033 | Rogo | Announcing Rogo's New Partnership with PitchBook | Rogo customers will first gain access to PitchBook's best-in-class private company, deal, and fund data. |
| SR034 | Fast Company | Deloitte to refund Australian government after AI hallucinations found in report | Deloitte had reviewed the 237-page report and 'confirmed some footnotes and references were incorrect.' |
| SV001 | Securities and Exchange Commission / FactSet Research Systems Inc. | FactSet Research Systems Inc. Form 10-K (fiscal year ended August 31, 2025) | As of August 31, 2025, organic annual subscription value ("Organic ASV") totaled $2,370.9 million, an increase of 5.7% over the prior year... Net income for fiscal 2025 was $597.0 million, an increase of 11.2% from the prior year. |
| SV002 | Securities and Exchange Commission / S&P Global Inc. | S&P Global Inc. Form 10-K (fiscal year ended December 31, 2025) | Revenue increased 8% driven by increases at all of our reportable segments... The increase at Market Intelligence was primarily due to subscription revenue growth in Data, Analytics & Insights. |
| SV003 | The Motley Fool | Intapp (INTA) Q3 2026 Earnings Transcript | Cloud ARR -- $459.3 million, up 31%, now representing 82% of total ARR. Total ARR -- $560 million, growing 23%. ... Cloud net revenue retention (NRR) -- 123%. |
| SV004 | Multiples.vc | Intapp - Multiples.vc Public Comps and Valuation Multiples | Intapp trades at 3.1x EV/Revenue multiple, and 15.1x EV/EBITDA... As of June 30, 2026, Intapp has market cap of $2B and EV of $2B. |
| SV005 | Stock Analysis | FactSet Research Systems (FDS) Statistics & Valuation | FDS has a market cap or net worth of $8.38 billion... EV / Sales 4.02... In the last 12 months, FDS had revenue of $2.40 billion and earned $587.79 million in profits. |
| SV006 | Macrotrends | FactSet Research Systems Market Cap 2010-2025 | FactSet Research Systems market cap as of October 03, 2025 is $10.67B. |
| SV007 | Multiples.vc | Public Software Valuation Multiples — June 2026 | Artificial Intelligence [sector, EV/Revenue NTM] 3.7x... Financial Services Software 2.9x [EV/Revenue]... Median [all horizontal SaaS] 2.1x [EV/Revenue], 9.2x [EV/EBITDA]. |
| SV008 | Premier Alternatives | Hebbia Valuation: $700.0M (2026) | Hebbia is currently valued at $700.0M as of February 24, 2026... With a capital efficiency ratio of 4.40x, Hebbia has achieved a valuation that is 4.40 times the total capital raised. |
| SV009 | MLQ.ai | Glean Crosses $300M ARR, Tripling Enterprise AI Search Revenue in 15 Months | Glean was last valued at $7.2 billion following a $150 million Series F round in June 2025... At a 24x revenue multiple on $300M annualized revenue, the valuation remains elevated but is not out of line with high-growth enterprise AI peers. |
| SV010 | The Economic Times (Reuters) | AI company Glean hits $7.2 billion in valuation in latest funding round | According to Schulman, Glean's 72x valuation multiple on revenue is "punchy", but investors are getting "early access to a franchise" since the company is cash-flow positive. |
| SV011 | Pulse2 | AlphaSense Raises $350 Million At $7.5 Billion Valuation While Surpassing $600 Million In Annual Recurring Revenue | AlphaSense... reported exceeding $600 million in annual recurring revenue (ARR) during the first quarter of 2026, up from $500 million in October 2025. |
| SV012 | ARR.club | AlphaSense ARR hit $600M with $350 raised at $7.5B valuation | CEO Jack Kokko revealed that annual recurring revenue (ARR) recently topped $600 million, up from $500 million last fall. |
| SV013 | Perspective Labs | Is the AI Bubble About to Burst? The Numbers Behind the Hype | With $400 billion in annual investment generating only $100 billion in enterprise revenue... AI startup valuations have declined 23% since late 2025, signaling investor skepticism. |
| SV014 | MarketWise | Is the AI Bubble Set to Burst? Tech Stocks Are Giving Mixed Signals | The tech-heavy Nasdaq composite tanked, closing the week ending June 5 at 4.7% lower -- its worst week in more than a year -- as AI and tech stocks experienced a significant sell-off. |
| SV015 | Financial Advisor Magazine (Bloomberg) | Junior Bankers Sick Of Grunt Work Build $2 Billion AI Tool To Do The Job | Rogo... just notched a $2 billion valuation in a fundraising round. That's up from $750 million three months ago. The new $160 million series D round was led by Kleiner Perkins. |
| SV016 | Blockchain Council | AI Funding News 2026: Records and Key Trends | AI venture funding reached $212 billion in 2025, up from $114 billion in 2024... Q1 2026 set a record for global venture investment: $300 billion invested globally... with AI taking 80%. |
