Basis
Basis 已是可信的代理式会计自动化品类领跑者,背后有一线投资人和头部事务所验证;但经济数据仍过于不透明,不能把 2026 年 2 月的独角兽估值视为已被验证。
Basis 看起来已是智能体会计自动化的真实品类龙头,但当前估值跑在公开经济指标前面。
封面要素
公司概况
Basis 是一家位于纽约的私营软件公司,面向会计事务所打造长周期 AI 代理。公开材料显示,平台覆盖客户会计服务、税务、审计和咨询流程,重点不是只给建议的 copilot,而是端到端完成任务后交由人工复核。公司拼出了一支混合创始团队,既有商业运营者,也有深厚的概率编程和研究人才;随后从专业金融科技和前沿 AI 投资人处完成种子轮、Series A 和 Series B 融资。公开验证最强的部分来自 Wiss、Baker Tilly、UHY 等具名事务所,以及 Basis 覆盖约 30% 美国前 25 大会计事务所的更宽口径说法。主要限制是经济数据不透明:收入、ARR、留存、利润率、烧钱速度和客户集中度均未披露。
- 成立时间
- 2023-01-01
- 创始人
- Matt Harpe, Mitchell Troyanovsky, Zenna Tavares, Emily Mackevicius, Eli Bingham
- 创立地点
- New York City, New York, United States
- 总部
- New York City, New York, United States
- 产品
- AI 会计代理可跨 CAS、税务、审计和咨询执行多步骤流程,接入 QuickBooks、Xero 等系统,并把结果交回给人工验证。
- 客户
- 美国中大型会计事务所,尤其是希望在不按相同比例增加人手的情况下扩大客户会计、税务和审计吞吐量的事务所。
- 商业模式
- 面向会计事务所销售的企业软件和流程自动化模式,可能叠加实施和支持层,但实际定价和服务组合未公开披露。
- 阶段
- Series B private company
- 融资情况
- 2026 年 2 月以约 $1.15 billion 估值完成 $100 million Series B;此前完成 $3.6 million 种子轮和 $34 million Series A。
执行摘要
主要优势
- 可见采用信号很强,既有头部会计师事务所背书,也声称约打入前 25 大事务所中的 30%。
- 一线投资机构组成股东团,$100 million Series B 降低了近期融资压力。
- 产品差异化押在长周期、可复核的会计智能体,而不是只给建议的 copilots。
主要风险
- 没有公开 ARR、留存、毛利率、烧钱速度或客户集中度数据,无法校验独角兽估值。
- 除非 Basis 私下能证明更强的工作流控制和合同保护,CPA、税务、审计和隐私责任仍留在客户侧。
- Basis 仍依赖偏窄的公开客户集、关键合作伙伴,以及可能被大型既有厂商挤压的基础模型和集成生态。
未决问题
- 2026 年 2 月估值背后的当前 ARR、收入结构、毛利率和服务强度。
- 具名会计师事务所客户的续约、扩张、集中度和合同期限数据。
- 税务和审计流程中的合同风险分配、DPA 条款、赔偿安排,以及可量化的错误率或人工覆盖率。
目录
01公司概览
1.1 身份、总部、阶段与产品范围
虽然 Basis 没有发布密集的公司事实表,但其公开身份相当一致。官网、关于页面和发布材料都把 Basis 描述为面向会计师的 AI 代理平台,而不是横向生产力 copilot。隐私政策中可见的法律记录显示,运营实体是 Essex Labs, Inc.,以 Basis 名义开展业务。CPA Practice Advisor 和 Crowdfund Insider 的公开融资报道把公司放在纽约市;官方网站本身没有披露地址或办公室列表,因此总部应视为外部报道口径,而非公司认证口径。最清晰的阶段标签是私营 Series B:Basis 在 2026 年 2 月公开宣布以 $1.15 billion 估值完成 $100 million 融资,保留来源集中没有任何信息显示其已上市或完成后续融资。对一家年轻公司来说,产品叙事也异常具体:代理要执行真实会计流程,再把完成的工作交回给会计师复核、判断和做客户咨询,而不是彻底替代这个职业。[CO001, CO002, CO003, CO005, CO008, CO009]
1.2 创始人、治理与关键人物集中度
创始人记录比融资报道中可见的运营叙事更宽。种子轮和后续融资材料持续把 Matthew Harpe 和 Mitchell Troyanovsky 列为最贴近商业发布的联合创始人,其中 Harpe 是最清晰的公开 CEO 声音,Troyanovsky 则在后来的 Codex 客户聚焦中作为面向产品的联合创始人出现。另一批保留的创始人材料和 Citybiz 的 Series A 报道还把 Zenna Tavares、Emily Mackevicius 和 Eli Bingham 识别为联合创始人,因此本章可支撑 Basis 拥有五人创始团队,尽管公开材料强调的是不同子集。五人中四人的简要背景也有支撑:外部资料称 Harpe 曾任 Boston Consulting Group 和 SoftBank 高管;Basis 和 Tavares 自己的网站称 Zenna 是推理研究者、前 Columbia 创新学者和 MIT Ph.D.;Basis 称 Mackevicius 是 Columbia 训练的博士后研究员和 MIT Ph.D.,负责 Collaborative Intelligent Systems;Basis 称 Bingham 是 Broad Institute 研究员、Pyro 共同创造者,曾任 Uber AI Labs 高级研究科学家。相比之下,保留的公开材料把 Troyanovsky 识别为联合创始人和面向产品的代表,但未提供更完整的 Basis 之前履历。公开披露仍未清晰映射五位创始人的运营头衔、治理角色或股权。Annette Garcia 被明确披露为 2024 年加入的第五名全职员工,如今负责 Accounting Product Operations;这说明 Basis 正把从业者专长嵌入产品开发,而不是把平台完全交给工程师。Series B 之后,治理也更可读:Accel 的 Miles Clements 加入董事会,董事会此前已通过 Khosla Ventures 纳入 Keith Rabois。即便如此,已披露的管理层仍然集中:创始人仍主导产品叙事、战略定位和外部沟通,因此在管理层发布更完整的高管名单和董事会图谱之前,关键人物依赖应被视为真实存在。[CO006, CO007, CO016, CO017, CO030, CO031]
| 人物 | 职务 | 背景 | 创始人市场匹配或职能覆盖 | 关键人物依赖 |
|---|---|---|---|---|
| Matthew Harpe | 联合创始人,CEO | 曾任 Boston Consulting Group 和 SoftBank 高管;融资报道称其为 CEO,也是官方融资帖共同作者。 | 将研究型创始团队商业化,并负责 Basis 的外部叙事、投资人沟通和商业定位。 | 高 |
| Mitchell Troyanovsky | 联合创始人 | 在发布和融资材料中持续出现,并出现在 Codex 客户聚焦中;保留下来的公开来源未提供更完整的 Basis 之前履历。 | 面向产品的可见创始人,连接工作流设计、模型合作叙事和工作流架构。 | 高 |
| Zenna Tavares | 联合创始人,董事 | 曾任 Columbia 创新学者,MIT 博士;Basis 和其个人网站都聚焦其在推理和因果概率编程上的工作。 | 支撑公司的推理能力和科学研究基础。 | 中 |
| Emily Mackevicius | 联合创始人,高级研究科学家 | 负责 Collaborative Intelligent Systems;Basis 描述其 Columbia 博士后经历和 MIT 博士学位。 | 为智能体设计增加协作智能和实验研究深度。 | 中 |
| Eli Bingham | 联合创始人,董事 | 据 Basis 介绍,其为 Broad Institute 研究员、Pyro 共同创建者,并曾任 Uber AI Labs 高级研究科学家。 | 为创始团队带来概率编程和 ML 系统可信度。 | 中 |
| Annette Garcia | 会计产品运营负责人 | 在审计和企业内部会计经历后,于 2024 年作为第五位全职员工加入。 | 把从业者标准带入产品、质量、税务和审计扩张。 | 中 |
| Keith Rabois | Khosla Ventures 派驻董事 | Series B 披露公开称其作为 Khosla 代表进入董事会。 | 增加治理、融资杠杆和后期阶段网络资源。 | 中 |
| Miles Clements | Accel 派驻董事 | Series B 披露称 Accel 合伙人加入董事会。 | 代表新领投方监督,并带来 go-to-market 模式识别。 | 中 |
公开领导层披露只是部分而非完整清单。该表列出可支撑的五位创始人、最清晰的面向会计师的运营负责人,以及公开投资方董事姓名。
[CO006, CO007, CO016, CO030, CO031, CO037]1.3 融资历史、投资人基础与利益相关方图谱
即便股权结构细节仍属私密,Basis 现在已有连贯的公开资本历史。公司在 2023 年 10 月以 $3.6 million 种子轮公开亮相,2024 年 12 月完成由 Khosla Ventures 领投的 $34 million Series A,随后在 2026 年 2 月宣布以 $1.15 billion 估值完成 $100 million Series B。合并这些轮次后,累计融资约 $137.6 million,几家外部媒体四舍五入为约 $138 million。Series B 财团显著拓宽了利益相关方基础:Accel 和 GV 以领投方身份进入,Khosla 继续加码,来自软件、基础设施和金融领域的其他支持者和天使投资人仍然参与。投资人名单重要,不只是因为它给 Basis 资本;它还增加了公司面向会计事务所、模型实验室和未来企业买家的分发、招聘和战略可信度。关键公开缺口在于未披露的部分:保留来源集中没有债务、仓储融资、要约收购或其他二级交易证据,因此股权故事可见,但完整资本结构故事不可见。[CO011, CO012, CO013, CO014, CO015, CO016]
| 利益相关方 | 角色 | 控制权或经济重要性 | 尽调问题 |
|---|---|---|---|
| Accel | Series B 领投方和董事席位 | $1.15B 台阶上升中的新领投方;Miles Clements 加入董事会。 | 获取董事会权利、按比例认购条款和任何治理契约。 |
| GV (Google Ventures) | Series B 领投方 | 在成长期轮次增加前沿 AI 信号和分发可信度。 | 厘清战略商业期待与纯财务所有权之间的边界。 |
| Khosla Ventures / Keith Rabois | Series A 领投方和持续董事会影响力 | 连接 2024 年 Series A 和 2026 年 Series B,维持治理连续性。 | 复核保护性条款、持股和董事会委员会影响力。 |
| Better Tomorrow Ventures | 种子轮领投方 | 最早的机构支持者,具备 fintech / 会计专业化。 | 梳理当前持股和任何跟投参与权。 |
| Wiss | 参考客户和公开采用方 | 提供 CAS 中 beta 使用、上线和运营价值的第三方证据。 | 请求部署指标、续约条款和跨业务线扩张数据。 |
| Baker Tilly | 类分发合作伙伴 | 显示其有潜力走出单事务所试点动态,进入规模化上线。 | 厘清经济模式是直接 SaaS、托管服务,还是类似渠道的收入分成。 |
该利益相关方图谱混合了资本提供方和运营伙伴,因为两类群体都会实质影响 Basis 的商业位置和尽调路径。
[CO011, CO012, CO013, CO014, CO015, CO016]1.4 规模信号、客户验证与企业就绪姿态
Basis 披露的牵引力质量多于绝对规模。当前最强信号不是原始客户数量,而是切入大型事务所:Basis 称自己服务约 30% 的 Top 25 会计事务所;其招聘页面还称公司已部署于头部事务所,并在每位会计师背后运行数百个代理。公开验证点也比泛泛的 logo 列表更强。Wiss 披露 2023 年 11 月开始 beta,2024 年 9 月扩大 CAS 部署;Basis 则宣布 2026 年与 Baker Tilly 合作,目标是在数千名会计客户中推进自动化。产品宽度也已超出早期 CAS 定位;到 2026 年,公开材料描述了跨 CAS、税务和审计的部署,包括一个展示的端到端 1065 税务流程。安全姿态是另一个当前信号。Basis 称其已获得 SOC 2 Type II、ISO 27001 和 ISO 42001 认证,数据留在美国,不用客户数据训练模型,并为代理动作提供审计轨迹。主要但书是披露不透明:没有保留来源支撑当前 ARR、收入、精确员工数或准确办公室足迹,安全页面仍把 GDPR 和 CCPA 列为即将推出。[CO018, CO019, CO020, CO021, CO022, CO023]
| 指标 | 数值 / 状态 | 日期 | 置信度 | 缺口 / 备注 |
|---|---|---|---|---|
| 法律实体 | Essex Labs, Inc.(以 Basis 名义经营) | 2026 年网站状态 | 高 | 直接来自隐私政策。 |
| 成立时间 | 2023 | 2023-10-18 发布披露 | 高 | 由种子轮发布材料和后续 2026 年融资报道支持。 |
| 总部 | New York City(外部报道) | 2026-02-24 至 2026-02-26 | 中 | 独立报道将公司定位在 New York City;官方网站未公布地址。 |
| 阶段 | 私营 Series B / 独角兽估值 | 2026-02-24 | 高 | 官方 Series B 材料支持最新阶段标签。 |
| 累计融资(USD millions) | 137.6 | 2026-02-24 | 中 | 根据披露的种子轮、Series A 和 Series B 计算;外部报道约为 138。 |
| 最新估值(USD millions) | 1150 | 2026-02-24 | 高 | 公司和独立报道都指向 $1.15B。 |
| 客户规模信号 | 前 25 大会计师事务所中约 30% | 2026-02-24 | 高 | 这是渗透率指标,不是绝对客户数。 |
| 收入 / ARR | 保留下来的公开来源没有披露收入、ARR、运行率或留存指标。 | |||
| 员工数 | 招聘和团队资料暗示公司在增长,但当前员工数未公开披露。 | |||
| 地点 / 办公足迹 | New York City 有外部报道,但公司未公布完整办公室或地点清单。 |
供后续章节使用的标准封面指标。公开记录只支持方向性信号或第三方地点引用的地方,null 单元格是有意保留。
[CO005, CO008, CO009, CO010, CO013, CO015]展示会计痛点、Basis 智能体平台、人工审阅、模型合作和客户验证如何连成一个公司系统。
[CO002, CO003, CO018, CO024, CO025, CO033]用可支撑的当前信号,而不是未经支持的经营指标,概括成熟度、牵引力、披露质量和信任状态。
这张图有意混合数字和状态类指标,因为 Basis 未发布足够原始经营指标,无法做纯财务 KPI 面板。
[CO013, CO015, CO018, CO019, CO026, CO038]1.5 里程碑、模型合作与需带入后文的谨慎信号
带日期的公开时间线如今足够丰富,可以作为后续章节复用的记录。Basis 创立于 2023 年,2023 年 10 月随种子轮融资公开亮相,披露 Wiss beta 从 2023 年末开始,2024 年 12 月完成 $34 million Series A,2025 年 8 月发布详细的 OpenAI 进展叙事,并在 2026 年 2 月的 Series B 后进入新的规模区间。2026 年 3 月和 4 月又加入 Codex 聚焦和 Baker Tilly 合作;2026 年 6 月,Annette Garcia 的公开评论把税务和审计扩张路径说得更明确。因此正面故事很清楚:资本更强、流程宽度扩大、合作伙伴验证可见。谨慎故事同样清楚。Basis 所在领域里,可信 AI、网络安全治理和可复核审计轨迹不是可选项。SEC、NIST 以及更广泛的专业服务研究都指向金融流程中 AI 系统更高的合规门槛。这不削弱公司论点,但意味着后续风险工作应把信任、治理和披露完整性视为一阶尽调事项,而不是锦上添花。[CO020, CO024, CO025, CO033, CO034, CO035]
| 日期 | 事件 | 类型 | 金额 / 估值 / 状态 | 参与方 | 含义 |
|---|---|---|---|---|---|
| 2023 | 公司成立 | 成立 | Basis 成立 | Matthew Harpe;Mitchell Troyanovsky;Zenna Tavares;Emily Mackevicius;Eli Bingham 等联合创始人 | 建立公开公司记录的起点,同时注明后续材料在商业化叙事和研究型创始人叙事之间有所拆分。 |
| 2023-10-18 | 通过种子轮和首个产品发布公开亮相 | 融资 | $3.6M 种子轮 | Basis;Better Tomorrow Ventures;BoxGroup;Avid Ventures 等参与方 | 创建关于身份、产品和投资人支持的首个耐久公开证据。 |
| 2023-11 | Wiss 开始在 CAS 中 beta 使用 | 产品 | Beta 部署启动 | Wiss;Basis | 来源集中最早保留下来的第三方客户证据点。 |
| 2024-09-16 | Wiss 合作公开宣布 | 合作 | 全面 CAS 部署叙事 | Wiss;Basis | 显示 Basis 从 beta 推进到全事务所部署措辞。 |
| 2024-12-17 | Series A 宣布 | 融资 | $34M Series A | Basis;Khosla Ventures;现有支持方 | 标志着公司从种子阶段发布进入资本更充足的建设阶段。 |
| 2025-08-12 | OpenAI 进展叙事发布 | 产品 | 模型进步到智能体的投资逻辑 | Basis;OpenAI 模型 | 显示 Basis 将模型改进定位为核心运营优势,而非后端管道。 |
| 2026-02-24 | Series B 宣布 | 融资 | $100M,估值 $1.15B | Basis;Accel;GV;Khosla;其他投资人 | 将 Basis 推入独角兽层级,并扩大董事会 / 投资人监督。 |
| 2026-02-24 | 前 25 大渗透率和 1065 demo 披露 | 规模 | 前 25 大事务所约 30%;端到端 1065 工作流 | Basis;会计师事务所客户 | 提供公开记录中最强的当前牵引力和工作流深度信号。 |
| 2026-03-10 | Codex 客户聚焦发布 | 合作 | OpenAI 相关产品聚焦 | Basis;Mitch Troyanovsky | 让 OpenAI 关系继续出现在公司的技术可信度叙事中。 |
| 2026-04-29 | Baker Tilly 合作宣布 | 合作 | 覆盖数千名会计客户的上线 | Basis;Baker Tilly | 暗示从点状部署扩张到更大的托管服务分发。 |
| 2026-06-09 | Annette Garcia 访谈详述 APO 角色及税务 / 审计扩张 | 治理 | 领导层和工作流扩张细节 | Basis;Annette Garcia | 说明 Basis 如何随工作流拓宽而配置领域专业能力。 |
| 2026-06-11 | 安全页面仍显示 GDPR 和 CCPA 即将推出 | 反向 | 合规工作明显仍在推进 | Basis | 提醒后续章节:企业就绪主张很强,但部分公开领域仍不完整。 |
这是本章唯一的记录时间线。它混合公司披露、合作伙伴证据和带日期的网站状态观察,让后续章节可以复用一致的里程碑序列。
[CO005, CO011, CO012, CO013, CO016, CO018]展示从创立和首次发布,到融资、客户验证、技术定位更新,以及最新公开就绪度提醒的顺序。
公开来源只能把创立记录支撑到年份层面,所以开头里程碑使用 2023 年,而不是编造精确日期。
[CO005, CO011, CO012, CO013, CO016, CO023]1.6 图表
02市场分析
2.1 市场定义与范围
分析 Basis 时,应对标专业会计事务所流程自动化市场,而不是所有财务软件支出或经济体中每一名会计从业者。公司自己把产品定位在 CAS、税务、审计和咨询之间,这与面向客户出售经常性和项目制专业服务的公共会计事务所最直接契合。BLS 职业定义确认,会计劳动力横跨公共会计师、管理会计师、政府会计师和审计师,因此广义会计劳动力市场大于 Basis 的直接购买中心。这个区分重要,因为大量软件市场服务的是 controller、CFO、薪资团队或行业特定财务运营,而不是 CPA 事务所。 因此,本章纳入的支出包括:公共会计服务收入、销售给会计事务所的软件和流程工具,以及可在事务所内部替代或增强会计师劳动的邻近自动化预算。排除或仅部分纳入的支出包括企业 ERP、通用 AP 自动化,以及企业财务团队购买的内部 controllership 软件,除非这些工具可作为 Basis 想要自动化的同一会计工作的实际替代品。这个边界让本章保持诚实:它保留了会计支出的大外沿,同时避免把每一美元记账、薪资或 finance-ERP 支出都视为同样可由面向事务所的代理平台触达。[CM001, CM002, CM003, CM012, CM049, CM052]
| 细分 / 类别 | 纳入支出 | 排除支出 | 买方 / 付款方 | 与 Basis 的相关性 |
|---|---|---|---|---|
| 公共会计服务 | CPA 事务所获得的审计、税务、CAS、咨询和外包会计收入 | 企业内部会计薪酬和 ERP 许可证 | 管理合伙人、业务线负责人、事务所运营预算 | 主要 TAM 边界 |
| 记账 / 税务 / 薪资服务组合 | 事务所交付的经常性 SMB 合规和后台工作 | DIY 消费者税务软件和纯消费者金融 app | 事务所所有者、CAS 负责人、客户服务线预算 | 相关性高,因为 Basis 自动化重复性事务所工作 |
| 会计师事务所软件栈 | 执业工具、工作流软件、面向会计师的云套件、自动化预算 | 无关的企业 IT 支出 | CIO、COO、创新负责人、事务所合伙人 | 主要 SAM 边界 |
| 企业财务 / 主计长软件 | GL、关账、AP、报告、ERP 和财务运营工具 | 外部 CPA 服务收入 | 主计长、CFO、财务 IT | 相邻替代市场,不是 Basis 的第一个切入点 |
| 技术赋能会计替代方案 | 来自记账或财务提供商的软件加服务混合方案 | 没有技术层的纯人工本地记账 | SMB 所有者、财务负责人、部分会计师事务所 | 重要竞争品类 |
Basis 位于公共会计人工支出和用于压缩该人工的软件 / 工具预算之间。本章排除广义 ERP 支出,除非它替代的是同一项待完成工作。
[CM001, CM002, CM003, CM012, CM039, CM043]2.2 TAM、SAM 与受证据约束的规模测算
最宽且可支撑的 TAM 口径是美国会计服务收入。IBISWorld 把 2026 年美国市场规模定为 $157.4 billion,AnythingResearch 则显示 2025 年为 $145.4 billion,前四大份额为 42.8%。这些数字方向一致——Basis 所处的是一个很大的服务池——但它们并不直接等同于软件支出,也不等同于 Basis 可变现的表面。在软件侧,The Business Research Company、Mordor Intelligence 和 Precedence 都把 2025-2026 年全球会计软件市场放在约 $21 billion 到 $23 billion,披露区域份额的来源显示北美约占 38%。这意味着,在调整 Basis 面向会计事务所而非所有企业财务用户之前,北美软件支出的外沿大约是 $8 billion 到 $9 billion。 “跨 CAS、税务和审计销售给会计事务所的 AI 自动化”的精确 SAM 未公开披露。最好的公开方法是把北美会计软件作为上限,再按买方类型收窄。Intuit、Sage 和 Xero 都已经卖给会计师;Botkeeper、Pilot、Numeric 和 Digits 则说明,买方可以通过记账自动化、技术赋能服务、结账软件或 AI 原生总账来解决同一痛点。因此 Basis 的近期 SAM,是愿意在多项业务中采用垂直、可复核 AI 的大型美国会计事务所子集。公开来看,最强的 SOM 代理不是收入份额,而是切入约 30% 的 Top 25 会计事务所。这是楔入强度的有意义证据,但若没有定价、席位数或 ACV 披露,还不足以计算精确收入份额。[CM004, CM005, CM006, CM007, CM008, CM009]
| 发布方 / 视角 | 年份 | 地域 | 数值 | CAGR / 增长 | 方法 | 置信度 | 局限 |
|---|---|---|---|---|---|---|---|
| IBISWorld 会计服务 | 2026 | 美国 | $157.4B | 1.1% 年化(2021-2026) | 会计服务行业收入估计 | 高 | 广义服务市场,不是 Basis 专属软件预算 |
| AnythingResearch 会计服务 | 2025 | 美国 | $145.4B | 5.4% 五年增长 | 基于联邦数据的行业市场研究 | 中 | 2025 年数值,且行业组合比 Basis 精确切入点更宽 |
| The Business Research Company 会计软件 | 2026 | 全球 | $22.72B | 2030 年前 CAGR 13.4% | 全球软件收入估计 | 中 | 包含会计师事务所以外的企业财务和其他终端用户 |
| Mordor Intelligence 会计软件 | 2026 | 全球 | $23.47B | 2031 年前 CAGR 8.85% | 全球软件市场模型 | 中 | 不是仅美国或仅事务所口径 |
| Precedence 会计软件 | 2026 | 全球 | $23.1B | 到 2035 年 CAGR 9.15% | 自上而下 / 自下而上的市场估算 | 中 | 仍宽于会计师事务所自动化 |
| 推导出的北美软件上限 | 2025-2026 | 北美 | $8B-$9B | n/a | 将约 38% 的北美占比套用于全球会计软件估算 | 中 | 上限代理值,并非披露的会计师事务所 SAM |
宽口径服务 TAM 很大,但可直接变现的软件与自动化切片更小,披露也更不透明。Basis 的实际 SAM 是这些视角下的子集,而不是一条单独公开的市场项目。
[CM004, CM005, CM006, CM007, CM008, CM009]公开市场估算差异很大,因为它们衡量的层级不同:宽口径美国服务收入,或全球会计软件收入。不能只拿一个标题 TAM 来承销 Basis。
low/value/high 相同表示披露的是点估计,而不是统计区间。服务和软件数值放在一起,是为了展示边界差异,不代表可直接比较。
[CM004, CM005, CM008, CM010, CM011, CM053]Basis 的市场最好看成嵌套视角,而不是干净的瀑布:宽口径服务外壳、较窄的软件边界,再到更窄的会计事务所自动化楔子。
这是一个受证据约束的视角栈。只有外层直接发布;内层用公开买方和区域份额证据划定边界。
[CM004, CM005, CM008, CM010, CM011, CM047]2.3 需求驱动、采用节奏与约束
最大驱动因素是会计师劳动力短缺。BLS 仍预计每年有 124,200 个岗位空缺,薪资中位数已升至 $81,680,行业主导的报道持续描述结构性疲弱的人才管道。NPAG 专为解决短缺而成立;NYSSCPA 对 AICPA 工作的总结强调学生转入会计专业的转化疲弱;The CPA Journal 则把人手不足与上升的报告和审计质量风险联系起来。换句话说,事务所需要 Basis,不是因为会计很时髦;它们需要杠杆,是因为劳动力市场紧、贵,而且运营脆弱。薪酬成本上升进一步强化这种压力。2026 年 3 月 Employment Cost Index 显示,民用劳动者工资同比增长 3.4%;这不是会计专属数据,但仍支持劳动力成本持续通胀的大背景。 第二个驱动因素是流程复杂度。会计工作由截止日期驱动、受监管且复核密集。PCAOB 经 SEC 批准的技术辅助分析案卷强调,技术正更深嵌入审计程序,但控制门槛更高而不是更低。这为能让工作更可复核、更一致、更可追踪的工具创造了有利环境;也会压制任何承诺“完全自主”却没有清晰控制的供应商。Thomson Reuters 的 2026 年数据很好地捕捉了采用阶段:GenAI 使用快速上升,但代理式 AI 和 ROI 衡量仍处早期。买方想要 AI,但强制要求仍不常见,正式 ROI 纪律仍薄。这意味着 Basis 进入了一个接受度高的市场,但仍需要重度建立信任、实施支持,并证明能减少复核者工作量。[CM013, CM014, CM015, CM016, CM017, CM018]
| 驱动因素 / 约束 | 方向 | 时间 | 含义 | 尽调问题 |
|---|---|---|---|---|
| 会计人才短缺 | 正向 | 当前且具结构性 | 事务所层面更容易为替代人力的自动化找到理由 | 要求按业务领域和岗位层级提供部署指标 |
| 薪酬和用工成本上升 | 正向 | 当前 | 只要实施成本受控,软件 ROI 叙事会更强 | 要求提供回本周期、席位经济性和审阅时间节省 |
| 监管复杂度与审计质量压力 | 需求正向,速度负向 | 持续 | 支撑对可追踪自动化的需求,但抬高举证门槛 | 要求提供审计轨迹、控制和异常率证据 |
| AI 采用仍处早期 | 近期负向 | 当前 | 预算推进快于标准化 ROI 衡量 | 要求提供客户从试点走向广泛部署的推进路径 |
| Intuit / Sage / Xero 的现有厂商捆绑 | 负向 | 当前 | 独立供应商面临更高切换成本和定价压力 | 要求提供相对现有技术栈扩展的赢单 / 输单数据 |
| 人工审阅和信任要求 | 负向 | 持续 | 限制全自主渗透率,并拉长实施周期 | 要求提供审阅人投入、改写率和准确率基准 |
让 Basis 有吸引力的条件,也会拖慢品类成熟。短缺和复杂度推高紧迫性,但信任要求和现有厂商分发约束了支出迁移速度。
[CM013, CM014, CM015, CM016, CM017, CM018]劳动力稀缺和 AI 兴趣正把品类往前推,但规模化智能体部署和 ROI 纪律仍处早期。
[CM013, CM014, CM015, CM016, CM020, CM021]2.4 按事务所规模划分的买方分层与竞争动态
Basis 最好的公开适配对象是较大型公共会计事务所,尤其是业务量和流程标准化足以支撑多业务自动化的事务所。Basis 自身的公开足迹加上 Wiss 推广都支撑这一点。产品被放在 CAS、税务、审计和咨询之间,最强的公开客户标识验证也不是很小的本地事务所,而是一个 Top 100 风格的买方把产品用作力量倍增器。这很重要,因为采用路径会随事务所规模急剧变化。全国性和 Top 100 区域事务所可以用创新、CIO 或业务转型预算资助试验;它们也有足够的流程重复,让 AI 产生可见杠杆。中型市场事务所仍可能采用,但价格敏感度更高,也更可能把 Basis 与现有技术栈扩展或技术赋能服务替代方案比较。 竞争压力已经很宽。Intuit 正围绕 AI 驱动流程和统一数据为事务所重新包装 QuickBooks;Sage 和 Xero 把自己定位为面向会计师的云系统;Botkeeper 和 Pilot 销售人机结合的运营模式;Numeric 和 Digits 以明确的自动化主张攻击结账和总账流程。因此市场并没有等 Basis 来定义品类。Basis 的优势在于,它把垂直、端到端流程执行卖进会计事务所场景,在那里人才稀缺、质量门槛和事务所经济模型相互碰撞。它的风险在于,现有厂商掌握既有流程表面,更窄的供应商可能给出更清晰的 ROI 或更容易的实施。公开证据支持 Basis 在头部事务所形成强早期楔子,但还不能证明该品类已经被其普遍赢下。[CM033, CM034, CM035, CM036, CM037, CM038]
| 细分市场 | 买方 | 用户 | 付款方 | 工作流 | 预算负责人 | 采用触发因素 |
|---|---|---|---|---|---|---|
| Big 4 / 全国性网络 | 创新负责人、业务线负责人、CIO | CAS、税务、审计团队 | 事务所转型或技术预算 | 多业务线工作流执行 | COO / CIO / 管理合伙人 | 在保住控制的同时,把稀缺人力撬动到大客户群 |
| Top 25 / Top 100 区域事务所 | 业务负责人或运营高管 | CAS 和税务团队先用,随后扩到审计 / 咨询 | 业务改进预算 | 跨客户工作流标准化 | 管理合伙人 / COO / CIO | 需要在不按比例招人的情况下扩规模 |
| 中端市场区域和本地事务所 | 所有者合伙人或 CAS 负责人 | 记账员、高级员工、经理 | 合伙人管理的运营预算 | 选择性工作流自动化 | 管理合伙人 | 守住利润率和员工生活质量 |
| 购买外包服务的 SMB 客户 | 所有者 / 创始人 / 运营者 | 外部会计师或记账员 | 运营预算 | 外包账务、税务、报告 | 企业所有者 | 比内部招聘更便宜,也更可靠 |
| 企业财务团队 | Controller / CFO | 会计和结账团队 | 财务系统预算 | 结账、AP、报告、ERP | CFO / 财务 VP | 更快结账或合规,但不是 Basis 公开主打的切入口 |
Basis 最强的公开证据落在前两行。企业财务这一行重要,主要因为它界定了替代品和分析师市场边界,而不是因为 Basis 已公开在这里领先。
[CM001, CM033, CM036, CM037, CM040, CM041]公开证据显示,事务所会从劳动力痛点和试点赞助,走向受控推广,而不是直接切到全公司全自主部署。
[CM018, CM020, CM032, CM047, CM048, CM050]03竞争对手
3.1 竞争格局类别,以及什么才算真正竞争对手
Basis 的竞争集合应按待完成任务拆分,而不是把任何提到 AI 的产品都算进去。Basis 营销的是可跨 CAS、税务、审计和咨询执行会计流程的代理,然后把可复核输出交回给会计师。因此最接近的直接同业,是同样试图在事务所内部掌控会计流程执行的供应商,而不只是弹出聊天框或自动化一个子任务。基于保留的公开证据,Botkeeper、Truewind 和 Digits 是最相关的直接比较对象,因为它们会计原生、理解流程,并销售给会计师或会计事务所。Pilot 在客户结果层面重叠,但模式更偏服务。Vic.ai 和 BILL 是更窄的 AP 专家。QuickBooks 和 Microsoft 重要,是因为即便它们不是最深的会计原生系统,也可以通过现有软件预算和熟悉界面进入。这个类别拆分很重要,因为 Basis 即便在流程深度上最好,也可能在买方更重视既有安装基础、较低公开价格或外包结果,而非一个新平台时,把预算输给现有厂商、替代方案或服务模式。[CP001, CP002, CP003, CP016, CP017]
| 竞争对手 | 类别 | 规模信号 / 融资可见度 | 目标客户 | 产品范围 / 差异化 | 公开限制或注意事项 |
|---|---|---|---|---|---|
| Basis | 直接会计原生工作流平台 | $100M Series B,估值 $1.15B;保留来源称 / 支持约 30% 的美国 Top 25 事务所 | 中大型会计师事务所 | 覆盖 CAS、税务、审计和咨询的端到端会计代理,输出可进入审阅 | 公开定价和合同结构未披露 |
| Botkeeper | 直接同业 / 面向事务所的记账自动化 | 官方网站称有 200+ 家会计师事务所和 5,000+ 家企业客户;保留页面未披露融资 | 会计师事务所 | AI 记账,配套专门服务、会计师事务所专用工作流和公开的按许可证定价 | 公开叙事在记账和产能上最强,而不是税务 / 审计广度 |
| Truewind | 直接同业 / 结账自动化层 | 首页称获 500+ 名会计信任;保留页面未披露融资 | 使用 Sage Intacct 或 QBO 的会计团队和事务所 | 可入 GL 的日记账分录、对账、交易编码、明细表,以及审阅优先的结账自动化 | 公开范围更集中在结账和真实来源集成,而不是多业务线事务所工作流所有权 |
| Digits | 直接同业 / 总账替代 | 声称服务数千家初创公司、小企业和会计师事务所;定价公开;保留页面未披露融资 | 企业和会计师事务所 | AI 原生总账、代理式结账、账单支付、开票、事务所品牌、客户门户 | 比纯会计师事务所覆盖更多客群;专属事务所定价仍需销售沟通 |
| Pilot | 服务占比较高的替代品 | 声称服务 3,000+ 家增长驱动的初创公司和小企业;定价公开 | 初创公司和小企业 | 记账、税务、CFO 和外包运营,采用人加软件的交付方式 | 更接近外包财务,而不是会计师事务所原生代理软件 |
| Vic.ai | 相邻专科 / AP 自动化 | AP 效率指标公开;保留页面未披露融资 | 企业财务团队 | AP 的发票处理、审批、付款和自主财务 | 窄于 Basis,因为公开范围以 AP 为中心 |
| QuickBooks + Intuit Assist | 现有 SMB 套件替代品 | 保留页面未量化装机规模;AI 功能捆绑进现有 QuickBooks 路径 | SMB 和现有 QuickBooks 用户 | QuickBooks 内的工作流提示、开票帮助和财务问答 | 保留证据未显示会计师事务所原生的多业务线工作流所有权 |
| Microsoft 365 Copilot | 横向现有厂商替代品 | 保留页面未量化装机规模;捆绑杠杆来自现有 Microsoft 365 覆盖面 | 现有 Microsoft 企业用户 | 聊天、搜索、代理、应用集成和工作图谱上下文 | 横向生产力层,而非会计原生记录系统 |
| BILL | 相邻专科 / AP-AR 自动化 | 存在会计师事务所渠道;保留页面未显示标价 | SMB、中型企业、会计师 | 应付、应收、采购、审批和付款工作流 | 比 Basis 的运营切入口更窄,且没有保留证据显示税务 / 审计工作流范围 |
画像行优先呈现公开可见的范围、目标买方和规模信号。保留来源未披露融资或客户数量时,单元格直接说明未披露,而不是推断。
[CP003, CP006, CP007, CP009, CP011, CP013]3.2 直接会计原生同业:Botkeeper、Truewind 和 Digits
Botkeeper、Truewind 和 Digits 都与 Basis 故事中有意义的一部分重叠,但方式不同。Botkeeper 明确为会计事务所打造,并已声称拥有超过 200 家事务所和 5,000 多个企业客户,这给了它真实的安装基础可信度;同时,它仍把产品描述为 AI、支持和专属帮助的混合体,而不是多业务代理平台。Truewind 在复核优先的流程自动化上更接近,但其公开定位仍集中在结账、对账、交易编码和附表准备,并把 Sage 或 QBO 保持为事实来源。Digits 是三者中结构上最有野心的,因为它销售 AI 原生总账、公布入门价格,并推动事务所采用品牌化客户体验和代理式结账运营。放在一起看,这些同业说明 Basis 并不是唯一试图自动化会计师流程的公司;它们也说明市场在重服务记账、以结账为中心的 copilot 层和完整总账替代之间相当碎片化。当买方想要跨多个业务线的可复核执行,而不只是记账或结账提速时,Basis 的差异化主张最强。[CP004, CP005, CP006, CP009, CP010, CP011]
| 购买标准 | Basis | Botkeeper | Truewind | Digits | Pilot | Vic.ai | QuickBooks | Microsoft | BILL |
|---|---|---|---|---|---|---|---|---|---|
| 会计师事务所原生定位 | 强 | 强 | 中 | 中 | 弱 | 弱 | 弱 | 弱 | 弱 |
| 跨多项会计业务的端到端工作流执行 | 强 | 中 | 弱 | 中 | 弱 | 弱 | unknown | 弱 | 弱 |
| 结账 / 对账自动化 | 强 | 中 | 强 | 强 | 弱 | 弱 | 弱 | 弱 | 弱 |
| 税务工作流相关性 | 强 | 弱 | 弱 | 弱 | 中 | 弱 | 中 | 弱 | 弱 |
| 审计工作流相关性 | 强 | 弱 | 弱 | 弱 | 弱 | 弱 | unknown | 弱 | 弱 |
| 总账替代 / 记录系统野心 | 弱 | 弱 | 弱 | 强 | 弱 | 弱 | 强 | 弱 | 弱 |
| 默认包含人工服务层 | 弱 | 强 | 弱 | 弱 | 强 | 弱 | unknown | 弱 | 弱 |
| 公开企业信任姿态 | 强 | 中 | 中 | 中 | unknown | 强 | unknown | 强 | 中 |
矩阵值是基于保留官方页面和客户证据的有边界综合。标为未知的单元格反映缺少公开证据,不代表缺少能力。
[CP001, CP004, CP007, CP009, CP010, CP011]用 1-10 的序数分值比较工作流所有权深度(x 轴)与分销或装机基础杠杆(y 轴)。
分值是有证据支撑的综合判断,不是经审计的市场份额或功能数量指标。工作流所有权反映供应商公开声称能端到端执行多少会计工作;分销反映装机基础杠杆、服务足迹或采购熟悉度。
[CP001, CP003, CP006, CP011, CP013, CP016]3.3 现有厂商、邻近替代品与现状
竞争分析中最大的错误,是只把 Basis 与其他创业公司供应商比较。Pilot、Vic.ai、BILL、QuickBooks、Microsoft 以及内部自建现状,都从不同角度攻击同一个劳动力池。Pilot 为创业公司和小企业提供记账、税务和 CFO 服务,因此当买方更想要外部团队而不是新软件栈时,它就形成竞争。Vic.ai 和 BILL 更窄,但 AP 专家仍可在财务团队或会计事务所投入更广流程平台之前,吸收自动化预算和运营注意力。Intuit Assist 和 Microsoft 365 Copilot 又不同:基于保留公开证据,它们不是会计事务所原生产品,但继承了 QuickBooks 和 Microsoft 365 的分发,降低了已经生活在这些系统里的买方采用摩擦。现状同样重要。事务所已经运行总账、电子表格、业务工具、复核习惯和客户沟通闭环,这些都会形成惯性。因此 Basis 不只需要胜过某一个产品;它要打败的是一组既有软件、人工流程和熟悉的流程捷径。[CP007, CP008, CP013, CP014, CP015, CP016]
| 提供商 | 公开入门价格 / 合同模式 | 单位 | 保留证据中的包含能力 | 定价可见度 | 对 Basis 的含义 |
|---|---|---|---|---|---|
| Basis | 未知 / 销售主导 | Unknown | 覆盖 CAS、税务、审计、咨询的会计代理;合同细节未公开 | 低 | 工作流深度可能支撑溢价定价,但没有管理层材料就无法做公开基准 |
| Botkeeper | 1-4 个许可证 $149/mo;25+ 个按月 $59/mo;25+ 个按年 $53/mo | 按许可证 / 面向客户的产能 | 记账自动化、洞察、支持、安全、协作功能 | 高 | 透明的低端席位定价,给买方提供了对照 Basis 不透明性的即时基准 |
| Pilot | 记账入门 $99/mo;CFO 起价 $1,750、$3,150 和 $5,250/mo | 月度计划 / 服务包 | 记账、税务、CFO、外包运营,加人工支持 | 高 | Pilot 把预算框定为外包财务支出,而不是纯软件支出 |
| Digits | 每月 $65 / $100 / $250,外加合作伙伴提供的服务 | 按公司订阅 | AI 记账、账单支付、开票、看板、报告、事务所选项 | 高 | 低公开入门价可能让 Digits 成为较小买方和事务所的有吸引力默认选项 |
| QuickBooks + Intuit Assist | Intuit Assist 面向部分 QuickBooks 用户免费提供 | 订阅内捆绑功能 | 开票帮助、工作流自动化、财务问答 | 中 | 捆绑 AI 可能让买方觉得它比采用新的独立工作流平台更便宜 |
| Microsoft 365 Copilot | Copilot Chat 面向符合条件的用户免费提供;完整 Copilot 需要额外许可证 | 捆绑 / 附加许可证 | 聊天、代理、应用集成、企业控制 | 中 | 现有 Microsoft 买方可能先测试横向 AI,再购买会计专用软件 |
| Vic.ai | 演示 / 联系销售 | 企业合同 | APSuite、发票处理、审批、付款、分析 | 低 | AP 专科厂商可以在不公开标价的情况下拿到预算 |
| BILL | 演示 / 联系销售 | 保留页面未公开订阅 / 交易合同 | AP、AR、审批、采购、付款 | 低 | BILL 可以争取应付款自动化预算,而不必证明多业务线工作流广度 |
各行保留了公开标价、捆绑驱动定价和不透明企业定价之间的差异。未知表示保留来源集没有披露可用的公开基准。
[CP018, CP019, CP020, CP021, CP022, CP023]以商业模式为导向的矩阵,比较哪些供应商拥有工作流、总账、服务层或套件入口。
这张图有意不同于详细功能表。它强调买方匹配和运营模式,而不是逐行核对会计能力。
[CP007, CP008, CP011, CP012, CP015, CP016]3.4 定价、包装、分发与信任姿态
公开定价是竞争集合分化最清楚的地方之一。Botkeeper、Pilot 和 Digits 都发布了有意义的包装信息,让买方能快速对标入门成本:Botkeeper 发布按授权分层,Pilot 发布记账和 CFO 计划,Digits 发布 SaaS 计划加合作伙伴提供的服务。Basis、Vic.ai 和 BILL 在保留证据中没有提供可比的公开价格表;Intuit 和 Microsoft 则更强调捆绑和附加逻辑,而不是干净的一对一会计事务所合同。这有两层含义。第一,价格敏感或自助型买方拥有比 Basis 更透明的替代方案。第二,如果 Basis 赢在流程深度或企业部署经济性,不透明定价仍可能合理,但这必须由私下证明,因为公开记录没有显示。信任姿态类似。Basis 控制很强,但 Botkeeper、Vic.ai、Digits 和 Microsoft 也营销安全、隐私或合规保障。因此公开信任主张看起来像进入交易的前置条件,而不是独立护城河。分发验证可能更重要:来自 Wiss 的客户证据和通过 Baker Tilly 的合作伙伴杠杆显示,Basis 可以超越孤立工具试点;而捆绑型现有厂商仍会威胁凭熟悉度和采购便利性胜出。[CP018, CP019, CP020, CP021, CP022, CP023]
| 护城河主张 | 威胁 | 证据 | 严重性 | 缓释措施 / 尽调问题 |
|---|---|---|---|---|
| Basis 凭会计原生的端到端工作流深度胜出 | 模型质量提升后,通用 copilot 工具和点状工具可能已经足够好 | Wiss 将 Basis 与 Copilot 作了对比;OpenAI 和 Microsoft 都显示出智能体能力在快速扩张 | 高 | 按业务线复核当前相对 QuickBooks、Microsoft 和直接同业的赢单 / 输单记录 |
| 可复核性和可追溯性支撑高风险工作流里的信任 | 各厂商的安全和隐私控制正在趋同,并变成基本门槛 | Basis、Botkeeper、Vic.ai、Digits 和 Microsoft 都在营销有分量的信任姿态 | 中高 | 要求提供并排控制证据、审计日志、升级规则和客户安全审查结果 |
| 大型事务所渗透形成有引用价值的滩头阵地 | Botkeeper 已有装机基础分发,既有套件厂商也已占住相邻工作流界面 | Botkeeper 声称覆盖 200+ 家事务所;QuickBooks 和 Microsoft 通过熟悉的技术栈进入 | 高 | 要求 Basis 按业务线提供扩张率、多业务附着率,以及买方选择 Basis 而非捆绑既有厂商的原因 |
| 合作伙伴和客户证明可加速分发 | 合作伙伴经济性和渠道结构未公开 | Wiss 和 Baker Tilly 显示出杠杆,但公开收入分成或转售条款缺失 | 中 | 要求提供合作伙伴管线、转售经济性,以及渠道影响的 ARR 占比 |
| 多业务范围扩大单客钱包份额机会 | AP 专项厂商可以先拿预算,并推迟更宽工作流平台的采用 | Vic.ai 和 BILL 聚焦更窄但 ROI 更清晰的品类,买方故事更容易讲 | 中 | 梳理 AP 或关账点状方案已落地后,Basis 如何进入 |
| 不透明定价可能支持价值型销售 | 不透明定价也可能拖慢与 Digits、Pilot、Botkeeper 等透明替代方案的评估 | 多个竞争对手有公开标价,Basis 没有 | 中高 | 要求提供价格手册、折扣走廊、实施费用,以及续约或扩张案例 |
严重性反映的是对 Basis 竞争耐久性的可能影响,而非损失确定性。该登记表聚焦公开证据,并有意把未知合同经济性留作尽调问题,而不是当成假设。
[CP030, CP031, CP032, CP034, CP037, CP039]用紧凑的公开代理指标概括 Basis 哪里最强、竞争耐久性哪里最脆弱。
各项是公开代理指标和分析摘要,不是经审计的份额或留存指标。它们是方向性信号,用来概括本章判断。
[CP018, CP019, CP020, CP023, CP032, CP035]3.5 转换成本、护城河耐久性与反向角度
Basis 的最好情况是,流程深度、可复核性和多业务会计原生设计,形成通用 copilots 和点状解决方案不易匹配的耐久楔子。公开客户验证支撑这个论点:Wiss 明确把 Copilot 与 Basis 作了对比,并称真正改变业务的是集成式端到端自动化,而不是泛化 AI 可用性。OpenAI 和 Basis 也把产品进展描述为随模型改进而复利,这可能扩大 Basis 能安全掌控的流程集合。反向情形同样重要。模型进步也可能商品化助手功能,缩小专业平台与捆绑资源丰富的现有厂商之间的差距。各供应商的公开信任主张正在收敛,因此安全本身不是护城河。Botkeeper 和 Digits 的公开价格会压低感知价值,而 QuickBooks 和 Microsoft 可以利用既有分发,让“够用”的 AI 显得更便宜、更容易采用。结果是一个看起来真实但有条件的护城河:当事务所想要跨多个业务的会计原生、可复核流程所有权时,Basis 看起来最强;当买方优化的是捆绑杠杆、外包或透明低入门价格,而不是深层流程控制时,Basis 看起来较弱。[CP030, CP031, CP032, CP035, CP036, CP041]
04财务
4.1 收入模式与定价架构
Basis 的公开财务表面更像卖给会计事务所的企业软件,而不是消费者 SaaS。官网展示了横跨 CAS、税务、审计和咨询的业务模块,但没有结账流程或价格表;法律页面反复把商业使用导向书面 MSA、订单和单独企业协议。这个组合意味着报价制合同、谈判范围,以及按流程定制部署,而不是透明的按席位自助定价。 因为 Basis 把自动化卖进高风险会计工作,可能的收入模式会把经常性平台费与上线导入、流程配置、支持,或许还有特定伙伴部署工作混在一起。官网证言横跨多家事务所和业务领域,Wiss 与 OpenAI 则描述多步骤流程、人工复核和可衡量的产能提升。公开证据足以说明 Basis 通过面向会计事务所的企业流程自动化变现;但不足以披露实际 ACV、折扣区间、最低合同规模,或软件与服务的拆分。[CI002, CI010, CI011, CI012, CI013, CI017]
| 来源 | 机制 | 单位 / 合同 | 当前公开状态 | 收入质量 | 尽调问题 |
|---|---|---|---|---|---|
| 企业平台订阅 | 面向会计事务所工作流的报价平台访问,覆盖 CAS、税务、审计和咨询 | MSA / 订单;可能按年或多年期 | 核心商业界面可见,但没有公开价格卡 | 如果续约强劲,可能是经常性且质量较高的收入 | 要求提供已签订单、计费节奏和续约条款 |
| 业务领域扩张 | 在事务所部门内增加部署,或进入新工作流 | 部门 / 工作流扩张 | Wiss 全面 CAS 推出和多业务定位提供支撑 | 如果标准化,质量高;如果每次扩张都要定制范围,质量较低 | 拆分扩张 ARR 和实施收入 |
| 实施和配置 | 工作流搭建、事务所特定标准、集成和入职 | 项目制或固定费用服务 | 企业协议和部署较重的客户证明暗示这一项存在 | 如果占比高,会稀释利润率 | 提供服务 SOW、实施周期和附着率 |
| 持续支持和治理 | 人工复核工作流、可审计性、合规文档、客户成功 | 捆绑或单独定价的支持 | 运营上必要,但未公开定价 | 如果深度嵌入,会有粘性;如果人力密集,可能压缩利润率 | 披露支持人员配置、包含的服务级别和升级经济性 |
| 合作伙伴 / 托管服务渠道 | 通过会计和咨询事务所把 Basis 嵌入客户交付来获取收入 | 渠道或类似转售的经济性 | Wiss 和 Baker Tilly 显示出渠道式分发潜力 | 如果伙伴能自助,分发杠杆好;如果 Basis 仍需深度介入,收入质量较低 | 披露合作伙伴经济性、收入分成和渠道贡献 ARR 占比 |
各行概括公开可见的变现界面;未披露商业条款不等于为零,需要尽调跟进。
[CI002, CI013, CI017, CI020, CI022, CI033]| 价格 / 合同元素 | 公开数值 / 状态 | 标价与实际 | 重要性 | 尽调问题 |
|---|---|---|---|---|
| 标价 | 网站没有公开标价 | 没有价格锚,外部无法测试交易规模或采用摩擦 | 要求提供当前价格表或企业报价样例 | |
| 合同载体 | 书面 MSA / 订单约定服务访问 | 实际条款经谈判确定,不是自助式 | 表明采用企业合同,也为定制范围或折扣留出空间 | 提供 MSA 模板、订单和付款条款 |
| 试点 / beta 条款 | Wiss 提到从 beta 到全面部署,但没有公开试点经济性 | 试点结构决定 CAC、转化率和服务负担 | 披露试点时长、转化率,以及试点是否收费 | |
| 实际 ACV / 折扣区间 | 没有公开 ACV、最低额或折扣指引 | 需要用它把采用证明连接到经常性收入规模 | 按事务所规模提供 ACV 分布和折扣审批政策 | |
| 计费基础 | 可能按事务所范围、工作流或业务领域报价 | 从企业部署证据推断,未直接披露 | 决定增长来自席位、工作流,还是更广的事务所采用 | 澄清主要计费单位和增购路径 |
空值表示未披露;合同结构可见,但实际定价和折扣未公开。
[CI010, CI011, CI012, CI020, CI045]Basis 似乎把会计师事务所需求转化为报价型企业合同、已部署工作流,以及附带支持层的经常性平台收入。
[CI002, CI010, CI020, CI033, CI045]4.2 GTM 代理指标与收入质量
公开需求代理指标真实但有选择性。Basis 和第三方反复称,公司服务约 30% 的前 25 大会计事务所;Wiss 称 CAS beta 扩展为全面部署,内部采用逐月上升。Baker Tilly 的 2026 年合作又加入一个规模化买方,用 Basis 做交易处理、日记账分录支持、对账和结账后报告。官网还展示了 Wiss、UHY、Clark Nuber、Boulay 和 BPB 的证言;对早期垂直 AI 供应商来说,这是异常具体的客户验证。 缺失的是投资人真正要承销的部分:合同期限、净收入留存、赢单率、销售周期长度、实际定价和客户集中度。OpenAI 案例研究称,事务所平均报告最高 30% 的时间节省;公司相关融资报道引用 20% 到 50% 的效率提升,因此 Basis 显然在销售产能扩张和更快复核周期。但这些 ROI 信号仍由公司渠道或合作伙伴引用,并非经审计收入数据。GTM 故事是可信的企业采用;收入质量故事仍未被证明。[CI003, CI016, CI017, CI018, CI019, CI020]
| 指标 | 公开数值 / 代理指标 | 置信度 | 重要性 | 尽调问题 |
|---|---|---|---|---|
| 收入 / ARR | 低 | 没有营收规模锚点,估值和回本都无法承保 | 按来源提供季度 ARR 和收入桥 | |
| 毛利率 | 低 | Basis 未披露毛利率;利润率路径是核心承保问题 | 分别披露软件、支持和服务毛利率 | |
| CAC / 回本 | 低 | 企业直销叠加合作伙伴销售,可能带来可观销售和入职成本 | 按细分提供 CAC、销售周期、赢单率和回本周期 | |
| 留存 / NRR | 低 | 全事务所部署可能有粘性,但没有公开队列证明 | 按客户队列提供总留存率、NRR 和扩张 | |
| ROI 代理指标 | 20%–50% 效率声明;合作伙伴 / 公司背书来源称最高可节省 30% 时间 | 中 | 如果 ROI 可重复,需求可能真实,但生产率证明不等于变现证明 | 分享与签约客户绑定的实测 ROI 研究 |
| 软件 vs. 服务组合 | 可比文件显示,企业平台通常把订阅收入与专业服务搭配销售 | 中 | 服务占比高会推迟软件式毛利率 | 披露经常性软件占比、实施占比和支持成本负荷 |
公开单位经济性披露基本缺失,因此该表把可观察代理指标与缺失的私有指标分开。
[CI018, CI021, CI028, CI030, CI031, CI037]公开证据支持一条从工作流验证走向扩张的路径,但定价和利润率节点大多仍未披露。
这座桥混合了公开证据点和明确未知项;它是逻辑图,不是公司披露的漏斗。
[CI015, CI018, CI021, CI025, CI036, CI047]4.3 成本结构与单位经济模型
Basis 不像是在通用模型外面套一层、低支持成本的包装层。安全页面宣传 SOC 2 Type II、ISO 27001、ISO 42001、仅美国托管、MFA、审计日志、备份和第三方测试。法律条款还补充,AI 输出不能替代专业判断,需要人工复核和验证。这些承诺在会计领域有商业帮助,但也意味着在每增加一美元毛利之前,公司要先承担固定合规、基础设施和客户成功成本。 公开可比公司强化了这种可能组合。Workiva 的 2025 10-K 称,92% 收入来自订阅和支持费,其余来自专业服务,并警告服务组合可能影响毛利率。BILL 的 2025 10-K 提到部分年度订阅,也提到开放式安排、定价敏感度和销售成本,同时报告规模化后的 81.7% GAAP 和 85.0% non-GAAP 毛利率。Basis 最终也许会像高毛利软件,但今天的公开记录更支持一种近期模型:企业软件加上线导入、伙伴推广和模型供应商开支,而不是已经实现的纯软件经济模型。[CI014, CI015, CI021, CI023, CI025, CI028]
Basis 的现金画像可能混合了有吸引力的经常性软件经济性,以及规模化前不轻的合规、部署和模型成本负担。
这个定性矩阵基于公开产品、法律、安全和可比备案证据,对现金强度驱动因素排序,而非依据内部财务报表。
[CI014, CI015, CI035, CI036, CI047, CI049]4.4 资本充足性与融资依赖
Basis 已分三步披露股权融资:$3.6 million 种子轮、$34 million Series A,以及以 $1.15 billion 估值完成的 $100 million Series B。自成立以来,披露股权融资约 $137.6 million;没有检索来源描述风险债、项目融资或其他杠杆。相对大多数垂直 AI 创业公司,这是一笔强资本基础。Series B 募资用途——加速平台开发并扩张工程和机器学习团队——也读起来像进攻性投入,而不是救援融资。 限制在于,融资不等于现金。Basis 没有披露现金余额、烧钱速度或收入贡献,因此现金跑道只能示意,不能确知。只用新到位的 $100 million 融资做一个简单情景,如果月烧钱约 $3.0 million 到 $5.5 million,意味着大约 18 到 33 个月现金跑道;如果此前资本仍有余额,或经常性收入抵消支出,实际现金跑道可能更长。即便如此,融资依赖并未消失。Basis 正在扩张一个安全重、依赖模型、企业部署型业务;下一次融资触发点很可能取决于采用头条能否转化为实际定价、扣除支持后的毛利率,以及在大型事务所内可重复扩张。[CI005, CI006, CI007, CI008, CI009, CI039]
| 项目 | 公开数值 / 状态 | 置信度 | 重要性 | 尽调问题 |
|---|---|---|---|---|
| 已披露股权融资 | 累计 $137.6M($3.6M 种子轮 + $34M Series A + $100M Series B) | 中 | 设定公司成立以来已披露外部资本的上限 | 对账扣除费用后的轮次实收款,以及任何老股交易部分 |
| 最新融资 | $100M Series B,估值 $1.15B(2026 年 2 月) | 高 | 新资本降低短期破产风险,并支持产品继续扩展 | 提供股权结构表、投资人权利和任何分期条件 |
| 账面现金 | 低 | 已融资金额不等于当前流动性 | 提供当前现金余额和月度现金流 | |
| 示例月度烧钱 | 每月 $3.0M–$5.5M(估计) | 低 | 需要用它把本轮融资换算成现金跑道;该估计基于情景,未披露 | 提供过去 12 个月月度运营费用和现金消耗 |
| 仅基于 Series B 的示例现金跑道 | 18–33 个月(估计) | 低 | 显示近期资本充足性对烧钱假设的敏感度 | 确认基准和下行情景运营计划下的现金跑道 |
| 债务 / 项目融资义务 | 未公开披露 | 低 | 没有可见杠杆降低复杂度,但不能证明没有义务 | 提供债务明细、租约和任何表外承诺 |
烧钱和现金跑道是仅基于已披露 Series B 的情景估计;当前现金、收入贡献和此前现金留存均未公开。
[CI005, CI006, CI007, CI009, CI039, CI040]只有融资情况已披露;烧钱速度和现金跑道是情景区间,不是公司报告数据。
烧钱速度和现金跑道按最新一轮融资规模及 Basis 披露的扩张姿态估算。它们仅作示意,不应视为已披露结果。
[CI009, CI039, CI040, CI041, CI042, CI046]4.5 财务结论与公开缺口
公开多头论点很直接:Basis 有真实的企业级验证点、快速融资速度、可信投资人、匹配买方要求的合规姿态,以及多项客户和伙伴来源指向节省劳动力的 ROI。随着会计劳动力市场保持紧张,专业服务中的 AI 采用从试验走向战略部署,公司看起来正卖进一个真实痛点。 公开空头论点也同样直接:收入、ARR、现金余额、毛利率、CAC、续约或集中度数据仍未披露。Journal of Accountancy 和 NIST 都强调了这件事为什么重要:幻觉、prompt injection、隐私和监督失败是会计流程中的现实风险,模型或供应商承诺可能跑在运营验证前面。因此诚实的财务结论是:品类需求和近期资本充足性偏正面,但收入质量和利润率耐久性仍无定论。严肃承销流程需要实际定价、软件与服务组合、留存和月度财务报表,之后才能把 Basis 视为财务风险已被化解,而不仅是商业前景可观。[CI021, CI025, CI026, CI027, CI034, CI038]
| 缺失私有指标 | 对分析的影响 | 精确尽调路径 |
|---|---|---|
| 当前 ARR / 收入运行率 | 没有营收规模,估值和效率无法基准化 | 要求提供月度收入桥和过去八个季度 ARR 历史 |
| 毛利率及软件 vs. 服务拆分 | 实施和支持负担仍不透明时,利润率耐久性无法判断 | 按软件、支持和专业服务提供 P&L |
| 现金余额和月度烧钱 | 仅靠融资标题无法承保现金跑道 | 提供当前现金、过去 12 个月烧钱和 18 个月运营计划 |
| 留存、扩张和集中度 | 没有队列数据和大客户敞口,无法判断收入质量 | 提供总留存率、NRR、队列扩张和前 10 大客户集中度 |
| 实际定价和折扣 | 无法把采用证明转化为经常性合同经济性 | 提供价格手册、ACV 分布、折扣审批和试点转生产数据 |
这些缺口是把 Basis 从叙事故事转为可承保对象所需的最低财务披露。
[CI021, CI034, CI038, CI041, CI043, CI044]4.6 图表
05产品与技术
5.1 用客户流程定义产品
Basis 用流程语言而不是助手语言描述产品。官网称代理在后台运行、端到端执行会计流程,并把完成的工作交回复核;关于页面强调,公司搭建的是会计逻辑系统,而不是叠在会计数据上的聊天机器人层。这个区别重要,因为 Basis 最强的公开证据不是泛化 AI 生产力话术;而是反复主张会计师仍在关键阶段掌控,系统完成执行密集步骤后,由人来复核工作产出并作判断。OpenAI 的 Basis 案例研究把这套营销落在具体例子上,例如对账、日记账分录和财务摘要;Series B 报道又加入端到端 1065 税表示例。Basis 还把产品映射到 CAS、税务、审计和咨询,因此客户流程故事比狭窄 AP 或记账楔子更宽。官网和 Wiss 的公开客户验证显示,公司确实有多业务采用;但公开记录在定性流程结果上仍强于按流程标准化的生产 KPI。[CE001, CE002, CE003, CE004, CE005, CE006]
| 模块 / 资产 | 主要用户 | 当前公开成熟度 | 差异化信号 | 尽调缺口 |
|---|---|---|---|---|
| CAS 智能体 | CAS 会计师和事务所复核人 | 公开界面中证明最充分 | 从核心会计切入,并有 Wiss 和官网证言的客户证明 | 需要按 CAS 任务提供工作流级准确率和人工覆盖指标 |
| 税务智能体 | 税务编制人和复核人 | 已演示但仍在扩张 | 1065 工作流证明,以及公开强调税务范围正在从核心会计之外扩展 | 需要确切税表覆盖、税务引擎集成和复核率数据 |
| 审计智能体 | 审计团队和项目复核人 | 早期公开扩张信号 | Basis 将审计作为全事务所范围的一部分来营销,Annette 称团队正在为审计判断复杂性而构建 | 需要真实审计生产范围、证据管理集成和 QA 指标的证据 |
| 咨询 / 全事务所层 | 合伙人、复核人和面向客户的会计师 | 定位层面的证据 | 官网把会计师描述为转向更高价值咨询,而智能体承担执行工作 | 需要明确工作流自动化与咨询分析功能之间的产品边界 |
| 信任 / 复核层 | 复核人、合规负责人和 IT / 安全团队 | 明确声称的控制界面 | 审计轨迹、理由、来源可见性、基于 token 的访问和认证都是其宣传话术的明确组成部分 | 需要 NDA 下的客户界面截图、管理员控制和安全审查输出 |
该表区分 Basis 明确营销的内容,以及仅由更宽定位暗示的内容。成熟度指公开证明质量,不是内部产品信心。
[CE001, CE005, CE006, CE007, CE008, CE010]| 用户任务 | 当前人工工作流 | Basis 方案 | 公开收益信号 | 当前限制 |
|---|---|---|---|---|
| 对账复核人 | 拉取支持材料、追踪交易、解决例外、起草对账 | 智能体完成对账步骤,并返回可复核输出 | OpenAI 将对账列为 Basis 自动化工作流 | 没有公开例外率或对账周期基准 |
| 财务总监或高级会计 | 研究支持材料、映射科目、起草日记账分录、解释逻辑 | 智能体准备日记账分录,并展示数据、映射逻辑和置信度 | OpenAI 描述的日记账分录会向会计师展示解释 | 没有公布错误率或过账批准指标 |
| 准备管理输出的事务所团队 | 汇编来源、总结活动、起草财务评论 | 智能体把财务摘要作为重复会计工作的一部分自动化 | OpenAI 将财务摘要列为重复工作流之一 | 没有按事务所类型或报告包公开模板覆盖 |
| 税务编制人 | 收集源文件、提取数值、起草申报包、复核 | Basis 公开演示端到端 1065 工作流,并称税务范围正在扩张 | Series B 报道和 Annette 访谈都支持税务工作流进展 | 没有公开支持表单、边缘案例或税务引擎集成清单 |
| CAS 团队负责人 | 跨客户协调月结工作和入职 | 客户证明显示,Basis 将执行自动化与入职和支持结合 | UHY 引述称月末关账时间减少,入职体验强 | 公开记录没有区分软件投入和服务投入 |
| 事务所创新负责人 | 比较 copilot 工具、助手和工作流系统 | Basis 把自己定位为会计原生端到端自动化,而不是通用助手 | Wiss 称 Copilot 和私有 ChatGPT 有用,但不等同于 Basis | 公开定价和部署经济性仍未披露 |
收益限于有来源支撑的信号,例如节省时间、减少关账工作量或改造工作流。该表避免在披露声明之外做无支撑 ROI 量化。
[CE002, CE003, CE009, CE010, CE012, CE013]公开描述中,Basis 在后台执行会计工作,并在决策点把可复核输出交给会计人员。
[CE002, CE003, CE009, CE010, CE012, CE021]5.2 架构与运营模式
最清晰的公开架构描述来自 OpenAI:每个 Basis 任务都从一个监督代理开始,由它根据任务、复杂度、延迟和输入类型把工作路由给专门子代理。在该描述中,监督角色从 o3 迁移到 GPT-5,而 GPT-4.1 仍留在技术栈中处理速度敏感交互。Basis 称这套架构由基准测试驱动:公司运行内部评估,以判断哪些模型适合哪些流程,以及代理何时能安全承接更多工作。同一来源还称,代理通过中心层共享上下文,并向复核者暴露假设、源材料和逻辑,这与 Basis 可追踪 AI 的公开承诺一致。另一批 Basis 工程帖补充了重要细节:公司似乎运行一种代理原生的内部运营模式,包含 APO、Atlas、重评测工程实践,以及通过统一 MCP 做内部工具编排。这些帖子是有价值的开发者信号,说明工程严肃度和内部流程成熟度;但它们不能直接证明每个面向客户的流程都使用完全相同的基础设施。公开来看,客户产品架构可信且具体;私下仍未披露精确服务栈、非 OpenAI 依赖和备用行为。[CE015, CE016, CE017, CE018, CE019, CE020]
| 层 / 流程 | 公开描述角色 | 关键依赖 | 主要风险 |
|---|---|---|---|
| 复核界面和工作包 | 为会计复核展示输出、来源、变更、理由和置信度 | 清晰的人工复核 UI 和底层数据来源 | 如果证据视图太薄,复核界面可能不足以支持审计级签署 |
| 事务所上下文和标准层 | 让输出适配事务所、业务、项目和既有偏好 | 持久结构化上下文和政策输入 | 上下文漂移或过期指令可能制造可重复错误 |
| 监督智能体 | 协调整个工作流,并在专门子智能体之间路由工作 | 前沿模型推理质量和编排逻辑 | 模型路由失败会沿长工作流传播 |
| 专门子智能体 | 按复杂度、延迟和输入类型处理聚焦步骤 | 模型基准和工具访问边界 | 覆盖缺口或工具误用可能制造隐藏失效模式 |
| 工具 / 连接器层 | 拉取执行所需的会计数据、文档和操作 | 身份认证、外部系统和集成覆盖 | 公开集成不透明,买方无法验证部署范围 |
| 控制、评测和内部运营层 | 内部 APO、Atlas 和评测实践似乎提升产品质量和迭代速度 | 人类管理者、基准和运营纪律 | 内部工程成熟度不等于面向客户的工作流证明 |
这是基于 Basis 和 OpenAI 材料合成的公开架构。它区分面向客户的工作流层,以及内部工程或运营模型信号。
[CE014, CE015, CE016, CE017, CE018, CE019]公开来源显示,Basis 可能采用分层架构,覆盖复核界面、事务所上下文、监督智能体编排、专用子智能体、工具和控制功能。
该图综合公开描述而成,并非供应商撰写的架构图。应把它视为尽调地图,而不是精确部署蓝图。
[CE015, CE016, CE017, CE018, CE019, CE020]5.3 部署、集成、可靠性与支持
Basis 拥有有意义的公开部署验证,但公开实施细节仍不完整。Wiss 称其在 2023 年末 beta 测试 Basis,成功后扩大到完整 CAS 部署;Baker Tilly 则把与 Basis 的关系定位为横跨数千名会计客户的规模化合作。官网补充了关于月结提速和强上线导入支持的具名证言,招聘页面称 Basis 已在头部会计事务所部署,并在每位会计师背后拥有数百个代理。安全材料支撑其对备份、事件响应、基于 token 的访问、加密和 MFA 的严肃态度。缺失细节恰恰在买方最需要的地方:经复核的公开记录没有发布完整生产集成列表、具体 ERP 或税务引擎连接器、可用性承诺、公开状态历史,或软件与实施服务之间按流程划分的服务边界。这种不透明不意味着集成或 SLA 不存在;它意味着公开尽调表面不完整。对投资人来说,这是典型的企业软件承销问题:客户验证可见,但部署可复现性和可靠性边界仍需直接尽调。[CE028, CE029, CE033, CE034, CE035, CE036]
| 维度 | 公开状态 | 已验证内容 | 未公开验证内容 | 含义 |
|---|---|---|---|---|
| 已点名部署 | 已验证 | Wiss 和 Baker Tilly 公开描述了与 Basis 的关系;官网包含多家点名事务所 | 按工作流、席位数和续约状态衡量的生产广度 | 真实部署证明存在,但规模经济性仍不透明 |
| 入职和支持 | 部分验证 | 首页客户证言称,上线和实施表现强 | 标准实施周期、人员配置模式和支持层级 | 支持可能是真正差异点,也可能带来服务交付负担 |
| 身份认证和访问 | 原则层面已验证 | 安全页面称访问使用可撤销令牌和 MFA | 各第三方系统的具体 SSO、API 或连接器模式 | 能看到重视安全的访问设计,但集成可复现性还看不清 |
| 数据驻留和连续性控制 | 原则层面已验证 | 安全和隐私页面称仅在美国托管,并提供加密、备份和事件响应 | RPO/RTO 数值、可用性目标和外部状态历史 | 控制姿态清楚,但企业级可靠性边界仍未公开 |
| 生态集成相关性 | 外部已验证 | Intuit 和 Xero 发布会计 API 与 OAuth 指南,支持生态连接 | Basis 面向这些或其他系统的专属连接器矩阵 | 即使 Basis 未公布准确清单,市场集成负担也是真实存在的 |