| SV017 | AIMojo | AI Startup Funding Report 2026: What the Numbers Actually Say | More AI companies crashed and burned this year than the last three years combined... Deal count dropped ~14%. Total capital went up... the middle is collapsing. |
| SV018 | Angel Investors Network | AI Valuations at 12x ARR: What Investors Should Know | AlphaSense has zero public liquidity, no disclosed path to profitability, and a financial services concentration that bleeds risk... That matches public AI software companies trading at 8-15x forward ARR. |
| SV019 | Qubit Capital | AI Startup Funding Trends 2026: Data, Rounds & What's Next | Valuations in adjacent categories have compressed because the marginal dollar is going to AI, not fintech, climate, or SaaS without an AI wedge. |
| SV020 | Atonement Licensing | Microsoft 365 Copilot Pricing 2026: The Complete Cost Reference | Microsoft 365 Copilot for Finance [is] $15 per agent per month, capped... the realised seat cost after the mandatory Microsoft 365 E3 or E5 prerequisite is $66 to $87 per user per month. |
| SV021 | Tech Insider | Microsoft AI Spending 2026: $150B Capex [Analysis] | Only 3.3% of the Microsoft 365 commercial installed base has converted to paid Copilot seats, and its share of the U.S. paid AI subscriber market dropped 39% in six months. |
| SV022 | AI2.work | Rogo Raises $160M to Build Wall Street's Agentic AI Platform | Rogo... closed its Series D on April 29, 2026, led by Kleiner Perkins... The round pushes the company's total funding past $300 million and its valuation to $2 billion -- up from $750 million just three months ago. |
| SV023 | TechCrunch | AI startup Hebbia raised $130M at a $700M valuation on $13 million of profitable revenue | The $700 million valuation implies that investors valued Hebbia at about 54 times ARR... Hebbia's closest analogues, Glean and Harvey, had valuations of slightly over 60x ARR, according to the Information's reporting. |
| SV024 | Built In | 2026 IPO Watchlist: OpenAI, SpaceX and Other Tech Giants | Databricks: Profitable, $5.4B annualized revenue, positive FCF, H2 2026 possible IPO... OpenAI: $25B+ annualized revenue, eyeing a $1T+ listing, but not profitable until 2029-2030. |
| SV025 | Eqvista | Pre-IPO Startups in 2026: Down Rounds, Multiples & Exit | Nearly half the startups eyeing an IPO in 2026 are stuck between an expired disclosure window and valuations the public market simply won't accept. Only about one in four is genuinely ready to move. |
| SV026 | AI Funding Tracker | AI IPO Tracker 2026: SpaceX, OpenAI, Anthropic, Databricks | Databricks is the only profitable company in the AI IPO pipeline, with $5.4 billion in annualised revenue growing 65%, positive free cash flow, and a net retention rate above 140%... AI and AI-adjacent companies account for roughly 92% [of the pipeline]. |
| SV027 | Premier Alternatives | Rogo Technologies Valuation: $750.0M (2026) | Rogo Technologies is currently valued at $750.0M as of January 28, 2026. The company has raised a total of $153.6M in funding across 7 funding rounds... capital efficiency ratio of 4.88x. |
| SV028 | Fanatical Futurist (Bernard Marr) | Perplexity AI's Computer AI clones Bloomberg's $30,000 terminal | That dominance generated $12.6 billion in annual revenue last year -- largely from terminal subscriptions. But that reign may be starting to crack... Perplexity AI introduced a new product called "Computer". |
| SV029 | S&P Global | AI impact on employment 2026: Labor market data and outlook | Large companies forecast a net negative employment impact of -13 [percentage points]... whereas medium-sized firms also forecast a positive employment effect from AI in 2026, with a net balance of +2 percentage points. |
| SV030 | LangChain | How Kensho built a multi-agent framework with LangGraph to solve trusted financial data retrieval | Kensho's goal is to ensure that as AI transforms industries, its outputs remain grounded in trusted data... Grounding, a multi-agent framework that serves as a core access layer for S&P Global data. |
| SV031 | Kleiner Perkins | Rogo: The AI Platform for Global Finance | We're thrilled to lead Rogo's Series D and partner with Gabe, John, Tumas, and the entire team. The best analysts on the Street now have a platform that works as hard as they do. |
| SV032 | Silicon Valley Bank | Enterprise Software Report 2026: AI & VC trends | 65% of US enterprise software venture capital went to AI startups in 2025... 356 US VC-backed enterprise software unicorns now exist. |