| 公开运营披露 | 未验证 | 已审阅来源均未列出完整 ERP、税务引擎或文档系统支持,也未给出公开 SLA | 具体生产集成和可靠性边界 | 这是公开记录中最主要的产品技术尽调阻碍 |
公开状态指证据表面,而不是判断 Basis 内部是否可能具备该能力。「未公开验证」的项目应转化为直接尽调请求。
[CE028, CE029, CE033, CE034, CE035, CE036]Basis 的执行质量取决于模型提供商、客户系统、结构化上下文、人工复核者和信任控制;连接器和 SLA 条款仍有多个公开披露缺口。
节点代表外部可见的依赖和控制点,并非完整供应商或基础设施清单。
[CE023, CE028, CE033, CE034, CE046, CE047]5.4 成熟度、路线图与范围扩张
Basis 的公开时间线支持这样一种产品图景:从核心会计起步,做出真实 CAS 牵引力,现在推进税务和审计,而不是声称所有业务领域从第一天起就已成熟。Annette Garcia 的访谈是这里最坦率的来源:她说 Basis 从核心会计开始,是因为系统必须先知道如何做会计,才能在税务或审计中成功;两个较新的业务领域都会引入性质上更难的判断和监管问题。这一表述与其余记录一致。Wiss 提供 CAS 部署的历史验证,OpenAI Basis 文章显示更深的流程编排和模型路由成熟度,Series B 报道则把公开范围扩展到 CAS、税务和审计,包括 1065 流程。与此同时,Basis 的 2025-2026 年工程帖显示,公司正在构建内部上下文、eval 和代理管理系统,应能提高迭代速度。正确解读是:Basis 已不再是早期原型,但也没有公开证明其营销的每个高风险流程都达到统一质量水平。成熟度故事在流程宽度和内部技术严肃性上最强,在公开质量指标和标准化企业运营数据上较弱。[CE007, CE008, CE010, CE011, CE020, CE036]
| 日期 / 阶段 | 功能或里程碑 | 状态 | 含义 | 来源 |
|---|---|---|---|---|
| 2023 基础阶段 | 核心会计 / CAS 起点 | 后续回顾已验证 | Basis 选择先学核心会计,再进入更难的税务和审计领域 | Annette Garcia 访谈 |
| 2023 年 11 月至 2024 年 9 月 | Wiss 先测试版试用,后全面部署 CAS | 已验证 | 说明产品至少在一家大型事务所里,从试点走到全业务线使用 | Wiss 官方发布 |
| 2025 年 8 月 | 可信智能体 / 模型路由架构公开讨论 | 已验证 | 公开技术细节增加,不再只是泛泛的 AI 营销 | Basis OpenAI 进展博客和 OpenAI Basis 文章 |
| 2025 年 8 月之后 | Atlas 和智能体原生的内部运营模式 | 内部运营已验证 | 暗示 Basis 正把上下文、评测和内部工具投入产品杠杆 | AGI 时代和工程文章 |
| 2026 年 2 月 | Series B 融资叙事把范围扩到 CAS、税务和审计,并突出 1065 自动化 | 已验证 | 在市场叙事中,Basis 不再只是 CAS 工具 | Business Wire 和 CPA Practice Advisor |
| 2026 年 6 月 | 税务和审计扩张被描述为下一前沿,但明确提到判断和监管复杂性 | 带保留地验证 | 支持后续路线图,同时也确认这些工作流比核心会计更难 | Annette Garcia 访谈 |
路线图状态仅限公开披露的里程碑,不能解读为每个营销中的业务领域都已达到同等成熟度。
[CE007, CE008, CE010, CE011, CE036, CE040]公开证据显示,Basis 目前在核心会计工作流执行和可复核性上最强;税务和审计广度明显推进,但证据还不充分。
强度标签基于公开证据质量,而非内部产品遥测或供应商提供的记分卡。
[CE007, CE008, CE010, CE020, CE036, CE038]5.5 差异化、信任、安全与合规
Basis 最可防守的公开差异化,不在于它是唯一把 AI 用在会计里的公司,而在于它把会计特定流程所有权、可复核性和多业务野心结合起来。Wiss 的比较尤其重要,因为它明确说 Copilot 和一家私有事务所 ChatGPT 实例有用,但仍不等同于 Basis 的集成式端到端自动化。这与竞争集合对应得很清楚:QuickBooks AI 和 Microsoft Copilot 是嵌入更大套件的宽助手;Botkeeper、Vic.ai 和 Digits 则分别营销更窄的记账、AP 或总账中心表面。相比之下,Basis 销售的是一个会计事务所原生平台,代理可以执行工作并交回可复核交付物。安全和信任姿态强化了这个故事:Basis 公开声称不使用客户数据训练模型、数据隔离、动作可审计、token-based 访问、仅美国数据托管和多项认证。但会计是受监管的信任业务,不是通用自动化市场。NIST、PCAOB、IRS、AICPA 和 Thomson Reuters 都强调,高风险金融 AI 需要治理、可复核性和专业判断。Basis 的信任叙事方向上强,但证明责任仍高,因为精确集成覆盖、SLA 条款和审计质量指标大多仍属私密。[CE012, CE013, CE024, CE025, CE026, CE027]
| 控制或要求 | 公开状态 | 范围 | 剩余缺口 | 重要性 |
|---|---|---|---|---|
| SOC 2 Type II | 已声称 | 安全性、可用性、保密性 | 报告仅在 NDA 下提供 | 企业买家会要求直接审阅报告 |
| ISO 27001 | 已声称 | 信息安全管理 | 认证边界未公开 | 支撑信任姿态,但不说明实施细节 |
| ISO 42001 | 已声称 | AI 管理体系 | 没有公开范围声明或控制摘要 | Basis 把 AI 卖进高风险工作流,因此该认证相关 |
| 不用客户数据训练模型 | 已声称 | 敏感客户和事务所数据 | 已审阅材料中没有公开企业隐私附录 | 这是会计事务所的核心信任问题 |
| 可追溯 AI 和审计日志 | 已声称 | 智能体动作、来源、变更、理由 | 没有公开截图或工作流级留存细节 | 可审阅性是会计场景采用的核心 |
| 仅美国托管和基于令牌的访问 | 已声称 | 驻留和身份认证模型 | 没有公开子处理方清单或逐连接器认证设计 | 对隐私审查和采购重要 |
| 专业人员复核要求 | 外部指南支持 | 审计、税务和会计判断仍由人承担责任 | Basis 未公开覆盖或升级率指标 | 监管和专业标准让人工监督不可选 |
| GDPR / CCPA 项目成熟度 | 部分声称 | 公开信任表面 | 安全页面仍写着即将推出 | 说明公开披露和合规叙事仍在成熟 |
本表把 Basis 的声称和外部治理背景放在一起,因为产品可信度既取决于供应商控制,也取决于约束会计工作的专业标准。
[CE024, CE025, CE026, CE027, CE028, CE029]5.6 图表
06客户
6.1 客户分层与买方图谱
Basis 卖给会计事务所,而不是以自助方式卖给个人记账员。公开信息反复围绕事务所业务领域组织产品——CAS、税务、审计、咨询、外包会计和托管服务;客户引用来自 CEO、合伙人、CIO 和托管服务负责人,而不是自己刷卡购买的一线员工。这个模式强烈指向事务所级买方和付款方,会计师及交付团队是主要用户,会计事务所的终端客户则通过更快结账、更好报告和更多咨询产能间接受益。分层宽度也不止单一记账楔子。Basis 从 CAS 起步,因为该领域流程重复且深度会计原生;但公开材料现在也把平台定位到税务和审计。实用的分层视角因此应按事务所类型和流程成熟度来划分:以 CAS 为主的外包会计事务所、更大型多学科事务所、托管服务业务,以及希望用一层 AI 横跨多个部门的跨业务事务所。[CU001, CU008, CU011, CU020, CU021, CU022]
| 分层 | 买方 / 用户 / 付款方 | 主要用例 | 公开规模 / 证据 | 战略价值 | 缺口 |
|---|---|---|---|---|---|
| CAS / 外包会计事务所 | 买方和付款方是事务所或业务线负责人;用户是 CAS 会计师和复核人员 | 日记账分录、对账、月结、报告、临近记账的工作流 | Wiss 从测试版到全面部署 CAS;UHY 月结引述;Basis 从 CAS 起步 | 证据最强的滩头阵地,因为工作流重复、高频且原生于会计 | 没有按 CAS 客户披露的席位数、ACV 或实施周期 |
| 大型多业务线会计事务所 | 买方和付款方似乎是事务所管理层;用户覆盖 CAS、税务、审计和咨询团队 | 跨业务线自动化和工作流委派 | Basis 称进入约 30% 的美国前 25 大事务所;首页引用的客户包括 Wiss、Boulay、Clark Nuber 和 BPB | 如果跨业务线使用属实,客户标识价值高,参考质量也强 | 公开证据未说明前 25 大关系中有多少是广泛部署,有多少是窄试点 |
| 托管服务型会计和财务业务 | 买方和付款方是托管服务负责人;用户是服务终端客户的交付团队 | 交易处理、JE 支持、对账、关账后分析、管理报告、预算和预测 | Baker Tilly 官方合作列名这些工作流,并称其先评估了多个平台 | 说明 Basis 能卖进服务交付运营,不只是后台实验 | 没有关于实施时间表、客户覆盖或上线经济性的公开信息 |
| 税务团队 | 买方可能是税务业务线负责人;用户是税务专业人士 | 税务底稿、纳税申报表和研究支持 | Basis 称平台现在覆盖税务;BPB 引述称可信智能体重塑税务工作;公司称已演示 1065 纳税申报表 | 如果准确性和可审阅性守得住,CAS 之外还有上行空间 | 没有公开的税务专项留存、胜率或生产量指标 |
| 审计团队 | 买方可能是审计负责人;用户是审计师和复核人员 | 审计测试、工作底稿、证据收集,支持可复核工作流 | Basis、媒体报道和公司材料都把审计列为目标业务领域 | 审计风险高、工作流丰富,战略价值大 | 公开证据弱于 CAS 和托管服务;未检索到具名审计客户案例 |
| 会计事务所终端客户受益方 | 付款方仍是会计事务所,但受益方是事务所客户 | 更快关账、更好报告、更多顾问产能 | UHY 提到客户月结时间缩短;Baker Tilly 把价值表述为一致性、可见性和可直接用于决策的洞察 | 客户结果很重要,因为它能强化事务所内部粘性和扩张 | 终端客户引用是间接的;留存来源没有披露直接客户满意度得分或案例研究分母 |
各行结合公司、客户和独立证据;公开证据对工作流类型最强,对花费、席位和续约行为最弱。
[CU001, CU008, CU010, CU011, CU013, CU020]Basis 通常从事务所领导层解决产能和质量瓶颈切入;团队信任可复核的工作流自动化后,再逐步扩张。
[CU008, CU009, CU020, CU022, CU023, CU031]6.2 具名账户与采用轨迹
公开客户证据最强的地方,是客户自己开口。Wiss 称,它从 2023 年 11 月开始在 beta 阶段测试 Basis,随后在 CAS 业务全面部署,内部采用率逐月上升,其他业务线也在排队跟进。Baker Tilly 称,它评估了多个平台后,选择 Basis 用于会计和财务托管服务,并点名计划自动化的工作流:交易处理、日记账分录支持、对账、关账后分析、管理报告、预算和预测。Basis 网站上 UHY 的引述给出更具体的结果证据:Basis 显著缩短了客户的月结时间,入职和实施支持也按承诺交付。跳出具名客户层面,Basis 和多次报道都称,该平台现在覆盖美国前 25 大会计师事务所中的约 30%。这是有意义的规模说法,但仍然只是渗透率标题,不是客户数、ARR 或使用率分母。[CU002, CU003, CU004, CU005, CU006, CU007]
| 指标 | 数值 | 日期 / 锚点 | 来源 | 置信度 | 含义 | 缺失分母 |
|---|---|---|---|---|---|---|
| 头部事务所渗透标题 | 约 30% 的美国前 25 大会计事务所 | 2026-02-24 报道窗口 | BusinessWire + CPA Practice Advisor + Our Accounting World 等公开资料 | 中 | Basis 已打入相当一部分大型事务所客户 | 没有公开列出具体前 25 大事务所、部署深度或收入贡献 |
| Wiss 初始证据 | 测试版从 CAS 开始 | 2024 年合作发布中引用 2023-11 启动 | Wiss 官方公告 | 中 | 最早的具名客户证据已经超出泛泛试点声称 | 未披露实施周期或转化经济性 |
| Wiss 生产证据 | 成功结果推动 CAS 全面部署 | 2024-09-16 公告 | Wiss 官方公告 | 中 | 确认至少一个具名客户从测试版进入生产工作流使用 | 未披露使用量、席位数或续约期限 |
| Wiss 内部扩张信号 | 内部采用逐月快速增加;其他业务领域也急于采用 | 2024-09-16 公告 | Wiss 官方公告 | 中 | 暗示初步成功后,单个事务所内存在落地再扩张潜力 | 没有量化用户数或跨业务线推广时间表 |
| UHY 运营结果 | 客户月结显著缩短 | 2026-06-11 访问 Basis 首页 | Basis 首页客户证言 | 中 | 公开材料中最清楚的客户结果陈述之一 | 没有基准期、缩短比例或样本量 |
| OpenAI ROI 代理指标 | 使用 Basis 的事务所平均节省 30% 时间 | 2026-06-11 访问 | OpenAI 案例研究 | 中 | 支持一个不依赖单一客户、可重复讲述的节省人工叙事 | 没有客户数、分层拆分或方法论 |
| 投资人效率声称 | 各业务领域效率提升 20% 到 50% | 2026-02-24 报道窗口 | BusinessWire + CPA Practice Advisor | 低 | 如果全所部署按描述跑通,存在上行空间 | 这是投资人引述,不是经审计运营数据 |
| 托管服务上线证据 | Baker Tilly 评估多个平台后选择 Basis,并列名多个工作流 | 2026-04-28 公告 | Baker Tilly 官方 | 中 | 大型会计和财务服务提供商采用的新近证据 | 没有公开规模指标说明平台上有多少客户或员工 |
本表混合公司声称、客户陈述和第三方复述;每一行都是采用代理指标,而不是已披露收入或活跃用户指标。
[CU002, CU005, CU006, CU009, CU012, CU017]| 客户 | 分层 | 部署 / 用例 | 生产还是试点 | 结果 / 公开信号 | 限制 |
|---|---|---|---|---|---|
| Wiss | 前 100 大 CPA / 外包会计和咨询事务所 | 2023 年 11 月启动 CAS 测试版,随后全面部署 CAS,并计划跨业务线推广 | CAS 已生产使用;更广泛推广仍在扩张 | 全面部署、内部快速采用、更快月结 / 更高效率声称、多位高管引述 | 没有席位数、续约期限或量化 ROI 分母 |
| Baker Tilly | 大型会计和财务托管服务提供商 | 交易处理、日记账分录支持、对账、关账后分析、管理报告、预算、预测 | 供应商评估后宣布生产意向;当前上线规模未披露 | 客户自有公告称,评估多个平台后选择 Basis | 暂无公开证据说明部署广度、客户数或留存 |
| UHY | CAS / 外包会计业务 | 在 UHY 全所实施,面向客户月结用例 | 基于客户引述看似已生产使用;启动日期未披露 | 引述称关账时间显著缩短,上线和支持也强 | 证据位于 Basis 自有页面,而非 UHY 自有的 Basis 公告 |
| Clark Nuber | 聚焦 CAS 的会计事务所参考客户 | CAS 部门用例和工作流精简 | 生产使用口吻的客户证言 | 合伙人称团队不断发现新用例,Basis 是力量倍增器 | 未检索到独立客户自有文章 |
| Boulay | 跨业务线会计事务所参考客户 | 全所税务、审计、咨询及更多业务 | 生产使用口吻的客户证言 | CIO 称 Basis 服务整个事务所,并扮演 AI 合作伙伴 | 未检索到部署指标、时间或独立客户自有案例研究 |
列举范围覆盖本轮检索到、公开证据最具体的具名参考;没有结果细节的客户标识已排除。
[CU004, CU005, CU006, CU007, CU009, CU010]证据在供应商选择和初始生产使用阶段最强;到了留存、ARR 集中度和队列扩张阶段,就明显变弱。
[CU005, CU009, CU017, CU018, CU026, CU031]6.3 ROI、部署摩擦与证据质量
ROI 叙事可信,但仍大量依赖替代指标。OpenAI 称,使用 Basis 的事务所平均报告节省 30% 时间,并且随着信任加深持续扩大智能体职责;Basis 融资报道还加入了一位投资人的说法,称各业务线效率提升 20% 到 50%。问题不在于没有 ROI 信号——信号有好几个——而在于最强的公开数字仍来自公司渠道或合作伙伴渠道,而不是经审计的客户经营数据。独立实施研究解释了原因。CPA Practice Advisor、Thomson Reuters、ACCA、Journal of Accountancy 和 IBM 都描述了同一组采用瓶颈:数据隐私担忧、系统割裂、培训和技能缺口、治理要求、人工复核义务,以及对黑箱 AI 输出的怀疑。Basis 自己的产品运营访谈也从内部说明同一点:会计师要的是一致性、审计轨迹、可预测性,以及后续复核时站得住的输出。因此,对现有证据最好的解读是:在合适的事务所里已经有真实生产价值,但部署仍需要大量控制和变更管理。[CU017, CU018, CU019, CU023, CU024, CU025]
| 摩擦 / 控制领域 | 公开证据 | 对客户为何重要 | 当前判断 | 尽调要求 |
|---|---|---|---|---|
| 隐私和客户数据处理 | Basis 称不训练模型、完整数据隔离,并提供单独企业协议;Thomson Reuters 和 ACCA 称隐私是头号关切 | 会计事务所处理高度敏感的客户数据,如果使用条款不清楚,会阻止部署 | 控制姿态方向上不错,但买家仍需做合同和运营验证 | 要求提供 DPA / MSA 隐私条款、模型提供商清单和客户数据流图 |
| 可审计性和人工复核 | Basis 强调动作可追溯、理由透明;Annette Garcia 称会计师需要一致性和可审阅性 | 客户需要输出经得起后续复核,也符合专业标准 | 这看起来既是核心产品强项,也是核心采购要求 | 要求提供审计轨迹样本、复核人看板,以及异常处理工作流。 |
| 集成复杂度 | Thomson Reuters 和 IBM 称,系统割裂和工作流集成是主要障碍。 | 多系统会计环境可能拉长见效时间,并推高服务成本。 | 即便客户愿意购买,这也很可能是真实的部署负担。 | 要求提供实施计划、系统前置条件和平均上线周期。 |
| 培训和变更管理 | CPA Practice Advisor、Thomson Reuters、ACCA 和 IBM 都强调技能缺口和人员提升。 | 培训薄弱会拖慢采用,也限制账户内扩张。 | 这很可能是持续工作,不是一次性上线步骤。 | 要求提供客户赋能计划、管理员培训小时数和采用手册。 |
| 客户信任和预期 | JOA、ACCA 和 Thomson Reuters 都警示幻觉、黑箱信任和披露预期。 | 即便事务所购买产品,面向客户的使用仍可能受信任和专业风险约束。 | 平台看起来围绕这一反对意见设计,但公开证据仍主要停留在叙述层面。 | 要求提供客户披露措辞、QA 控制,以及被拒绝或升级输出的案例。 |
各行把产品主张与独立反向证据放在一起;这是一张部署风险表,不是失败记录。
[CU023, CU024, CU025, CU026, CU027, CU028]Wiss 拥有最丰富的客户自有证据,Baker Tilly 给出最新的规模化采用信号,UHY 的公开运营结果最清晰;其余多是较薄的证言证据。
[CU002, CU005, CU007, CU009, CU012, CU014]6.4 耐久性、扩张与集中度
Basis 最清晰的扩张路径不是消费级座席增长,而是在已经信任它的事务所内部,继续打深工作流和业务线渗透。Wiss 从 CAS beta 走到全面部署,并出现更广泛的内部兴趣。Baker Tilly 把此次上线表述为托管服务扩张的下一阶段。Basis 自己的会计产品运营团队称,公司从 CAS 起步,因为必须先吃透核心会计,再延伸到审计和税务。这些都是实打实的落地后扩张信号。尚未被证明的是投资人能够承保的耐久性。公开来源没有披露 GRR、NRR、流失、续约率、合同期限或队列留存。集中度图景同样有两面:覆盖前 25 大事务所中约 30% 是很强的 logo 质量信号,但也可能意味着相当一部分商业价值集中在少数大客户手里。没有收入集中度、按 logo 拆分的部署深度或合同到期数据,客户故事在战略价值上有吸引力,但经济耐久性的文件仍不够。[CU005, CU006, CU009, CU018, CU020, CU022]
| 指标 | 数值 | 分层 | 置信度 | 尽调要求 |
|---|---|---|---|---|
| 总收入留存率(GRR) | null | 全部客户 | 低 | 按客户队列和业务领域提供 2024、2025 以及过去 12 个月 GRR |
| 净收入留存率(NRR) | null | 全部客户 | 低 | 提供 NRR,并拆分席位增长、工作流扩张和价格驱动 |
| 流失 / 续约率 | null | 全部客户 | 低 | 按客户层级提供客户流失、收入流失和续约率 |
| 重复使用代理指标 | Wiss 内部采用逐月上升;随着信任增长,事务所扩展职责范围 | 具名企业账户 | 中 | 展示头部账户的月活用户、任务量和工作流扩张曲线 |
| 满意度 / 实施质量代理指标 | UHY 称上线和支持按承诺交付;首页客户证言非常正面 | 具名参考账户 | 中 | 提供客户参考访谈、CSAT/NPS,以及实施到价值实现时间指标 |
关于耐久性的公开证据大多是轶事;null 表示留存公开来源中未找到该指标,并不代表表现弱。
[CU006, CU012, CU018, CU036, CU037, CU047]| 扩张驱动因素 | 集中度 / 持久性风险 | 影响 | 公开证据 | 尽调路径 |
|---|---|---|---|---|
| 从 CAS 扩到税务和审计 | 后期业务线采用速度可能慢于 CAS。 | 如果 Basis 能在事务所内部标准化,上行空间很大。 | Basis 首页、关于页和产品运营访谈都指向跨业务线野心。 | 要求按业务领域拆分 ARR 和活跃工作流。 |
| 单一客户内的工作流宽度 | 如果客户只买一个工作流而不买其他工作流,使用可能仍很窄。 | 即便初始大客户标识很强,也可能限制 NRR。 | Wiss 和 Baker Tilly 暗示存在多条工作流,但公开材料对广度的量化仍很薄。 | 要求提供头部客户逐工作流采用历史。 |
| Top-25 事务所渗透 | 大型事务所标识占比高,也可能意味着收入集中。 | 少数大账户可能主导收入和续约风险。 | Series B 材料反复提到前 25 大事务所中约 30% 使用。 | 要求按 ARR 和任务量披露前 10 大客户集中度。 |
| 高质量背书集中 | 最强公开证据集中在 Wiss、Baker Tilly 和 UHY。 | 少数标识之外证据偏弱,可能夸大覆盖广度。 | 首页推荐语集合比客户自有证据集合更宽。 | 要求提供更多客户自有案例研究和背调电话。 |
| 留存不透明 | 没有公开 GRR、NRR、流失或合同期限指标。 | 仅靠公开证据无法承销持久性。 | 公司、客户和行业媒体来源中都一致缺少这些指标。 | 要求提供队列留存、续约日历,以及扩张 / 降级桥表。 |
本表把增长机制和仍缺失的证据拆开;这些风险是商业未知项,不是采用失败的证据。
[CU020, CU022, CU036, CU038, CU042, CU043]Basis 似乎从 CAS 滩头扩张到相邻业务领域;但大型事务所 logo 质量既支撑上行,也因流失和续约数据未披露而放大集中度风险的重要性。
[CU005, CU006, CU009, CU020, CU021, CU022]6.5 客户尽调缺口
买方可以为“Basis 已有真实企业采用、具名生产参考和符合品类逻辑的 ROI”这句话辩护。但买方还不能为强留存、低集中度或部署风险已充分消除辩护。公开记录仍缺客户数分母、按客户层级拆分的 ARR、留存队列、合同期限、续约时点、实施到价值实现的时间,以及前 25 大事务所与长尾客户之间的客户组合。即便具名客户,证据也不均衡:Wiss、Baker Tilly 和 UHY 有实质性证明,而 Clark Nuber、Boulay 和 BPB 大多停留在推荐语层面。因此,下一步尽调不该再重点问客户是否存在,而应追问这些客户有多耐久、经济意义有多大。应要求 Basis 提供带 ARR 区间的客户名单、前十大客户集中度表、续约排期、实施时间线、按业务线划分的扩张历史,以及包含分母定义和时间窗口的客户级 ROI 研究。[CU036, CU037, CU038, CU043, CU044, CU047]
6.6 图表
07风险
7.1 风险排序总览,以及为什么承保仍被卡住
Basis 周围最高严重度的风险不是泛泛的“AI 风险”,而是 AI 输出可能出错,与承担法定、标准化和声誉责任的专业服务工作之间的相互作用。Basis 现在把一个平台卖进 CAS、税务、审计和咨询,这意味着单个模型、控制或工作流弱点可能跨多个受监管会计场景扩散,而不是被限制在低风险后台任务里。公开缓释因素真实存在。Basis 称其通过 SOC 2 Type II、ISO 27001 和 ISO 42001 认证,数据留在美国,不用客户数据训练,并记录智能体动作。法律条款也异常明确:AI 输出不能替代专业判断,在用于申报、客户交付物或专业工作成果之前必须复核。这些控制重要,但没有回答承保人更难的问题:税务和审计工作流中观测到的幻觉率或异常率是多少,企业合同如何分配责任,客户基础有多集中,如果关键模型提供商、安全认证或集成路径突然变化会发生什么。[CR001, CR002, CR003, CR005, CR006, CR007]
按发生可能性、影响、缓释成熟度和剩余严重性,序数化呈现 Basis 当前最重要的风险领域。
评级是基于已审阅 2026 年来源集综合出的承保判断,不是概率预测。
[CR023, CR027, CR034, CR036, CR042, CR045]输出质量、监管和合同责任分配风险如何传导为客户流失和承保失败。
[CR006, CR007, CR017, CR020, CR041, CR042]7.2 专业准则、CPA 责任与监管风险是一阶问题
Basis 明确卖进了贴近 GAAP 解读、纳税申报处理、审计证据和 SOX 控制流程的工作。这很关键,因为治理准则把判断责任放在持牌事务所和项目团队身上,而不是工具供应商身上。Basis 自己的条款写明,输出仅供信息和辅助使用,可能包含错误或遗漏,不能替代会计、税务、审计、法律或财务建议。Journal of Accountancy 和 CPAI 材料从买方侧强化了同一点:生成式 AI 可能误用准则、产生幻觉,或不当处理机密客户数据;盲目信赖会把工作流便利变成疏忽或专业责任问题。税务侧尤其敏感。IRS Publication 4557 称保护纳税人数据是法律要求,并提示事务所承担 FTC Safeguards Rule 义务;26 U.S.C. §7216 则将某些未经授权披露或使用申报信息的行为刑事化。审计侧,PCAOB AS 1105 要求充分、适当的审计证据,并特别看重证据来源的可靠性和独立性,因此任何由 AI 摘要或 AI 转换过的证据链都是尽调项。Basis 可以支持工作流,但公开记录尚未给出足够量化的控制证据,不能假设这些责任已经在经济或运营层面从事务所转移出去。[CR006, CR007, CR008, CR010, CR011, CR012]
| 风险 | 当前公开证据 | 可能性 | 严重性 | 缓释成熟度 | 残余暴露 | 尽调路径 |
|---|---|---|---|---|---|---|
| AI 输出执业过失 / 疏忽暴露 | Basis 称输出不是专业建议,必须经过复核;CPA 指引称盲目信赖可能产生责任。 | 高 | 关键 | 中 | 如果税务、审计或咨询工作出错,事务所仍承担复核责任。 | 要求提供业务约定条款样本、QA 工作流控制、错误升级阈值和索赔历史。 |
| 纳税申报保密 / IRC 7216 误用 | IRS 和 26 U.S.C. §7216 约束申报信息的披露或使用;CPA 指引指出,涉及 AI 时需要分析同意要求。 | 中高 | 关键 | 中 | 未经授权使用数据可能引发法律、声誉和客户流失后果。 | 审查数据流图、同意措辞、留存做法和税务工作流隔离控制。 |
| PCAOB / 审计证据不足 | PCAOB AS 1105 强调充分、适当证据和独立可靠性;如果 AI 在缺少控制的情况下转换证据,这两点可能被削弱。 | 中 | 高 | 低中 | 审计客户仍需要证明 AI 输出不会替代证据测试。 | 要求提供审计工作流设计文档、证据血缘、覆盖日志和审计师接受案例。 |
| 网络安全 / 隐私治理缺口 | Basis 宣传认证和保障措施,但企业隐私条款和详细合规附件未公开,GDPR/CCPA 仍标为即将推出。 | 中 | 高 | 中 | 信任姿态可见,但法律和运营完整性仍难从公开材料验证。 | 要求提供最新 SOC 2 报告、ISO 范围声明、DPA、子处理方名单和事件预案。 |
| SOX / GAAP / 披露控制误用 | 面向合规的现有厂商把 AI 用于披露、风险因素和 SOX 准备度营销,说明这些工作流对控制失效高度敏感。 | 中 | 高 | 低中 | 如果 Basis 用在财务报告上游,错误输出可能污染控制叙述或披露。 | 要求提供客户批准的控制矩阵,以及在结账、报告或 SOX 场景中有边界使用的案例。 |
各行按残余承销严重性排序;公开证据对规则存在性最强,对 Basis 专属合同分配最弱。
[CR006, CR007, CR008, CR014, CR015, CR016]7.3 安全、保密与治理是可见强项,但承保仍嫌不完整
Basis 的公开安全姿态强于许多私营 AI 初创公司。公司声称拥有 SOC 2 Type II、ISO 27001 和 ISO 42001 认证;不使用客户数据训练模型;严格租户隔离;仅在美国托管;MFA;静态和传输中加密;智能体动作可审计。对处理纳税申报、审计工作底稿和其他敏感客户数据的会计师而言,这些都是有意义的缓释因素。但同一套公开材料也说明,风险章节不能简单写成“安全已解决”。Basis 的安全页面仍写着 GDPR 和 CCPA “即将推出”,公开隐私政策称企业产品隐私条款存在于单独书面协议中,SOC 2 报告和其他合规文件也只在 NDA 或请求后提供。NIST 的 AI RMF、NIST 更广泛的 AI 治理工作和 SP 800-53 都指向同一个方向:风险管理是一套持续控制纪律,不是一次性徽章。因此,承保问题不是缺少安全主张,而是缺少公开证据来说明这些控制在真实客户工作负载下如何表现,模型供应商和子处理方如何治理,以及生产环境中可度量的残余错误、事件或覆盖率长什么样。[CR003, CR004, CR005, CR009, CR017, CR018]
| 失效模式 | 重要性 | 可能性 | 严重性 | 当前缓释 | 残余暴露 | 未解决缺口 |
|---|---|---|---|---|---|---|
| 幻觉或不完整的会计输出 | 税务、审计和 GAAP 任务高度依赖上下文,一个看似合理的答案也可能变成专业责任事件。 | 高 | 关键 | 人工复核、明确的输出免责声明,以及带引用的工作流步骤。 | 公开材料仍看不到错误逃逸率。 | 没有按工作流披露任务准确率、异常率或返工的公开基准。 |
| 机密客户或纳税人数据处理不当 | CPA 事务所和税务申报服务商处理高度敏感的财务和个人信息。 | 中高 | 关键 | 不用于训练声明、仅美国托管、MFA、加密和租户隔离。 | 客户仍必须验证数据流、留存和同意控制。 | 没有公开的企业 DPA 或工作流级留存披露。 |
| 认证之下的安全控制漂移 | 认证证明流程成熟度,但承销人更关心控制运行、范围和事件历史。 | 中 | 高 | SOC 2 Type II、ISO 27001、ISO 42001、备份和事件响应声明。 | 公开材料仍停留在摘要层面,且部分受 NDA 限制。 | SOC 2 报告、ISO 适用性声明和近期渗透测试摘要未公开。 |
| 提示注入 / 影子 AI / 用户绕行 | 客户文件和员工行为可能引入隐藏指令,或绕过获批工作流。 | 中 | 高 | 审计日志和基于角色的访问控制有帮助,专业指引也建议采用限制性政策。 | 人为错误仍是活的控制弱点。 | 没有沙盒、内容过滤或工作流隔离指标的公开证据。 |
| 规模扩张中认证或隐私承诺缺口扩大 | 进入更多事务所和更多工作流,可能跑在隐私、区域或行业控制承诺前面。 | 中 | 中高 | 可见的信任定位和标准对齐。 | 公开信号仍显示隐私 / 合规里程碑尚未完成。 | GDPR/CCPA 时间表和扩大的监管 / 客户要求仍不清楚。 |
| 声明或营销跑在实测表现前面 | FTC AI 执法显示,证据薄弱时,监管机构愿意挑战 AI 赋能主张。 | 中 | 高 | Basis 条款谨慎,把复核责任推回给用户。 | 营销、演示和客户轶事仍可能跑在量化证据前面。 | 需要按工作流披露经审计或客户验证的表现报告。 |
本登记表关注生产质量和保密风险,而不是泛泛的创业公司运营风险。
[CR003, CR004, CR005, CR010, CR011, CR012]7.4 基础模型依赖、既有厂商竞争与集中度都会压缩上行空间
Basis 不是在真空中竞争。公开市场和产品证据显示,AI 加持的财务与 ERP 市场增长很快,但赛场上也有大得多的供应商,已经把可信内容、披露工作流、税务引擎和嵌入式分发打包在一起。Workiva 推广用于财务报告、审计、风险和 SOX 准备的 AI;Thomson Reuters 推广带引用支撑的 CoCounsel Audit 和 CoCounsel Tax,以及注入 AI 的 ONESOURCE 工作流;Intuit 正把 Intuit Assist 推向 QuickBooks 和报税准备。Workiva 自己的 10-K 称,竞争横跨多元化企业软件提供商、垂直供应商、GRC 供应商和专业服务替代方案,并预计更大型和更专门化的供应商都会进一步加剧竞争。Basis 的客户证据令人印象深刻,但公开范围仍窄,集中在少数具名事务所和合作关系,而不是披露的广泛客户基础、留存画像或头部客户敞口。模型依赖也只露出一部分。Workiva 公开表示,其 AI 技术栈目前使用来自 Google、AWS/Anthropic 和 Microsoft/OpenAI 的前沿模型;Basis 的公开界面强调 OpenAI/Codex 和“可信智能体”叙事,但主安全 / 法律页面没有公开写清生产模型冗余或回退架构。因此,在既有厂商持续缩小产品差距的同时,Basis 也暴露在平台、模型成本或合作伙伴政策变化之下。[CR027, CR028, CR029, CR030, CR031, CR032]
| 依赖项 | 交易对手 / 类别 | 角色 | 集中度信号 | 失效情景 | 严重性 | 缓释 | 残余暴露 |
|---|---|---|---|---|---|---|---|
| 前沿模型供应商 | OpenAI / Anthropic / 超大规模云厂商级 LLM 生态 | 企业 AI 工作流的核心推理和生成层 | Basis 公开叙事提到 OpenAI 进展,而同业公开引用多家前沿模型供应商。 | 模型定价、政策、延迟或访问变化削弱产品经济性或质量。 | 高 | 人工复核、可审计性,以及潜在多模型架构。 | 公开材料未披露冗余架构。 |
| 云 / 安全基础设施 | 托管、身份和备份栈 | 存储并保护客户数据和工作流日志。 | 仅美国托管和认证可见,但基础设施架构未公开。 | 宕机、泄露或控制范围错配会伤害客户信任和续约。 | 高 | 加密、MFA、备份、事件响应和认证。 | 详细架构和近期事件历史仍为私密信息。 |
| 会计事务所渠道集中 | 具名大型事务所客户和合作关系 | 需求证明、背书和扩张渠道 | 公开证据集中在少数具名事务所,而不是披露出的广泛基盘。 | 失去一个标杆客户或合作伙伴,会损害可信度和增长叙事。 | 高 | 多业务线产品宽度和更多事务所背书。 | 公开材料未披露集中度和留存数据。 |
| 企业合同条款 | 客户 MSA、订单、DPA 和保险结构 | 分配责任、数据权利、复核义务和审计权。 | 公开网站条款把平台细节交给私下书面协议。 | 赔偿或责任条款弱于买方对受监管工作流的预期。 | 高 | 私下尽调中可用定制合同解决。 | 看不到合同前,承销不能假设分配有利。 |
| NDA 下的合规文档 | SOC 2 报告和其他合规文档 | 买方尽调加速器和续约支持 | 公开页面称这些材料只在 NDA 下或按请求提供。 | 如果文档不完整或范围过窄,销售周期会变慢,尽调也可能失败。 | 中高 | 信任导向定位和可见认证。 | 范围、例外和整改历史未公开。 |
依赖项按其直接损害客户信任、工作流质量或企业成交率的程度排序。
[CR005, CR009, CR032, CR033, CR036, CR038]塑造 Basis 剩余风险的主要外部依赖和竞争压力。
[CR001, CR032, CR033, CR034, CR035, CR036]7.5 承保卡点、缓释成熟度与投资逻辑击穿触发器
公开证据并不说明 Basis 不能投;它说明若要给出干净的风险折扣,材料还不够。公开记录支持真实的缓释因素——安全认证、不训练客户数据的主张、审计日志、人工复核要求、大型事务所客户证据,以及足够重要的产品覆盖面。但公开记录也留下多个未解卡点。没有公开证据显示税务和审计任务中的实测生产准确率、错误严重度、异常处理或返工;没有公开客户集中度或留存数据;没有公开 DPA、安全附件或赔偿排期来说明 AI 生成工作流导致错误申报或客户损失时责任落在哪里;也没有清晰的公开图谱说明模型提供商集中度或韧性。这些不是装饰性尽调问题。如果 Basis 不能证明低逃逸率、强覆盖控制、私有合同风险分配和多元化模型治理,那么正确做法就是把公司按高风险赋能型供应商承保,而不是按已经验证、关键任务级的会计基础设施承保。可度量的否决触发器很直接:重大安全事件、负面监管结论、流失标杆客户、无法维持认证,或无法证明税务和审计工作流的输出质量。[CR037, CR038, CR040, CR041, CR042, CR044]
| 角色 / 职能 | 公开依赖或缺口 | 可能性 | 严重性 | 缓释 | 尽调路径 |
|---|---|---|---|---|---|
| 负责任 AI / 质量治理领导力 | 公开材料更强调产品和信任结果,而不是具名的模型风险或保证负责人。 | 中 | 高 | ISO 42001 和审计日志主张暗示内部存在某种治理结构。 | 要求提供组织架构图、模型风险委员会节奏,以及高严重性输出失败的升级负责人。 |
| 客户成功和实施能力 | 横跨税务、审计、CAS 和咨询扩张,会提高上线和支持复杂度。 | 中高 | 高 | 具名大型事务所证据表明 Basis 能服务复杂客户。 | 要求按队列提供实施人员配置、支持 SLA 和续约驱动因素。 |
| 熟悉准则的主题专家深度 | 专业工作流需要深懂税务、审计和报告规则,不能只靠通用 AI 工程。 | 中 | 高 | 产品条款要求人工复核,Basis 也公开雇用面向会计的产品人员。 | 审查领域专家人数、复核人资质,以及各工作流验证的负责人。 |
| GTM 纪律与炒作 | 市场快速增长可能奖励激进 AI 定位,即使控制尚未充分证明。 | 中 | 中高 | Basis 公开法律条款谨慎,减少了明显过度声明。 | 要求提供营销主张、案例研究和基准发布的内部审批规则。 |
公开执行风险更多来自控制负责人深度和服务能力,而不只是创始人能见度。
[CR001, CR014, CR031, CR034, CR037, CR039]| 风险 | 可监控触发项 | 阈值 / 事件 | 行动含义 |
|---|---|---|---|
| 受监管工作流输出质量失败 | 经验证的错误或异常指标 | 税务或审计工作流逃逸率超过约定阈值,或造成重大客户补救事件。 | 暂停扩张到该工作流,直到控制、复核和基准证据改善。 |
| 安全 / 隐私恶化 | 认证、事件披露或客户通知 | 重大事件、失去认证,或信任材料范围受限。 | 上调风险评级,并要求更新控制包后再推进。 |
| 模型供应商集中冲击 | 供应商或政策变化 | 重大模型访问、定价、安全政策或延迟变化,且没有证明可行的后备路径。 | 套用依赖折价,并要求冗余计划。 |
| 客户集中度破裂 | 具名标识流失或合作关系逆转 | 失去标杆事务所、续约失败,或背书集合缩小。 | 重新评估增长持久性和商业可信度。 |
| 监管 / 准则压力上升 | IRS、PCAOB、SEC、FTC 或州隐私行动 | 新规则、指引或执法重大改变允许的工作流设计或披露预期。 | 延后承销,直到客户和供应商控制更新。 |
| 尽调包仍不完整 | 私下证据请求完成度 | 尽调周期后仍没有量化准确率、合同分配、集中度或模型治理证据。 | 将其视为继续研究 / 不承销结果,而不是强行给出干净通过。 |
触发项这样设计,是为了让投资人或买方不用依赖模糊情绪变化也能监控。
[CR017, CR020, CR023, CR027, CR036, CR037]7.6 图表
08估值
8.1 对价格敏感的结论与入场纪律
对一家年轻的垂直 AI 公司来说,Basis 已经越过最重要的标题门槛:2026 年 2 月的 Series B 将公司估值推到约 $1.15B,背后有蓝筹投资方,也有真实会计师事务所渗透。这很重要。相比泛泛的“AI for finance”营销,顶线故事更强,因为证据包里有前 25 大事务所渗透、Baker Tilly 类渠道合作,以及一份指向可度量省时的 OpenAI 案例研究。但这些还不足以把当前估值变成自动通过的承保结论。公开记录仍未披露 ARR、收入、留存或烧钱速度,所以投资人实际上是在为一个看起来真实、但经济性仍不透明的品类领导地位付费。Intuit、Workiva、BlackLine 和 BILL 的公开可比倍数大致聚集在约 2x 到 4x 销售额,2026 年 SaaS 市场环境也明显不如 2021 年宽松。因此,建议取决于价格,而不是绝对分类:Basis 具有战略吸引力,商业上也说得通,但在当前估值下,纪律性结论是继续研究,中等信心、高风险、估值偏高;而不是先买入、后尽调。[CV001, CV002, CV003, CV006, CV009, CV010]
| 建议 | 置信度 | 风险评级 | 估值立场 | 决策含义 |
|---|---|---|---|---|
| 继续研究 | 中 | 高 | 偏高 | 只有在完成 ARR、留存、利润率和优先股堆叠尽调之后,或以显著更低入场价,才承销 2026 年 2 月估值。 |
该建议刻意对价格敏感:公司质量不等于估值支撑。
[CV002, CV003, CV034, CV035, CV054, CV062]赞助方质量和商业验证把建议往上推,但缺少分母这一关仍然过不了。
这是一张逻辑图,不是概率模型。即便赞助方和客户信号很强,结论仍停在继续研究,原因在这里。
[CV002, CV004, CV006, CV009, CV010, CV034]头部 KPI 足够有吸引力,值得继续做功课;但信息仍太不完整,无法在当前估值标记下干净放行。
[CV002, CV003, CV006, CV030, CV033, CV034]8.2 当前融资背景、资本效率与稀释
估值本身足够真实;问题在分母。Basis 融资 $100M,估值约 $1.15B,目前披露累计融资约 $138M。纸面上看,这显得资本效率不错:累计融资仅约为当前标题估值的八分之一。若把所报估值视为投后估值,最新一轮的新钱稀释也只是约高个位数。这是好消息。更难的问题是,表面效率究竟反映真实经济杠杆,还是投资人在公司披露收入基础前,就提前为品类领导地位定价。公开来源没有显示风险债、老股交易、清算优先权,也没有说明普通股入场比独角兽标题暗示的更干净还是更复杂。实用的估值结论是,Basis 大概降低了短期融资风险,但没有消除稀释风险。如果公司后来在经济性得到广泛验证前还需要再融资,股权结构表和优先股堆叠里的隐藏变量,可能比当前投后估值更影响回报结果。[CV001, CV003, CV004, CV005, CV008, CV052]
| 视角 | 当前判断 | 什么会改变判断 |
|---|---|---|
| 品类需求 | 正面:会计自动化需求和劳动力压力是真实的,Basis 也有投资人质量和头部事务所证据。 | 如果部署被证明狭窄或集中,而不是能跨事务所和工作流复制,判断会转弱。 |
| 商业证明 | 建设性:Baker Tilly、OpenAI 和 Top-25 渗透表明市场拉力真实存在。 | 如果 Basis 披露续约、扩张和实施标准化数据,判断会明显改善。 |
| 估值支撑 | 薄弱:本轮融资真实存在,但公开经济性不存在。 | 如果 ARR 已经高于大约 $75M-$100M,且利润率接近软件公司,判断会改善。 |
| 资本结构 | 混合:最新一轮降低了近期融资风险,但优先权和稀释条款仍不公开。 | 完成股权结构表、清算优先权和任何老股悬垂尽调后,判断会改善。 |
各行把业务质量支撑和价格支撑拆开,避免章节在只想表达产品品类有吸引力时,意外变成推荐公司。
[CV003, CV004, CV006, CV008, CV009, CV010]8.3 可比框架,以及为什么传统倍数承保受限
Basis 不完全落在某一个纯可比桶里,所以估值纪律才重要。私营市场上,Numeric 和 Truewind 显示投资人仍在积极资助会计自动化,但这些公司更常披露轮次规模和产品雄心,而不是估值。Ramp 距离更远,但有参考价值,因为它说明,只要收入规模可见,2026 年私营市场对 AI 驱动的 fintech 平台仍可能胃口巨大:$44B 私有估值之下有超过 $1B 的年化收入。公开市场侧,可比信号更冷静。Yahoo Finance 在运行日显示,Intuit 约 3.8x 销售额、Workiva 约 3.0x、BlackLine 约 2.6x、BILL 约 2.1x。监管文件进一步证明,这些是真正的软件企业,收入经审计,毛利率也可观。Aventis 更广泛的 2026 年软件研究方向类似:公开市场 EV / 收入中位数约 3.4x,私营 SaaS 并购约 3.1x。Basis 可能配得上高于该组公司的溢价,因为它可能更早期、增长更快,也在塑造品类;但公开证据还不足以假设溢价宽到估值支撑不再重要。换句话说,今天还没有足够披露的收入、利润率或烧钱数据,无法用传统 EV / 收入倍数干净承保 Basis;用情景逻辑比假精确更诚实。[CV012, CV013, CV017, CV018, CV020, CV021]
| 可比对象 | 类型 | 当前披露标记 | 公开倍数 / 规模信号 | 与 Basis 的相关性 | 局限 |
|---|---|---|---|---|---|
| Basis | 私有会计 AI | 2026 年 2 月:$100M 融资,估值约 $1.15B | 无公开 ARR 或收入分母 | 当前承销决策的参考估值 | 标题估值可见;经济性不可见 |
| Numeric | 私有会计自动化 | 2025 年 11 月:$51M Series B;累计融资约 $89M | 声称自动匹配率超过 90%,并向财务运营扩张 | 最接近的工作流自动化同业,且刚拿到融资验证 | 估值未公开披露 |
| Truewind | 私营会计 AI | 2025 年 1 月:$13M Series A;累计融资超过 $17M | 声称收入增长 4x,并吸收 47% 的结账任务 | 可作为 AI 结账自动化采用率的早期同业参照 | 估值未公开披露 |
| Ramp | 私营金融科技 AI | 2026 年 6 月:估值 $44B;年化收入超过 $1B | 70k+ 客户,累计融资超过 $3B | 展示 AI 金融科技里可见分母能支撑什么体量 | 支出和支付平台远比 Basis 宽 |
| BILL | 上市金融科技 / 会计基础设施 | FY2025 收入约 $1.46B;GAAP 毛利率 81.4% | 约 2.08x P/S;约 1.84x EV/revenue | 财务自动化经济性的上市低端可比公司 | 收入结构含支付和交易收入,不只是 SaaS |
| Workiva | 上市报告 / 审计软件 | 2024 年收入约 $738.7M;订阅 / 支持约占 90% | 约 3.03x P/S;约 2.91x EV/revenue | 可比照审计就绪报告和披露工作流 | 规模更大、更成熟,且面向上市公司 |
| BlackLine | 上市会计自动化 | 2024 年 GAAP 毛利率 75.2% | 约 2.59x P/S;约 2.49x EV/revenue | 最接近结账和会计自动化的上市可比公司 | 比 Basis 更成熟,但 AI 原生属性较弱 |
| Intuit | 上市在位平台 | 大型在位者,覆盖广泛 SMB 会计场景 | 约 3.80x P/S;约 3.64x EV/revenue | 可信财务软件的高端在位者倍数参照 | 税务和 SMB 生态更宽,直接可比性并不完美 |
私营公司行使用已审阅来源实际披露的标记:有估值就用估值,否则用轮次规模和规模信号。上市公司行把监管文件与当前市场数据页面配对,让可比集同时锚定基本面和当前定价。
[CV002, CV012, CV013, CV017, CV018, CV020]$1.15B 估值对应的 ARR 门槛差异很大,关键看最终被市场接受的是哪一个收入倍数。
每个柱状值都按估值除以倍数反推。这些是敏感性门槛,不是 Basis 披露的经营指标。
[CV034, CV035, CV044, CV045, CV046, CV047]8.4 乐观、基准与悲观情景区间
情景测算把承保问题讲得很清楚。如果 Basis 仍在 $25M 到 $40M ARR 区间,即便给出慷慨的 8x 到 12x 软件倍数,隐含价值也会比 2026 年 2 月估值低数亿美元。$50M 到 $75M ARR、12x 到 16x 倍数的中间情景,才刚开始覆盖当前价值;并不能让这一轮显得明显便宜。真正有吸引力的上行情景需要更强条件:ARR 已经更接近 $80M 到 $120M,利润率像软件,且客户扩张证据足以让 15x 到 20x 的溢价区间可信。悲观路径不是产品崩盘,而是普通软件重估叠加税务和审计工作流中的 AI 控制风险;一旦出现幻觉、隐私失败或监督缺口,买方会迅速惩罚。上行路径是标准化:如果 Basis 能把头部事务所 logo 和合作伙伴证据转化成可复制、高毛利部署,当前估值仍可能成立。因此,估值立场仍是偏高,而不是崩坏:战略可信度存在,但当前价格只有在保守的基准到乐观区间内才说得通。[CV038, CV039, CV040, CV041, CV042, CV043]
| 情景 | 核心假设 | 估值逻辑 | 概率信号 | 关键风险 |
|---|---|---|---|---|
| 悲观 | ARR 只有 $25M-$40M;买方把 Basis 重新定价到普通软件风险区间。 | 8x-12x 收入意味着价值约 $0.2B-$0.48B。 | 如果部署停滞、错误率反噬,或软件倍数继续压缩,这最可能发生。 | 该轮估值会明显高于最终跑出来的经济性。 |
| 基准 | ARR 达到约 $50M-$75M,经济性像软件但不算顶尖。 | 12x-16x 收入意味着价值约 $0.6B-$1.2B。 | 如果 Basis 确实成立、但阶段仍早于私募估值所暗示的成熟度,这是最可信情景。 | 当前价格只有在 Basis 已接近该区间顶部时才说得通。 |
| 乐观 | ARR 达到约 $80M-$120M,留存、标准化和高质量毛利都很强。 | 15x-20x 收入意味着价值约 $1.2B-$2.4B。 | 需要异常强的增长,再叠加可复制的头部事务所扩张。 | 主要风险是在经济性披露前就假设高端 AI 倍数。 |
| 当前参考估值 | 未披露公开分母。 | 2026 年 2 月融资把当前市场价格标在约 $1.15B。 | 可作为私募清算价参考,不能证明公允价值。 | 价格可能比怀疑者预期更久地高于上市可比逻辑。 |
情景数值是分析敏感性区间,不是公司指引。它把估值反拆为 ARR 和倍数组合,让下行和上行明示出来,而不是藏在假设里。
[CV002, CV044, CV045, CV046, CV047, CV048]当前估值标记高于熊市情景,只落在有纪律基准情景的上半段。
区间是以十亿美元计的估值情景,由明确 ARR 和倍数假设搭出,不是公司指引。
[CV002, CV059, CV060, CV061]8.5 退出准备度、投资逻辑击穿点与最终尽调问题
退出准备度是战略兴奋与可投资估值之间缺失的桥。Basis 可能正在为会计师事务所搭建品类领先的工作流层,但公开记录还没到 IPO-ready,也没有展示足够经济性或控制细节,支撑传统退出承保。现实的近期路径是继续在私营市场复利,而不是干净交接给公开市场。因此,下行触发器异常重要:如果经济性持续不透明,如果税务或审计工作中暴露 AI 控制失败,或者合作伙伴证据始终无法转化为耐久扩张,投资案例可能比产品叙事恶化得更快。最终尽调清单因此不是可选项。在 ARR、留存、服务收入组合、控制证据和优先股堆叠变得可见之前,Basis 应该留在活跃工作队列里,但当前价格不应被视为已经充分验证。[CV038, CV039, CV040, CV041, CV042, CV043]
| 触发器 | 阈值 | 对投资逻辑的传导 | 行动含义 |
|---|---|---|---|
| 经济性持续不透明 | 尽调开放后,ARR、NRR、毛利率和服务收入占比仍不披露。 | 估值仍靠信念驱动,而不是证据驱动。 | 不要把 2026 年 2 月价格当成可投资价格。 |
| 实际工作流出现质量或控制失败 | 税务或审计生产使用中出现重大幻觉、隐私或监督问题。 | 高风险会计工作流买方会很快重估信任。 | 除非价格大幅重置且控制已修复,否则转为回避。 |
| 可比公司倍数继续压缩 | 上市会计软件区间跌破 2026 年 3 月已压缩水平。 | Basis 失去支撑私募增长例外所需的高端区间支撑。 | 重切悲观 / 基准估值,并提高回报门槛。 |
| 合作伙伴验证无法转化为持久扩张 | 具名事务所仍只是参考客户,看不到续约或跨业务线增长。 | 投资逻辑从可扩展平台降级为亮眼试点供应商。 | 从主动尽调转入观察名单。 |
这些是可监控断点,不是泛泛风险。每个触发器都用于告诉投资人:什么时候当前价格不再配得上假设中的溢价。
[CV034, CV035, CV038, CV039, CV040, CV041]| 主题 | 缺失证据 | 重要性 | 负责人或尽调路径 |
|---|---|---|---|
| ARR 与留存 | MRR 或 ARR 桥接、NRR、Logo 留存,以及按队列拆分的扩张 ARR。 | 没有分母,就无法用任何倍数纪律检验估值。 | 财务尽调:索取月度董事会材料中的 KPI 页和队列表。 |
| 股权结构表与优先权 | 完全稀释所有权、优先股堆叠、期权池刷新、老股交易,以及任何债务。 | 普通股回报不只取决于企业价值标题数字。 | 法律与融资尽调:索取股权结构表、公司章程、投资条款清单和附属协议。 |
| 毛利率和服务负担 | 软件与服务收入占比、按工作流拆分的毛利率、上线强度和支持人员配置。 | 只有自动化经济性真正像软件,高端软件倍数才站得住。 | FP&A 尽调:索取分部 P&L 和实施经济性。 |
| 客户质量 | 头部客户集中度、续约表现、标准实施周期,以及赢单 / 输单数据。 | Top-25 渗透只有在能复利增长并续约时才重要。 | 商业尽调:客户访谈,加上 CRM 和续约导出数据审阅。 |
| 风险控制 | 线上错误率跟踪、升级路径、隐私控制、审计轨迹完整性和模型治理文档。 | 税务和审计自动化的容错空间远窄于通用 copilots。 | 产品与合规尽调:索取控制矩阵、事故日志和 QA 评分卡。 |
这些是最低限度尽调请求;只有拿到这些,才能把一家有意思、价格真实的公司,推进为同一价格下可承销的投资。
[CV038, CV039, CV040, CV041, CV042, CV043]8.6 图表
免责声明
本报告仅基于 2026-06-11 或之前检索并审阅的公开可访问信息;任何投资决定前,都应补充管理层材料、客户访谈和法律尽调。
证据索引
| 编号 | 陈述 | 可信度 | 来源 |
|---|---|---|---|
| CO001 | Basis publicly describes itself as an AI agent platform built specifically for accountants and accounting firms. | 中 | SO001, SO006 |
| CO002 | Basis says its product footprint spans client accounting services, tax, audit, and advisory workflows. | 中 | SO001, SO008 |
| CO003 | Basis positions its agents as completing accounting workflows end to end and returning finished work for human review rather than acting as a generic chat assistant. | 中 | SO001, SO002, SO015 |
| CO004 | Basis frames trust, transparency, and verification as core design principles for accountant adoption. | 中 | SO006, SO012 |
| CO005 | Basis was founded in 2023. | 高 | SO006, SO013, SO018 |
| CO006 | Matthew Harpe and Mitchell Troyanovsky are the co-founders named across Basis financing and company materials. | 高 | SO006, SO019 |
| CO007 | Matt Harpe is the public CEO voice for Basis in financing and media coverage. | 中 | SO013, SO014, SO016 |
| CO008 | Independent 2026 coverage places Basis in New York City. | 中 | SO014, SO018 |
| CO009 | Basis's website privacy policy identifies the operating entity as Essex Labs, Inc. doing business as Basis. | 中 | SO005 |
| CO010 | The supportable current stage is private Series B with a unicorn valuation, not a public company. | 高 | SO008, SO013 |
| CO011 | Basis announced a $3.6 million seed round led by Better Tomorrow Ventures with participation from BoxGroup and Avid Ventures. | 高 | SO006, SO019 |
| CO012 | Basis announced a $34 million Series A led by Khosla Ventures in December 2024. | 中 | SO007 |
| CO013 | Basis announced a $100 million Series B at a $1.15 billion valuation on 2026-02-24. | 高 | SO008, SO013 |
| CO014 | Accel and GV led the disclosed Series B syndicate alongside Khosla Ventures and other existing backers. | 高 | SO008, SO013, SO014 |
| CO015 | The public funding record implies about $137.6 million of cumulative capital raised, which outside coverage rounds to about $138 million. | 中 | SO006, SO007, SO013, SO018 |
| CO016 | Businesswire and independent follow-on coverage indicate Keith Rabois was already on the board and Accel's Miles Clements joined at the Series B. | 中 | SO013, SO017, SO018 |
| CO017 | The disclosed investor bench also includes NFDG, Better Tomorrow Ventures, BoxGroup, Avid Ventures, and a roster of technology executives and operators. | 中 | SO007, SO008 |
| CO018 | Basis says it is already working with roughly 30% of the Top 25 accounting firms. | 高 | SO008, SO014, SO017 |
| CO019 | Public descriptions say Basis deployments now cover CAS, tax, and audit practices. | 中 | SO008, SO014, SO017 |
| CO020 | By February 2026 Basis said it had demonstrated the first AI agent to autonomously complete an end-to-end Form 1065 tax return. | 高 | SO008, SO014, SO016 |
| CO021 | Basis and supporting coverage claim 20% to 50% efficiency gains across participating firm workflows. | 中 | SO008, SO014, SO017 |
| CO022 | Basis's careers page says the company is deployed with top accounting firms and runs hundreds of agents behind every accountant. | 中 | SO004 |
| CO023 | Wiss said it began testing Basis in beta for client accounting services in November 2023. | 中 | SO020 |
| CO024 | Wiss publicly announced a strategic partnership and full CAS deployment with Basis in September 2024. | 高 | SO009, SO020 |
| CO025 | Basis announced a Baker Tilly collaboration in April 2026 to expand AI-powered automation across thousands of accounting clients. | 中 | SO010 |
| CO026 | Basis says it is SOC 2 Type II, ISO 27001, and ISO 42001 certified. | 中 | SO003 |
| CO027 | Basis says customer data is stored in the United States and segregated by tenant. | 中 | SO003, SO005 |
| CO028 | Basis says customer data does not train AI models and that every agent action is logged and auditable. | 中 | SO003 |
| CO029 | Basis's privacy policy says website data is processed primarily in the United States and enterprise-product privacy terms live in separate customer agreements. | 中 | SO005 |
| CO030 | Annette Garcia joined Basis in 2024 as its fifth full-time employee. | 中 | SO012 |
| CO031 | Basis has an Accounting Product Operations team that embeds accountants with ML engineers to shape agent behavior and quality. | 中 | SO012 |
| CO032 | Basis says it started in CAS and is extending the platform into tax and audit as of June 2026. | 中 | SO012 |
| CO033 | Basis publicly ties product progress to collaboration with foundation labs and specifically to OpenAI model improvements. | 中 | SO008, SO011, SO025 |
| CO034 | Basis said GPT-5 became the supervising agent model in its workflow stack while GPT-4.1 serves speed-critical interactions. | 中 | SO011 |
| CO035 | Basis says firms using its platform report about 30% time savings on average. | 中 | SO011 |
| CO036 | Basis says benchmarked model orchestration and parallel tool calling are part of how it improves workflow coverage and accuracy. | 中 | SO011 |
| CO037 | Public materials disclose little non-founder executive depth beyond named accountant-facing leaders such as Annette Garcia and investor-board representatives. | 中 | SO002, SO012, SO013 |
| CO038 | Public sources do not disclose Basis revenue, ARR, exact customer count, or a current headcount figure. | 中 | SO002, SO004, SO008 |
| CO039 | Basis's careers page signals active hiring and an intense in-person operating model, but it does not publish a current employee total or office footprint. | 中 | SO004 |
| CO040 | Basis's security page still listed GDPR and CCPA as “coming soon” on 2026-06-11, indicating enterprise-readiness progress is public but not obviously complete. | 中 | SO003 |
| CO041 | Thomson Reuters says 2026 marks a strategic phase of AI adoption across professional services, supporting the timing tailwind behind Basis's category. | 中 | SO023 |
| CO042 | BLS reports about 1.58 million accountant and auditor jobs in 2024 with roughly 124,200 annual openings, supporting the labor-shortage context Basis emphasizes. | 中 | SO024 |
| CO043 | SEC cybersecurity disclosure rules illustrate the governance and incident-reporting expectations that any future public-company path or acquirer scrutiny would impose on Basis. | 中 | SO021 |
| CO044 | NIST's AI Risk Management Framework shows that trustworthy-AI controls are now explicit evaluation criteria for AI systems in high-stakes work. | 中 | SO022 |
| CO045 | Basis's mission framing is that accountants should shift from doing repetitive work to reviewing, judging, and advising. | 中 | SO001, SO002 |
| CO046 | Wiss says Basis can reduce manual work while improving work-life balance and the quality of service for clients. | 中 | SO020 |
| CO047 | The Baker Tilly announcement suggests Basis is moving from one-firm pilots toward larger managed-services and distribution-style rollouts. | 中 | SO010 |
| CO048 | Basis was still using a public OpenAI Codex customer-spotlight narrative in March 2026, reinforcing that the OpenAI relationship is part of its go-to-market story. | 中 | SO025 |
| CO049 | The retained public record supports a five-person founding cohort spanning Matthew Harpe, Mitchell Troyanovsky, Zenna Tavares, Emily Mackevicius, and Eli Bingham, even though different public materials emphasize different subsets. | 中 | SO002, SO006, SO019, SO025, SO026 |
| CO050 | External coverage describes Matt Harpe as a former Boston Consulting Group and SoftBank executive, giving him the clearest commercial-operator background in the founder set. | 中 | SO026, SO028 |
| CO051 | Basis and Zenna Tavares's personal site describe him as a co-founder focused on reasoning and causal probabilistic programming, with Columbia innovation-scholar experience and an MIT Ph.D. | 中 | SO002, SO027 |
| CO052 | Basis describes Emily Mackevicius as a co-founder and senior research scientist who leads Collaborative Intelligent Systems and previously trained at Columbia and MIT. | 中 | SO002 |
| CO053 | Basis describes Eli Bingham as a co-founder and director, a Broad Institute fellow, a Pyro co-creator, and a former senior research scientist at Uber AI Labs. | 中 | SO002 |
| CO054 | Retained public materials identify Mitchell Troyanovsky as a co-founder and product-facing representative, but they do not provide a fuller pre-Basis biography. | 中 | SO006, SO019, SO025 |
| CM001 | Basis publicly positions itself as a platform built specifically for accountants and accounting firms. | 高 | SM001, SM018 |
| CM002 | Basis's public product framing spans CAS, tax, audit, and advisory workflows. | 高 | SM001, SM018 |
| CM003 | BLS occupation definitions show that accountants and auditors serve corporations, governments, individuals, and nonprofits, meaning the labor market is broader than Basis's immediate buyer set. | 中 | SM005 |
| CM004 | IBISWorld puts the U.S. accounting-services market at $157.4 billion in 2026. | 中 | SM002 |
| CM005 | AnythingResearch says the accounting-services industry reached $145.4 billion in 2025. | 中 | SM021 |
| CM006 | Public accounting-services estimates differ materially across publishers, so Basis should not be sized from a single headline TAM figure. | 高 | SM002, SM021 |
| CM007 | The Business Research Company says the global accounting-software market was $20.03 billion in 2025. | 中 | SM023 |
| CM008 | The Business Research Company says the global accounting-software market will reach $22.72 billion in 2026. | 中 | SM023 |
| CM009 | Mordor Intelligence values the global accounting-software market at $21.56 billion in 2025. | 中 | SM003 |
| CM010 | Mordor Intelligence estimates the global accounting-software market at $23.47 billion in 2026 and says North America held 38.35% revenue share in 2025. | 中 | SM003 |
| CM011 | Precedence Research estimates the global accounting-software market at $21.16 billion in 2025 and $23.1 billion in 2026. | 中 | SM024 |
| CM012 | Analyst accounting-software definitions include business owners, corporate professionals, and multiple industry verticals, so the software outer boundary is broader than Basis's firm-focused wedge. | 中 | SM023, SM024 |
| CM013 | Thomson Reuters reports that 40% of organizations now use GenAI in 2026, up from 22% in 2025. | 中 | SM004 |
| CM014 | Thomson Reuters reports that only 15% of organizations use agentic AI today. | 中 | SM004 |
| CM015 | Another 53% of organizations are planning or considering agentic AI adoption. | 中 | SM004 |
| CM016 | Only 18% of professionals say their organizations track AI ROI. | 中 | SM004 |
| CM017 | Two-thirds of corporate respondents want their outside firms to use AI, but fewer than 20% mandate it. | 中 | SM004 |
| CM018 | The professional-accounting market is therefore moving into AI adoption quickly, but agentic workflow automation remains early and under-measured. | 高 | SM004, SM018 |
| CM019 | BLS counted 1,579,800 accountant and auditor jobs in 2024. | 中 | SM005 |
| CM020 | BLS projects about 124,200 openings for accountants and auditors each year on average from 2024 to 2034. | 中 | SM005 |
| CM021 | The median annual wage for accountants and auditors reached $81,680 in 2024. | 中 | SM005 |
| CM022 | NPAG was convened in response to the accounting talent shortage and released its final report in August 2024. | 中 | SM006 |
| CM023 | NPAG said its final report incorporated nearly 8,000 survey responses. | 中 | SM006 |
| CM024 | NYSSCPA's summary of the AICPA advisory work says roughly 340,000 accounting professionals left the profession between 2019 and 2023. | 中 | SM007 |
| CM025 | NYSSCPA says only one in nine undergraduate business students chose an accounting major. | 中 | SM007 |
| CM026 | NYSSCPA says 57% of business majors who did not pursue accounting cited the 150-hour licensure path as a deterrent. | 中 | SM007 |
| CM027 | The CPA Journal says over 300,000 accountants left the field between 2019 and 2022. | 中 | SM008 |
| CM028 | The CPA Journal says accounting graduates fell to about 47,000, down 10% from 2021. | 中 | SM008 |
| CM029 | The CPA Journal cites reporting that more than 720 companies flagged insufficient accounting staff as a source of potential errors, up 30% from 2019. | 中 | SM008 |
| CM030 | BLS's March 2026 Employment Cost Index shows wages and salaries for civilian workers rose 3.4% year over year. | 中 | SM026 |
| CM031 | PCAOB Docket 052 on technology-assisted analysis was approved by the SEC. | 中 | SM020 |
| CM032 | Audit technology adoption is occurring under a rising control bar rather than a relaxed one, increasing the value of trustworthy, reviewable automation. | 高 | SM020, SM005 |
| CM033 | Intuit's annual-reports page shows a 2025 10-K filing dated September 3, 2025. | 中 | SM009 |
| CM034 | Intuit reported Global Business Solutions revenue of $3.2 billion for the quarter ended January 31, 2026. | 中 | SM010 |
| CM035 | Intuit reported Online Ecosystem revenue of $2.5 billion and QuickBooks Online Accounting growth of 24% in the same quarter. | 中 | SM010 |
| CM036 | Intuit Accountant Suite markets AI-powered workflows and unified data specifically to accounting firms. | 中 | SM011 |
| CM037 | Sage says it is trusted by more than 11,000 accounting and bookkeeping practices in the United States. | 中 | SM012 |
| CM038 | Xero markets cloud-accounting automation to U.S. small businesses and lists standard plan pricing from $25 to $90 per month before promotions. | 中 | SM013 |
| CM039 | Botkeeper positions itself as an automated bookkeeping solution built for accounting firms. | 中 | SM014 |
| CM040 | Pilot's pricing page starts bookkeeping at $99 per month and adds human bookkeeping and CFO layers in higher plans. | 中 | SM015 |
| CM041 | Numeric positions itself as AI-powered close automation and says customers use it to keep headcount low and time-to-close lower. | 中 | SM016 |
| CM042 | Digits positions itself as the first AI-native general ledger and lists plans at $65, $100, and $250 per month. | 中 | SM017 |
| CM043 | The market already spans incumbent accountant suites, tech-enabled services, bookkeeping automation, close software, and AI-native ledgers rather than a single unified AI category. | 高 | SM011, SM012, SM013, SM014, SM015, SM016, SM017 |
| CM044 | AnythingResearch says there are 52,535 companies in the industry. | 中 | SM021 |
| CM045 | AnythingResearch says the top four companies hold 42.8% market share. | 中 | SM021 |
| CM046 | AnythingResearch says average sales per location are $2.6 million and average pay per employee is $90,988. | 中 | SM021 |
| CM047 | CPA Practice Advisor says Basis is used by approximately 30% of the Top 25 accounting firms. | 中 | SM018 |
| CM048 | CPA Practice Advisor says Basis claims 20% to 50% efficiency gains across practices. | 中 | SM018 |
| CM049 | Basis's homepage and supporting coverage both frame the product across CAS, tax, audit, and advisory, matching the workflow mix of larger public-accounting firms rather than a narrow bookkeeping-only tool. | 高 | SM001, SM018 |
| CM050 | Wiss said it began testing Basis in beta for CAS in November 2023 and then deployed it across the practice after successful results. | 中 | SM027 |
| CM051 | Wiss said it already used Microsoft Copilot but viewed Basis's integrated end-to-end automation as transformative. | 中 | SM027 |
| CM052 | Public evidence most strongly supports Basis's wedge in larger accounting firms rather than in corporate controllership teams. | 高 | SM001, SM018, SM027 |
| CM053 | A supportable broad TAM for Basis is roughly $145 billion to $157 billion of U.S. accounting-services spend. | 高 | SM002, SM021 |
| CM054 | A supportable North American accounting-software upper bound is roughly $8 billion to $9 billion when ~38% regional share is applied to $21 billion to $23 billion global estimates. | 中 | SM003, SM024 |
| CM055 | Public data is insufficient to isolate exact accounting-firm AI SAM or Basis revenue-share SOM. | 中 | SM003, SM024, SM018 |
| CM056 | The best public SOM proxy is firm penetration — about 30% of the Top 25 accounting firms — rather than disclosed revenue share or seat count. | 中 | SM018, SM027 |
| CP001 | Basis positions itself as an accounting-firm agent platform spanning CAS, tax, audit, and advisory rather than a single-point bookkeeping tool. | 高 | SP001, SP003 |
| CP002 | Basis markets finished output as review-ready work that shifts accountants from doing work to reviewing it. | 高 | SP001, SP002 |
| CP003 | On retained public evidence, Basis’s closest direct peers are accounting-native workflow vendors such as Botkeeper, Truewind, and Digits, while Pilot, Vic.ai, BILL, Intuit Assist, and Microsoft Copilot are substitutes or adjacent competitors rather than perfect like-for-like peers. | 中 | SP001, SP006, SP008, SP010, SP013, SP016, SP017, SP018, SP020 |
| CP004 | Botkeeper markets itself as an AI accounting solution purpose-built for accounting firms. | 中 | SP006 |
| CP005 | Botkeeper explicitly layers dedicated services and support on top of automation instead of positioning as software-only self-serve infrastructure. | 中 | SP006 |
| CP006 | Botkeeper says its process has been used by more than 200 accounting firms and 5,000-plus business clients. | 中 | SP006 |
| CP007 | Pilot targets startups and small businesses with bookkeeping, tax, CFO, and outsourced operations services rather than accounting-firm practice automation. | 高 | SP008, SP009 |
| CP008 | Pilot describes its operating model as a combination of people and software rather than autonomous end-to-end agent execution. | 中 | SP008 |
| CP009 | Truewind keeps Sage Intacct or QBO as the source of truth and turns financial source materials into GL-ready journal entries, reconciliations, and SOPs for human review. | 中 | SP016 |
| CP010 | Truewind’s retained public positioning is concentrated on close management, reconciliation, transaction coding, and schedules, not tax or audit workflow ownership. | 中 | SP016 |
| CP011 | Digits positions itself as an AI-native general ledger spanning bookkeeping, month-end close, bill pay, invoicing, and real-time financials. | 高 | SP013, SP014 |
| CP012 | Digits for Firms sells ledger replacement plus firm branding, AI close agents, and client-facing reporting for accounting firms. | 高 | SP015, SP014 |
| CP013 | Vic.ai is centered on AP automation and autonomous finance, especially invoice processing, approvals, and payables. | 高 | SP010, SP011 |
| CP014 | BILL competes in AP/AR and accountant workflow tooling rather than end-to-end tax or audit execution. | 中 | SP020 |
| CP015 | Intuit Assist for QuickBooks focuses on automating invoicing, workflow tasks, and financial Q&A inside QuickBooks for SMB users. | 中 | SP017 |
| CP016 | Microsoft 365 Copilot is a horizontal work AI layer spanning chat, search, agents, and app integration rather than an accounting-native workflow product. | 高 | SP018, SP019 |
| CP017 | Basis’s retained public scope is broader across accounting-firm practice lines than Botkeeper, Truewind, Vic.ai, BILL, Intuit Assist, or Microsoft Copilot. | 中 | SP001, SP003, SP006, SP010, SP016, SP017, SP018, SP020 |
| CP018 | Botkeeper publishes tiered per-license pricing from $149 per month for 1-4 licenses down to $59 monthly or $53 annually for 25-plus licenses. | 中 | SP007 |
| CP019 | Pilot publishes bookkeeping plans starting at $99 per month and CFO plans starting at $1,750 per month, with higher public tiers at $3,150 and $5,250 per month. | 中 | SP009 |
| CP020 | Digits publishes Essentials, Core, and Pro plans at $65, $100, and $250 per month, plus partner-provided accounting, CFO, and tax services. | 中 | SP014 |
| CP021 | Intuit says Intuit Assist is available at no additional cost for certain QuickBooks users. | 中 | SP017 |
| CP022 | Microsoft says Copilot Chat is available at no additional cost for eligible Microsoft 365 users, while fuller Microsoft 365 Copilot is an added license. | 中 | SP019 |
| CP023 | Basis does not publish list pricing or enterprise contract terms in retained official pages. | 中 | SP001, SP002 |
| CP024 | Vic.ai and BILL use demo-led sales motions in retained pages and do not expose list pricing. | 中 | SP010, SP011, SP020 |
| CP025 | Basis markets SOC 2 Type II, ISO 27001, ISO 42001, US-only hosting, no model training on customer data, and traceable agent actions. | 中 | SP002 |
| CP026 | Botkeeper markets SOC2 Type 2 and bank-grade security. | 中 | SP006 |
| CP027 | Vic.ai says it maintains SOC 1 Type II and SOC 2 Type II, follows an ISO 27001 framework, and separates global derived-data learning from customer-specific local models. | 中 | SP012 |
| CP028 | Digits says it is SOC2 Type II and uses TLS plus revocable-token access patterns. | 中 | SP014 |
| CP029 | Microsoft says prompts and responses are not used to train the underlying models and remain inside Microsoft 365 enterprise data-protection boundaries. | 中 | SP019 |
| CP030 | PCAOB reported that current GenAI use in audits remains focused mainly on administrative and research activities and still needs strong supervision because of privacy and security risks. | 中 | SP022 |
| CP031 | NIST’s AI RMF and GenAI profile make trustworthy-AI controls a baseline requirement for high-stakes deployments, which raises procurement expectations for accounting AI vendors. | 中 | SP023 |
| CP032 | Wiss said Basis’ integrated end-to-end automation was transformative even though the firm already used Copilot and a private chat instance. | 中 | SP004 |
| CP033 | Wiss began with a CAS beta in late 2023 and expanded Basis after early success, implying a wedge from CAS into broader firm deployment. | 中 | SP004 |
| CP034 | Baker Tilly’s collaboration with Basis shows Basis is also pursuing distribution or managed-services leverage beyond direct firm-by-firm seat sales. | 中 | SP025 |
| CP035 | Basis says firms use it across CAS, tax, audit, and advisory, and OpenAI says the company supports a significant share of large U.S. accounting firms. | 高 | SP001, SP003 |
| CP036 | Crowdfund Insider says Basis already partners with about 30% of the top 25 U.S. accounting firms and sees estimated productivity lifts of 20-50% in key areas. | 中 | SP005 |
| CP037 | Botkeeper’s installed base gives it meaningful distribution inside accounting firms even though its public scope is narrower than Basis’s multi-practice agent story. | 中 | SP006, SP007 |
| CP038 | Pilot’s scale and services-heavy model make it a substitute for buyers who want outsourced finance outcomes rather than software-only workflow automation. | 中 | SP008, SP009 |
| CP039 | QuickBooks and Microsoft benefit from pre-existing suite distribution and familiar workflow surfaces, which can lower adoption friction versus a new standalone platform. | 中 | SP017, SP018, SP019 |
| CP040 | AP specialists like Vic.ai and BILL can capture automation budget before a firm adopts a broader accounting-workflow platform. | 中 | SP010, SP011, SP020 |
| CP041 | Basis’s moat is workflow depth plus reviewability, not public price leadership or suite bundling. | 中 | SP001, SP002, SP003, SP004 |
| CP042 | Model progress can commoditize assistant features over time, making generic-suite vendors and adjacent tools more credible substitutes even if Basis remains ahead on workflow depth today. | 中 | SP003, SP018, SP019, SP021 |
| CP043 | Public trust posture is necessary but not unique because Basis, Botkeeper, Vic.ai, Digits, and Microsoft all market meaningful security or privacy controls. | 中 | SP002, SP006, SP012, SP014, SP019 |
| CP044 | Switching costs sit in ledgers, ERP or QBO source-of-truth choices, reviewer habits, client portals, and firm-specific playbooks more than in one-time AI prompts. | 中 | SP001, SP013, SP015, SP016 |
| CP045 | The sharpest adverse angle is incumbent distribution power plus opaque enterprise pricing: Basis may be more workflow-native, but buyers can still default to bundled QuickBooks or Microsoft tools or cheaper published alternatives like Digits and Botkeeper. | 中 | SP007, SP014, SP017, SP019 |
| CI002 | Basis positions itself as an AI agent platform for accountants across client accounting services, tax, audit, and advisory work. | 中 | SI001, SI007 |
| CI003 | Public company-linked sources say Basis works with roughly 30% of the top 25 accounting firms. | 高 | SI004, SI012, SI017 |
| CI004 | Basis says accounting firms face tightening margins, rising client demands, and not enough experienced staff. | 中 | SI001, SI012 |
| CI005 | Basis announced a $100 million Series B at a $1.15 billion valuation in February 2026. | 高 | SI007, SI012, SI014 |
| CI006 | The Series B proceeds were described as funding continued platform development and expansion of engineering and machine learning teams. | 中 | SI012, SI014 |
| CI007 | Basis disclosed a $34 million Series A led by Khosla Ventures in December 2024. | 中 | SI008 |
| CI008 | Basis emerged from stealth with a $3.6 million seed round led by Better Tomorrow Ventures. | 中 | SI009 |
| CI009 | The disclosed seed, Series A, and Series B rounds sum to approximately $137.6 million of cumulative equity financing. | 中 | SI007, SI008, SI009 |
| CI010 | Basis exposes no public list pricing on its website and instead routes buyers to contact-led enterprise contracting. | 中 | SI001, SI006 |
| CI011 | Basis states that signed MSAs or order forms control service access when they exist. | 中 | SI006 |
| CI012 | Basis says privacy terms for enterprise products are governed by separate written customer agreements rather than the public website privacy policy. | 中 | SI005 |
| CI013 | Basis publicly markets modules and workflows across CAS, tax, audit, and advisory practices. | 中 | SI001 |
| CI014 | Basis says it is SOC 2 Type II, ISO 27001, and ISO 42001 certified and hosts customer data in the United States. | 中 | SI003 |
| CI015 | Basis's terms say AI outputs may contain errors and require human review and verification before use in professional work. | 中 | SI006 |
| CI016 | Wiss says it tested Basis in CAS beginning in November 2023 and then moved to full deployment across that practice, with adoption increasing rapidly month over month. | 中 | SI010 |
| CI017 | Basis's home page features testimonials from Wiss, UHY, Clark Nuber, Boulay, and BPB, indicating multi-firm customer proof across practice areas. | 中 | SI001 |
| CI018 | OpenAI's Basis case study says firms using Basis report up to 30% average time savings and increased capacity for higher-leverage work. | 中 | SI018 |
| CI019 | Company-linked funding coverage cites 20% to 50% efficiency gains across practices from using Basis. | 中 | SI012, SI013 |
| CI020 | Basis's GTM appears to be direct enterprise selling plus partner-style rollout rather than self-serve distribution. | 中 | SI001, SI006, SI010, SI011 |
| CI021 | No retrieved public source discloses Basis's ARR, revenue, gross margin, CAC, NRR, or current cash balance. | 中 | SI001, SI007, SI012 |
| CI022 | Basis said at launch that its platform was already serving clients across CFO advisory, assurance, client accounting services, and bookkeeping. | 中 | SI009 |
| CI023 | The BLS projects about 124,200 annual openings for accountants and auditors over 2024 to 2034. | 中 | SI019 |
| CI024 | Thomson Reuters says the 2026 professional-services AI market has moved beyond early adoption into a strategic phase. | 中 | SI020 |
| CI025 | Journal of Accountancy warns CPAs about hallucinations, prompt injection, privacy risks, and overreliance when deploying AI in accounting work. | 中 | SI021 |
| CI026 | NIST says its AI Risk Management Framework is intended to help organizations incorporate trustworthiness considerations into the design, development, use, and evaluation of AI systems. | 中 | SI022 |
| CI027 | Journal of Accountancy says some accounting firms have seen AI vendors fold before scheduled demos, underscoring buyer diligence risk in the sector. | 中 | SI021 |
| CI028 | Workiva's 2025 10-K says approximately 92% of revenue came from subscription and support fees, with the remainder from professional services, and notes services mix can affect gross margins. | 中 | SI025 |
| CI029 | Workiva says more than 6,600 organizations, including over 85% of the Fortune 1000, use its platform, showing how large reporting-software vendors monetize account expansion in CFO and audit offices. | 中 | SI025 |
| CI030 | Intuit's 2025 10-K says it sells financial-management and professional-tax products to accountants through both direct sales and partner channels. | 中 | SI024 |
| CI031 | BILL's 2025 10-K says some contracts are annual subscriptions, many arrangements can be terminated, and selling additional services can require more costly sales and marketing. | 中 | SI023 |
| CI032 | BILL reported 81.7% GAAP gross margin and 85.0% non-GAAP gross margin in fiscal 2025, indicating scaled finance software can be highly profitable once mature. | 低 | SI023 |
| CI033 | Basis likely monetizes recurring enterprise software with an attached onboarding and support layer rather than simple seat-based self-serve pricing. | 中 | SI001, SI006, SI010, SI011 |
| CI034 | Basis's public traction is stronger on logos, deployments, and adoption percentages than on disclosed units such as ARR, seats, or retention. | 中 | SI001, SI012, SI018 |
| CI035 | Basis's stated security and compliance posture implies a meaningful fixed-cost base spanning audits, controls, infrastructure, and documentation. | 中 | SI003 |
| CI036 | Basis's no-model-training, auditability, and human-oversight commitments likely add implementation and support burden relative to lightweight generic AI tools. | 中 | SI003, SI005, SI006 |
| CI037 | Accounting labor scarcity and strategic AI adoption trends support willingness to pay for tools that expand firm capacity. | 中 | SI019, SI020, SI001 |
| CI038 | Those market tailwinds do not prove realized pricing, CAC efficiency, or durable revenue quality for Basis. | 中 | SI001, SI020, SI021 |
| CI039 | Basis appears well-capitalized relative to many vertical SaaS startups because it has disclosed about $137.6 million of equity financing and no visible leverage, but actual cash on hand is unknown. | 中 | SI007, SI008, SI009 |
| CI040 | If only the fresh Series B is assumed available for current scaling, an illustrative burn range of $3.0 million to $5.5 million per month would imply roughly 18 to 33 months of runway. | 低 | SI004, SI007, SI014 |
| CI041 | Actual runway could be longer than the Series-B-only scenario if Basis retained prior cash or if recurring revenue offsets spend, but public sources do not disclose either input. | 中 | SI007, SI008, SI021 |
| CI042 | Basis's financing dependency is lower in the near term than that of earlier-stage peers, but not eliminated while the company expands engineering, machine learning, and enterprise deployments without public revenue disclosure. | 中 | SI004, SI012, SI014 |
| CI043 | Basis's next financing trigger likely depends on proving realized pricing, gross margin after support, and expansion within large firms rather than on adoption headlines alone. | 中 | SI010, SI018, SI025 |
| CI044 | Missing cash balance, ARR history, net retention, customer concentration, and software-versus-services mix block standard financial underwriting of Basis. | 中 | SI001, SI012, SI021 |
| CI045 | Basis's value messaging centers on capacity expansion, faster close cycles, and higher-value advisory work, which implies ROI-based selling rather than transparent seat pricing. | 中 | SI001, SI018 |
| CI046 | Basis moved from seed financing to a $34 million Series A and then to a $100 million Series B within roughly 27 months, signaling strong investor conviction and elevated growth expectations. | 中 | SI007, SI008, SI009 |
| CI047 | OpenAI's case study shows Basis depends on frontier model vendors for core capability gains, which improves product performance but also creates third-party model-cost exposure in the cost structure. | 中 | SI018 |
| CI048 | CAMICO guidance for CPA firms says autonomous AI and direct AI-to-client interactions require stronger monitoring, provider due diligence, contractual safeguards, and human review. | 中 | SI026 |
| CI049 | The Uniqus 2026 internal-controls paper says GenAI is already being used in journal entry preparation, account reconciliation, revenue contract analysis, and financial-close processes, and weak controls can create audit findings or material-weakness risk. | 中 | SI027 |
| CE001 | Basis publicly markets itself as an AI agent platform built specifically for accountants rather than a horizontal productivity assistant. | 高 | SE001, SE002 |
| CE002 | The homepage says Basis runs in the background, executes accounting workflows end to end, and returns finished output ready for review. | 中 | SE001 |
| CE003 | Basis frames the accountant role as shifting from doer to reviewer who manages agents, applies judgment, and advises clients. | 高 | SE001, SE006 |
| CE004 | Basis says outputs adapt to the standards of a firm, practice, and engagement rather than using one generic workflow for all users. | 中 | SE001 |
| CE005 | The public module map spans CAS, Tax, Audit, and Advisory. | 中 | SE001 |
| CE006 | Named homepage testimonials show public proof points across CAS, tax, audit, and firm-level adoption rather than a single narrow use case. | 中 | SE001 |
| CE007 | Annette Garcia says Basis started in CAS as core accounting before expanding into tax and audit. | 中 | SE006 |
| CE008 | Annette Garcia says tax has a large regulatory surface area and audit requires substantial professional judgment. | 高 | SE006, SE028, SE030 |
| CE009 | OpenAI says Basis automates repetitive accounting tasks such as reconciliations, journal entries, and financial summaries. | 中 | SE017 |
| CE010 | Series B coverage says Basis demonstrated an AI agent that autonomously completes an end-to-end 1065 tax return. | 高 | SE013, SE014 |
| CE011 | Independent coverage and OpenAI partner materials both describe Basis as using long-horizon agents that can work on accounting workflows over many hours. | 高 | SE013, SE014, SE017 |
| CE012 | Annette Garcia says ordinary chatbots are insufficient for accounting because accountants need consistency, audit trails, predictability, and outputs that hold up under review. | 中 | SE006 |
| CE013 | Wiss publicly contrasted Basis with its own private ChatGPT instance and Copilot usage, saying Basis delivered integrated end-to-end automation. | 中 | SE015 |
| CE014 | The About page says Basis chose to build a unified accounting intelligence rather than layer a chatbot on top of accounting data. | 中 | SE002 |
| CE015 | OpenAI says Basis treats accounting as a system of workflows with a multi-agent architecture. | 中 | SE017 |
| CE016 | OpenAI says each Basis task begins with a supervising agent that coordinates the full process. | 中 | SE017 |
| CE017 | OpenAI says the supervising agent was originally built on o3 and later migrated to GPT-5. | 中 | SE017 |
| CE018 | OpenAI says Basis routes work to specialized sub-agents based on task, complexity, latency needs, and input type. | 中 | SE017 |
| CE019 | OpenAI says GPT-4.1 is used for speed-critical interactions while GPT-5 is used for more complex multi-step accounting scenarios. | 中 | SE017 |
| CE020 | OpenAI says Basis uses an internal benchmark suite to choose models and determine when agents can safely take on new workflows. | 中 | SE017 |
| CE021 | OpenAI says Basis agents share context through a central layer that surfaces assumptions, data sources, and logic behind decisions. | 中 | SE017 |
| CE022 | OpenAI says function calling helped Basis agents complete multi-step processes like reconciliations and journal entries rather than only suggesting them. | 中 | SE017 |
| CE023 | The reviewed public Basis materials in this chapter do not enumerate non-OpenAI production model providers. | 低 | SE001, SE007, SE008 |
| CE024 | Basis security materials claim SOC 2 Type II, ISO 27001, and ISO 42001 certifications. | 中 | SE003 |
| CE025 | Basis says customer data never trains or improves any AI model. | 中 | SE003 |
| CE026 | Basis says it maintains strict tenant separation and data isolation across customers. | 中 | SE003 |
| CE027 | Basis says every agent action is logged and auditable and that AI steps show sources, changes, and rationale. | 中 | SE003 |
| CE028 | Basis says it does not store customer credentials and instead uses revocable tokens. | 中 | SE003 |
| CE029 | Basis says data is encrypted at rest and in transit, MFA is enforced, and daily backups are maintained. | 中 | SE003 |
| CE030 | Basis says customer data is stored in the United States and its website privacy policy says website data is processed primarily in the United States. | 高 | SE003, SE004 |
| CE031 | Basis security materials still label GDPR and CCPA as coming soon. | 中 | SE003 |
| CE032 | Basis privacy policy says enterprise product privacy terms are governed by separate customer agreements rather than publicly posted website terms. | 中 | SE004 |
| CE033 | The reviewed Basis public materials do not publish a public uptime SLA, status page, or incident history dashboard. | 低 | SE001, SE003, SE008 |
| CE034 | The reviewed Basis public materials do not enumerate a full production integration list covering ERP systems, tax engines, or document systems. | 低 | SE001, SE003, SE008 |
| CE035 | A UHY testimonial on the Basis homepage says month-end close time was significantly reduced for clients and that onboarding and implementation support were strong. | 中 | SE001 |
| CE036 | Wiss says it began beta testing Basis in November 2023 and later deployed it fully across CAS after successful results. | 中 | SE015 |
| CE037 | Baker Tilly says its collaboration with Basis is aimed at expanding AI-powered automation across thousands of accounting clients. | 中 | SE016 |
| CE038 | The careers page says Basis is deployed with top accounting firms and operates hundreds of agents behind every accountant. | 中 | SE005 |
| CE039 | Annette Garcia says Accounting Product Operations works closely with ML engineers to ensure Basis agents produce work that meets professional-accountant standards. | 中 | SE006 |
| CE040 | The AGI-era post says Atlas owns context architecture, context management, integrations, and agent development for internal agent use at Basis. | 中 | SE009 |
| CE041 | The unified MCP post says Basis centralized internal tool access, authorization, and telemetry for agent operations behind one gateway. | 中 | SE010 |
| CE042 | The monorepo engineering post says Basis optimizes its internal codebase for verifiability, canonical context, and interoperability to support agent-driven development. | 中 | SE011 |
| CE043 | The agent-managers post says strong evals and orchestration are required as models take on more valuable tasks. | 中 | SE012 |
| CE044 | QuickBooks AI and Microsoft 365 Copilot are marketed as broad assistants or cross-functional work agents rather than accounting-firm-native systems. | 中 | SE021, SE022 |
| CE045 | Botkeeper, Vic.ai, and Digits each market more bounded bookkeeping, AP, or ledger surfaces than Basis’s stated CAS-Tax-Audit firm-wide scope. | 中 | SE001, SE023, SE024, SE025 |
| CE046 | Intuit and Xero publish technical documentation for accounting APIs and OAuth flows, showing that major ecosystem integrations are a material implementation consideration in this market. | 中 | SE019, SE020 |
| CE047 | NIST and the PCAOB both frame trustworthy AI governance and human supervision as central requirements in high-stakes financial contexts. | 高 | SE026, SE028 |
| CE048 | IRS guidance says tax software cannot replace professional judgment, which supports Basis’s review-first framing for tax work. | 高 | SE006, SE030 |
| CE049 | The Thomson Reuters Institute says professional-services AI has entered a strategic phase focused on workflow redesign rather than experimentation. | 中 | SE031 |
| CE050 | AICPA and CIMA position AI adoption as a capability-building exercise for accountants rather than a purely autonomous replacement story. | 中 | SE029 |
| CE051 | OpenAI’s tax-agents case study shows that self-improving tax systems rely on practitioner feedback, production traces, and eval-backed iteration loops. | 中 | SE018 |
| CE052 | The reviewed public Basis materials do not disclose workflow-level accuracy rates, exception rates, override rates, or formal audit-quality SLAs. | 低 | SE001, SE003, SE013, SE017 |
| CU001 | Basis publicly positions itself as one platform serving accounting-firm workflows across CAS, tax, audit, and advisory rather than as a single-purpose bookkeeping tool. | 中 | SU001, SU002 |
| CU002 | Basis says it is working with roughly 30% of the top 25 U.S. accounting firms. | 中 | SU009, SU019 |
| CU003 | Independent trade coverage repeats the claim that Basis is deployed across CAS, tax, and audit at about 30% of the top 25 firms. | 中 | SU010, SU011 |
| CU004 | The Basis home page names Wiss, UHY, Clark Nuber, Boulay, and BPB as customer references, indicating the public customer set is broader than a single flagship logo. | 中 | SU001 |
| CU005 | Wiss says it began testing Basis in beta for its CAS practice in November 2023 and that successful results led to a full deployment across the practice. | 中 | SU006 |
| CU006 | Wiss says internal adoption of Basis increased rapidly month over month and other practice areas are eager to introduce it into their workflows. | 中 | SU006 |
| CU007 | Wiss said in late 2025 that Basis had significantly increased the effectiveness of its personnel and enabled faster monthly close processes with greater accounting accuracy. | 中 | SU013 |
| CU008 | The public Wiss evidence suggests a firm-led buying motion in which partners or leadership sponsor Basis while accountants inside the practice become day-to-day users. | 中 | SU006, SU013 |
| CU009 | Baker Tilly said it evaluated multiple platforms against complex client accounting scenarios before selecting Basis for the next phase of its managed-services expansion. | 中 | SU007 |
| CU010 | Baker Tilly said it plans to use Basis across transaction processing, journal entry support, reconciliations, post-close analysis, management reporting, budgeting, and forecasting. | 中 | SU007 |
| CU011 | Baker Tilly frames the rollout as a managed-services initiative, implying practice leadership is the buyer and payer while accounting and finance delivery teams are the users. | 中 | SU007 |
| CU012 | A UHY principal says Basis significantly reduced the month-end close for clients and that onboarding and support were delivered as promised. | 中 | SU001 |
| CU013 | UHY’s CAS page shows the surrounding use-case environment spans transaction processing through outsourced controller and CFO work, which helps explain where Basis can fit inside that customer segment. | 中 | SU008 |
| CU014 | Clark Nuber’s testimonial says Basis saves time, streamlines workflows, and keeps surfacing new use cases in the firm’s CAS department. | 中 | SU001 |
| CU015 | Boulay’s CIO says Basis can serve the entire firm across tax, audit, advisory, and beyond, which points to cross-practice aspiration rather than a single-workflow niche. | 中 | SU001 |
| CU016 | BPB’s tax-practice quote says trusted agents let the team focus on value-add work and client service instead of routine execution. | 中 | SU001 |
| CU017 | OpenAI says firms using Basis report 30% time savings on average. | 中 | SU005 |
| CU018 | OpenAI says firms continue expanding agent responsibilities as trust grows, which is a real but still narrative expansion signal. | 中 | SU005 |
| CU019 | Series B coverage repeats an investor claim that Basis is driving roughly 20% to 50% efficiencies across practices. | 低 | SU009, SU010 |
| CU020 | Basis says it is already at work with leading accounting firms across the country and is expanding into new segments and new types of organizations. | 中 | SU002 |
| CU021 | Basis’s seed materials say the platform was already serving clients across CFO advisory, assurance, client accounting services, bookkeeping, and more. | 中 | SU021 |
| CU022 | Basis says it started with CAS and is now extending into tax and audit because it first had to master core accounting workflows. | 中 | SU022 |
| CU023 | Basis’s accounting product operations lead says accountants want consistency, audit trail, predictability, and outputs that still make sense when reviewed later. | 中 | SU022 |
| CU024 | Basis publicly says every agent action is logged and auditable and that AI steps show sources, changes, and rationale. | 中 | SU003 |
| CU025 | Basis publicly says customer data does not train models, data is isolated by tenant, and enterprise-product privacy terms are governed by separate written agreements. | 高 | SU003, SU004 |
| CU026 | Thomson Reuters says the main AI-adoption barriers for accounting firms include privacy and security, integration cost and complexity, training, regulatory compliance, and managing client trust. | 中 | SU015 |
| CU027 | CPA Practice Advisor says CPA firms are adopting AI steadily but carefully because of security, accuracy, confidentiality, oversight, and skills gaps. | 中 | SU014 |
| CU028 | ACCA says accounting firms need governance, client-disclosure discipline, confidentiality controls, and vendor diligence before purchasing or rolling out AI tools. | 中 | SU016 |
| CU029 | Journal of Accountancy says accountants face hallucination, black-box trust, prompt-injection, privacy, and shadow-AI risks when deploying AI tools. | 中 | SU017 |
| CU030 | IBM says enterprise AI adoption stalls on fragmented data, weak governance, ROI proof gaps, skills shortages, and workflow-integration problems. | 中 | SU018 |
| CU031 | Basis’s home page says accountants shift from doing the work to reviewing it, with fewer review cycles and more time for higher-value work. | 中 | SU001 |
| CU032 | Basis markets its value as expanded revenue capacity for firms that are stretched thin on experienced accounting labor. | 中 | SU001, SU021 |
| CU033 | BusinessWire, CPA Practice Advisor, and Financial IT all frame accounting labor shortages and rising workload pressure as core demand drivers for Basis customer adoption. | 中 | SU009, SU010, SU012 |
| CU034 | Financial IT says several top accounting firms participated in Basis’s seed round, suggesting early customer or practitioner alignment even before the later named-account announcements. | 中 | SU012 |
| CU035 | Basis’s public materials and repeat coverage consistently frame the product as agents that complete work and return finished output for human review rather than as a fully unreviewed autopilot. | 中 | SU005, SU009, SU010 |
| CU036 | No retained public source discloses GRR, NRR, churn, renewal rate, or cohort retention for Basis customers. | 低 | SU001, SU005, SU006, SU007, SU009 |
| CU037 | The public record supports named deployments and positive testimonials more strongly than it supports durable renewal evidence. | 中 | SU001, SU006, SU007, SU013 |
| CU038 | Customer-proof quality is strongest for Wiss, Baker Tilly, and UHY, while Clark Nuber, Boulay, and BPB remain mostly testimonial-level references. | 中 | SU001, SU006, SU007, SU013 |
| CU039 | The visible commercial motion is firm-led enterprise adoption rather than individual self-serve purchasing by accountants. | 中 | SU001, SU006, SU007 |
| CU040 | The primary users appear to be accountants and delivery teams inside the firm, while the ultimate beneficiaries include the firms’ end-clients who receive faster closes and better visibility. | 中 | SU005, SU007, SU008 |
| CU041 | The public payer signal points to the accounting firm itself because the named voices are firm executives, partners, CIOs, and managed-services sponsors rather than end-client buyers. | 中 | SU006, SU007, SU013 |
| CU042 | Basis’s expansion path appears to be workflow breadth and adjacent practice areas inside existing logos, not a consumer-style seat ladder. | 中 | SU001, SU005, SU022 |
| CU043 | Reaching about 30% of the top 25 firms is commercially attractive, but it also raises the possibility that revenue concentration could be meaningful if a few large logos dominate spending. | 中 | SU009, SU010, SU011 |
| CU044 | The top-25 penetration claim still lacks a public denominator for customer count, ARR share, deployment depth, or exact logo list. | 中 | SU009, SU010, SU011 |
| CU045 | Independent accounting-industry sources suggest that even supportive customers will need approved use cases, training, governance, and human-review controls before scaling AI broadly. | 高 | SU014, SU015, SU016, SU018 |
| CU046 | Basis’s product language around trust, transparency, verification, audit trails, and reviewability indicates those objections are central to winning customer adoption. | 高 | SU003, SU021, SU022 |
| CU047 | Basis does not publicly disclose customer counts, contract duration, satisfaction scores, or concentration by ARR in the retained source set. | 低 | SU001, SU002, SU007, SU009 |
| CU048 | The freshest customer proof in the retained set is Baker Tilly’s 2026 collaboration and OpenAI’s 2026 case study, while the Wiss deployment proof is older but still substantive. | 中 | SU005, SU006, SU007, SU013 |
| CU049 | Baker Tilly broadens the Basis story from outsourced-accounting and CAS use cases into a wider accounting-and-finance managed-services motion. | 中 | SU007 |
| CU050 | OpenAI’s case study describes Basis automating reconciliations, journal entries, and financial summaries while preserving reviewable explanations for accountants. | 中 | SU005 |
| CR001 | Basis publicly markets one platform across CAS, tax, audit, and advisory. | 中 | SR049 |
| CR002 | Basis’s public multi-practice positioning enlarges the company’s regulatory and execution surface area because one platform touches several accounting workflows. | 中 | SR049, SR003 |
| CR003 | Basis says it is SOC 2 Type II, ISO 27001, and ISO 42001 certified. | 中 | SR001 |
| CR004 | Basis’s security page still lists GDPR and CCPA as coming soon. | 中 | SR001 |
| CR005 | Basis claims customer data never trains models and that agent actions are logged and auditable. | 中 | SR001 |
| CR006 | Basis’s terms say AI outputs are not a substitute for professional accounting, tax, audit, legal, or financial advice. | 中 | SR003 |
| CR007 | Basis requires users to review, verify, and validate AI outputs before relying on them or using them in work product or filings. | 中 | SR003 |
| CR008 | Basis disclaims that AI outputs will be accurate, complete, current, or error-free and disclaims liability for decisions based on them. | 中 | SR003 |
| CR009 | Basis’s public privacy policy says enterprise-product privacy terms live in separate written agreements, so public website terms do not show customer-specific data-processing protections. | 中 | SR002 |
| CR010 | Journal of Accountancy warns that entering confidential client data into generative AI means sharing that data with the tool owner and creates breach and liability risk. | 中 | SR006 |
| CR011 | Journal of Accountancy says generative AI can misapply tax laws, professional standards, and financial-reporting frameworks, creating professional-liability risk if human skepticism is removed. | 中 | SR006 |
| CR012 | Journal of Accountancy says hallucinations are common and that humans should fact-check source citations rather than trust polished answers. | 中 | SR007 |
| CR013 | Journal of Accountancy says prompt injection, hacker enablement, privacy clauses, and shadow AI all add operational risk for accounting firms. | 中 | SR007 |
| CR014 | CPAI says the legal environment for CPA use of generative AI is patchy and can implicate GLBA, FTC safeguards, state privacy laws, and other disclosure obligations. | 中 | SR008 |
| CR015 | CPAI says failure to review AI-assisted work and failure to disclose AI use can support negligence or trust-damage arguments if clients suffer a loss. | 中 | SR008 |
| CR016 | CPAI explicitly points firms to Section 7216 and related regulations when tax return information may be disclosed to or used within AI-enabled workflows. | 中 | SR008, SR053 |
| CR017 | IRS Publication 4557 says protecting taxpayer data is the law and that tax preparers must create and enact security plans under the FTC Safeguards Rule. | 中 | SR011 |
| CR018 | IRS Publication 4557 tells firms to use MFA, encryption, backups, limited access, and audit trails when handling taxpayer data. | 中 | SR011 |
| CR019 | The IRS AI governance manual treats AI affecting privacy or legally/materially significant outcomes as high-impact and subject to minimum risk management practices. | 中 | SR015 |
| CR020 | PCAOB AS 1105 requires auditors to obtain sufficient appropriate audit evidence to support an opinion. | 中 | SR016 |
| CR021 | PCAOB AS 1105 says evidence from knowledgeable independent sources is more reliable than evidence from internal company sources. | 中 | SR016 |
| CR022 | PCAOB AS 1105 says auditors using company-provided electronic information should evaluate its reliability and the controls around receiving, maintaining, and processing it. | 中 | SR016 |
| CR023 | SEC cyber rules require disclosure of material cyber incidents and annual disclosure of cyber-risk management, strategy, and governance processes. | 高 | SR019, SR020 |
| CR024 | NIST says the AI RMF is designed to manage risks to individuals, organizations, and society and that the GenAI profile addresses risks unique to generative AI. | 中 | SR025 |
| CR025 | NIST says it promotes a risk-based approach to maximize AI benefits while minimizing negative consequences and to build trust through standards and evaluation. | 中 | SR052 |
| CR026 | NIST SP 800-53 includes audit/accountability, incident-response, privacy, and supply-chain controls relevant to accounting AI vendors. | 中 | SR028 |
| CR027 | FTC AI enforcement actions against false AI-generated reviews and AI-powered earning claims show active regulatory scrutiny of AI claims and output misuse. | 中 | SR029 |
| CR028 | Grand View says finance is one of the fastest-growing ERP functions and that rising cybersecurity needs are pushing finance teams toward software with stronger security controls. | 中 | SR032 |
| CR029 | MarketsandMarkets projects rapid cloud ERP growth but says data security risks and vendor-selection complexity remain important restraints and challenges. | 中 | SR033 |
| CR030 | IDC treats AI-enabled finance and accounting applications as a distinct evaluated market, highlighting the importance of integration and real-time insight. | 中 | SR034 |
| CR031 | Workiva AI markets AI for financial reporting, audit, risk, and IPO/SOX readiness, which overlaps with Basis’s aspiration to sit inside high-stakes accounting workflows. | 中 | SR041 |
| CR032 | Workiva says its AI currently uses frontier models from Google, AWS/Anthropic, and Microsoft/OpenAI, indicating that enterprise accounting AI still depends on external model ecosystems. | 中 | SR041 |
| CR033 | Workiva’s 10-K says competition includes diversified enterprise vendors, niche providers, GRC vendors, and professional-services alternatives, and expects competition to intensify. | 中 | SR046 |
| CR034 | Thomson Reuters markets CoCounsel Audit, CoCounsel Tax, Guided Assurance, and citation-backed AI, intensifying compliance-focused competition around Basis. | 中 | SR050 |
| CR035 | Intuit Assist pushes generative AI across QuickBooks and tax preparation, giving Intuit a distribution advantage in SMB and midmarket accounting automation. | 中 | SR043 |
| CR036 | Basis’s public customer proof is strong but concentrated in a relatively small set of named firms and partnerships rather than a disclosed broad customer base. | 中 | SR004, SR049 |
| CR037 | Basis’s public materials do not disclose workflow-level accuracy, hallucination escape rates, exception rates, or rework metrics for tax or audit outputs. | 中 | SR001, SR039 |
| CR038 | Basis’s main security and legal pages do not publicly identify production model providers or explain model redundancy and fallback architecture. | 中 | SR001, SR003 |
| CR039 | Basis’s own blog index highlights OpenAI/Codex and trusted-agent narratives, showing that frontier-model progress is part of the company’s public positioning. | 中 | SR039 |
| CR040 | Basis’s security page says the SOC 2 report is available under NDA and that additional compliance documentation is available only on request. | 中 | SR001 |
| CR041 | Because public website terms defer enterprise platform specifics to private written agreements, investors cannot judge indemnity, audit rights, DPA scope, or liability allocation from public artifacts alone. | 中 | SR002, SR003 |
| CR042 | The combination of Basis’s output disclaimers with CPA, IRS, and PCAOB review obligations means residual liability stays with the licensed firm if controls fail. | 高 | SR003, SR006, SR008, SR011, SR016, SR053 |
| CR043 | Basis’s multi-practice deployment model creates failure-transmission risk because one output or control weakness can spread across tax, audit, CAS, and advisory workflows. | 中 | SR003, SR049 |
| CR044 | Public mitigants are real—certifications, no-training claims, audit logs, human-review rules, and standards-aligned governance—but they lower rather than eliminate residual risk. | 高 | SR001, SR003, SR025, SR028 |
| CR045 | Public evidence supports a high residual risk rating for Basis until private diligence resolves accuracy metrics, model redundancy, contract allocation, and concentration questions. | 中 | SR001, SR003, SR039, SR041, SR050 |
| CV001 | Basis announced a $100 million Series B financing in February 2026. | 高 | SV001, SV002, SV004 |
| CV002 | Official and independent coverage place Basis's February 2026 valuation at about $1.15 billion. | 高 | SV001, SV002, SV003, SV004, SV007, SV008 |
| CV003 | Basis had raised roughly $138 million in total disclosed funding by the February 2026 Series B. | 中 | SV002, SV008 |
| CV004 | Accel led the Series B and GV, Khosla Ventures, and Lloyd Blankfein participated. | 高 | SV002, SV004, SV007 |
| CV005 | Miles Clements joined Keith Rabois in Basis's board-level investor oversight at the Series B. | 中 | SV004, SV008 |
| CV006 | Basis says it works with about 30% of the top 25 accounting firms. | 高 | SV002, SV004, SV008 |
| CV007 | Tech Funding News reported that Basis also says it reaches about 20% of the top 150 U.S. accounting firms. | 中 | SV007 |
| CV008 | Basis said the $100 million round would fund platform development plus engineering and machine-learning hiring. | 中 | SV002, SV008, SV009 |
| CV009 | OpenAI says firms using Basis report about 30% time savings on average. | 中 | SV006 |
| CV010 | Baker Tilly said its Basis collaboration extends AI-enabled automation across accounting and finance managed services. | 中 | SV005 |
| CV011 | Basis publicly frames itself as a long-horizon agent platform spanning accounting, tax, and audit workflows. | 高 | SV001, SV004, SV010 |
| CV012 | Numeric raised a $51 million Series B in November 2025 led by IVP. | 高 | SV011, SV013 |
| CV013 | Numeric's total funding reached about $89 million after the Series B. | 中 | SV013 |
| CV014 | Numeric is pushing from close management into a broader finance-operations platform with cash management. | 中 | SV011, SV013 |
| CV015 | Numeric says its cash-management product reaches a 90%-plus auto-match rate. | 中 | SV011, SV013 |
| CV016 | TechCrunch reported that Numeric raised a $28 million Series A in October 2024 after a $10 million seed five months earlier. | 中 | SV012 |
| CV017 | Truewind raised a $13 million Series A and brought total funding to more than $17 million. | 高 | SV014, SV015 |
| CV018 | Truewind says revenue grew fourfold over the prior year. | 中 | SV014, SV015 |
| CV019 | Truewind says its product absorbs 47% of month-end close tasks. | 中 | SV014, SV015 |
| CV020 | Ramp raised $750 million at a $44 billion valuation in June 2026. | 高 | SV016, SV017 |
| CV021 | Ramp said annualized revenue was above $1 billion and TechCrunch cited Bloomberg for more than $1.5 billion of run-rate revenue. | 中 | SV016, SV017 |
| CV022 | Ramp reported more than 70,000 customers at the time of the June 2026 financing. | 高 | SV016, SV017 |
| CV023 | Ramp had raised more than $3 billion in total equity financing by June 2026. | 高 | SV016, SV017 |
| CV024 | CNBC reported Ramp was valued at $16 billion in June 2025, showing a rapid step-up to $44 billion one year later. | 中 | SV016, SV018 |
| CV025 | Workiva generated $738.7 million of revenue in 2024. | 中 | SV019 |
| CV026 | Approximately 90% of Workiva's 2024 revenue came from subscription and support fees. | 中 | SV019 |
| CV027 | Yahoo Finance showed Workiva at about 3.03x price-to-sales and 2.91x EV-to-revenue on 2026-06-11. | 中 | SV020 |
| CV028 | BILL generated about $1.463 billion of revenue in fiscal 2025. | 中 | SV021 |
| CV029 | BILL reported an 81.4% GAAP gross margin and an 85.0% non-GAAP gross margin in fiscal 2025. | 中 | SV021 |
| CV030 | Yahoo Finance showed BILL at about 2.08x price-to-sales and 1.84x EV-to-revenue on 2026-06-11. | 中 | SV022 |
| CV031 | BlackLine reported a 75.2% GAAP gross margin and a 79.3% non-GAAP gross margin in 2024. | 中 | SV023 |
| CV032 | Yahoo Finance showed BlackLine at about 2.59x price-to-sales and 2.49x EV-to-revenue on 2026-06-11. | 中 | SV024 |
| CV033 | Yahoo Finance showed Intuit at about 3.80x price-to-sales and 3.64x EV-to-revenue on 2026-06-11. | 中 | SV025 |
| CV034 | Aventis Advisors said the median public SaaS EV-to-revenue multiple stood at about 3.4x as of March 2026. | 中 | SV026 |
| CV035 | Aventis Advisors said median private SaaS M&A revenue multiples were about 3.1x as of March 2026. | 中 | SV026 |
| CV036 | Aventis Advisors said 2026 investors reward operational efficiency and meaningful AI integration rather than simple AI wrappers. | 中 | SV026 |
| CV037 | Aventis Advisors said AI-disruption fears and post-2021 rate discipline materially compressed software valuation multiples. | 中 | SV026 |
| CV038 | COSO guidance says generative-AI adoption creates new risks that require disciplined oversight and audit-ready control mapping. | 中 | SV027 |
| CV039 | Journal of Accountancy warned that generative AI can hallucinate facts and present flawed output in an authoritative tone. | 中 | SV028 |
| CV040 | Journal of Accountancy said AI in auditing raises security, confidentiality, privacy, transparency, and bias risks that still require human judgment. | 中 | SV028 |
| CV041 | PCAOB said current generative-AI use in audits is focused mainly on administrative and research activities. | 中 | SV029 |
| CV042 | PCAOB said firms acknowledged generative-AI limitations and the need for strong supervision to guard against data privacy and security risks. | 中 | SV029 |
| CV043 | NIST's AI Risk Management Framework exists to improve how organizations manage AI risks to people, organizations, and society. | 中 | SV030 |
| CV044 | At a $1.15 billion valuation, Basis would trade at roughly 57.5x ARR if ARR were only $20 million. | 中 | SV002, SV026 |
| CV045 | At a $1.15 billion valuation, Basis would trade at roughly 38.3x ARR if ARR were $30 million. | 中 | SV002, SV026 |
| CV046 | At a $1.15 billion valuation, Basis would trade at roughly 23.0x ARR if ARR were $50 million. | 中 | SV002, SV026 |
| CV047 | At a $1.15 billion valuation, Basis would trade at roughly 15.3x ARR if ARR were $75 million. | 中 | SV002, SV026 |
| CV048 | At a $1.15 billion valuation, Basis would trade at roughly 11.5x ARR if ARR were $100 million. | 中 | SV002, SV026 |
| CV049 | Matching the March 2026 public SaaS median of 3.4x would require roughly $338 million of revenue or ARR at a $1.15 billion value. | 中 | SV002, SV026 |
| CV050 | Matching the March 2026 private SaaS M&A median of 3.1x would require roughly $371 million of revenue or ARR at a $1.15 billion value. | 中 | SV002, SV026 |
| CV051 | Even a 10x to 15x premium software multiple would still require roughly $77 million to $115 million of ARR to support a $1.15 billion mark. | 中 | SV002, SV026 |
| CV052 | If the quoted $1.15 billion valuation is treated as post-money, a $100 million primary round implies about 8.7% new-money dilution. | 中 | SV002, SV004 |
| CV053 | Basis's roughly $138 million of total raised equals about 12% of the current $1.15 billion mark, which looks capital-efficient on paper. | 中 | SV002, SV008 |
| CV054 | Public evidence still does not disclose Basis ARR, revenue, burn, or net retention well enough to validate the current valuation. | 中 | SV001, SV002, SV004, SV009, SV010 |
| CV055 | Relevant public comps cluster around roughly 2.1x to 3.8x sales, well below the premium multiple Basis would need unless ARR is already far above the public record. | 中 | SV020, SV022, SV024, SV025, SV026 |
| CV056 | Basis's downside is amplified by AI-governance and verification risk because tax and audit workflows tolerate less error than generic productivity software. | 中 | SV027, SV028, SV029, SV030 |
| CV057 | The upside case depends on Basis turning top-firm penetration and partner proof into disclosed software-like ARR with durable margins, not just workflow demos. | 中 | SV005, SV006, SV026 |
| CV058 | The base case is that Basis deserves a private-growth premium to public comps, but not enough to remove the need for hard diligence on ARR, retention, services mix, and preference stack. | 中 | SV002, SV016, SV026 |
| CV059 | A downside case of $25 million to $40 million of ARR at 8x to 12x revenue implies roughly $200 million to $480 million of value, materially below the February 2026 mark. | 低 | SV002, SV026, SV028, SV029 |
| CV060 | A base case of $50 million to $75 million of ARR at 12x to 16x revenue implies roughly $600 million to $1.2 billion of value, which only brackets the current mark if Basis is already at meaningful scale. | 低 | SV002, SV026 |
| CV061 | An upside case of $80 million to $120 million of ARR at 15x to 20x revenue implies roughly $1.2 billion to $2.4 billion of value and requires strong growth plus high-quality economics. | 低 | SV002, SV006, SV026 |
| CV062 | At the February 2026 mark, the most supportable recommendation is research-more rather than buy because price cannot yet be validated on public evidence alone. | 中 | SV002, SV020, SV022, SV024, SV025, SV026, SV028, SV029 |
| CV063 | The current valuation stance is stretched rather than obviously broken because partner proof and top-firm penetration could still justify a premium if economics are later disclosed. | 中 | SV005, SV006, SV026 |
| 编号 | 出版方 | 标题 | 引文 |
|---|---|---|---|
| SO001 | Basis | Basis | Agents built specifically for accountants. |
| SO002 | Basis | Basis | About | |
| SO003 | Basis | Basis | Security | We’re SOC 2 Type II, ISO 27001, and ISO 42001 certified for Security, Availability, and Confidentiality. |
| SO004 | Basis | Basis | Careers | We're deployed today with the top accounting firms in the country. |
| SO005 | Basis | Privacy Policy | This Privacy Policy describes how Essex Labs, Inc. dba Basis ("Basis," "we," "us," "our") collects, uses, and discloses personal information. |
| SO006 | Basis | Basis Raises $3.6m with an AI Platform for Accounting Firms | |
| SO007 | Basis | Basis Raises $34m Series A Led by Khosla Ventures | |
| SO008 | Basis | Basis raises $100M Series B at a $1.15B valuation led by Accel alongside GV (Google Ventures) | |
| SO009 | Basis | Wiss Announces Strategic Partnership with Accounting AI Platform Basis | |
| SO010 | Basis | Baker Tilly and Basis Announce Collaboration to Expand AI-Powered Automation Across Thousands of Accounting Clients | |
| SO011 | Basis | Basis Scales Accounting by Turning OpenAI Model Progress into Trusted Agents | |
| SO012 | Basis | Seven Questions with Annette Garcia, Accounting Product Operations Lead | |
| SO013 | Business Wire | Basis Raises $100M at a $1.15B Valuation as Accounting Firms Adopt End-to-End Agents Across Accounting, Tax and Audit | |
| SO014 | CPA Practice Advisor | Basis Raises $100 Million to Deploy AI Agents for Accounting Firms | |
| SO015 | PYMNTS | Basis Raises $100 Million to Build Up AI in Accounting | |
| SO016 | Built In NYC | Basis Raises $100M Series B Round | |
| SO017 | Fintech News Switzerland | Basis Raises US$100m in Series B Funding at a US$1.15 Billion Valuation | |
| SO018 | Crowdfund Insider | Basis Announces $100M in New Funding at $1.15B Valuation to Enable AI-Driven Accounting Automation | |
| SO019 | PR Newswire | Basis Secures $3.6 Million to Bring AI to Accounting Firms | |
| SO020 | Wiss | Wiss Announces Strategic Partnership with Accounting AI Platform Basis | |
| SO021 | U.S. Securities and Exchange Commission | SEC Adopts Rules on Cybersecurity Risk Management, Strategy, Governance, and Incident Disclosure by Public Companies | |
| SO022 | National Institute of Standards and Technology | AI Risk Management Framework | The NIST AI Risk Management Framework (AI RMF) is intended for voluntary use and to improve the ability to incorporate trustworthiness considerations into the design, development, use, and evaluation of AI products, services, and systems. |
| SO023 | Thomson Reuters Institute | 2026 AI in Professional Services Report | |
| SO024 | U.S. Bureau of Labor Statistics | Accountants and Auditors | |
| SO025 | Basis | OpenAI (Codex) Customer Spotlight: Interview with Mitch Troyanovsky | |
| SO026 | citybiz | Basis Raises $34M To Advance AI Agents For Accounting | Founded by researchers Zenna Tavares, Emily Mackevicius and Eli Bingham, Basis is led by the fourth co-founder, Matt Harpe, a former Boston Consulting Group and Softbank executive. |
| SO027 | Zenna Tavares | Zenna Tavares | I was previously the inaugural Alan Kanzer Innovation Scholar at Columbia University's Zuckerman Institute and Data Science Institute. |
| SO028 | AI Market Watch | Basis | Co-founder & CEO at Basis (Jan 2023-Present). Previously Consultant at Boston Consulting Group (BCG) from 2018-2021. |
| SM001 | Basis | Basis homepage | |
| SM002 | IBISWorld | Accounting Services in the US - Market Size | |
| SM003 | Mordor Intelligence | Accounting Software Market Analysis | Industry Growth, Size & Forecast Report | |
| SM004 | Thomson Reuters Institute | 2026 AI in Professional Services Report | |
| SM005 | U.S. Bureau of Labor Statistics | Accountants and Auditors | |
| SM006 | National Pipeline Advisory Group | Advisory panel on accounting talent shortage releases final report | |
| SM007 | NYSSCPA | In Report, AICPA Advisory Group Offers Recommendations to Solve Talent Shortage | |
| SM008 | The CPA Journal | Intergenerational Solutions to Address the Crisis of the Leaking Accounting Pipeline | |
| SM009 | Intuit / SEC filings | Annual Reports | |
| SM010 | Intuit | Intuit Reports Strong Second-Quarter Results and Reiterates Full-Year Guidance | |
| SM011 | Intuit QuickBooks | Intuit Accountant Suite | |
| SM012 | Sage | Accountants and Bookkeeping Software | Sage US | |
| SM013 | Xero | Xero accounting software for U.S. small businesses | |
| SM014 | Botkeeper | Botkeeper | Bookkeeping for Accounting Firms | |
| SM015 | Pilot | Pilot Pricing: Transparent Bookkeeping, Tax, and CFO plans for businesses | |
| SM016 | Numeric | Numeric | AI-Powered Close Automation | |
| SM017 | Digits | Digits - AI-Native Accounting Software | |
| SM018 | CPA Practice Advisor | Basis Raises $100 Million to Deploy AI Agents for Accounting Firms | |
| SM019 | PYMNTS | Basis Raises $100 Million to Build Up AI in Accounting | |
| SM020 | PCAOB | Docket 052: Amendments Related to Aspects of Designing and Performing Audit Procedures that Involve Technology-Assisted Analysis of Information in Electronic Form | |
| SM021 | AnythingResearch | Accounting, Tax Preparation, Bookkeeping, and Payroll Services Industry Statistics | |
| SM022 | Kentley Insights | Accounting Services industry market research report | |
| SM023 | The Business Research Company | Accounting Software Global Market Report 2026 | |
| SM024 | Precedence Research | Accounting Software Market | |
| SM025 | The Business Research Company | Accounting Services Global Market Report | |
| SM026 | U.S. Bureau of Labor Statistics | Employment Cost Index – March 2026 | |
| SM027 | Wiss | Wiss Announces Strategic Partnership with Accounting AI Platform Basis | |
| SP001 | Basis | Basis home page | Basis runs in the background, executing accounting workflows end-to-end and updating you at key stages. |
| SP002 | Basis | Basis security page | Your data never trains or improves any AI model. |
| SP003 | OpenAI | Basis scales accounting by turning OpenAI model progress into trusted agents | Today, Basis supports a significant share of large accounting firms across the U.S. |
| SP004 | Wiss | Wiss Announces Partnership with Basis Accounting AI Platform | We are a Microsoft firm – we have a private Wiss ChatGPT instance with firm data and have found Copilot useful – but Basis’ capability for integrated, end-to-end automation is transformative. |
| SP005 | Crowdfund Insider | Basis Announces $100M In New Funding At $1.15B Valuation To Enable AI-Driven Accounting Automation | The company already partners with approximately 30 percent of the top 25 U.S. accounting firms, delivering measurable productivity lifts estimated at 20 to 50 percent in key areas. |
| SP006 | Botkeeper | Botkeeper home page | It's time to experience the ONLY AI Accounting solution purpose-built for Accounting Firms. |
| SP007 | Botkeeper | Botkeeper pricing page | 1 - 4 Licenses $149 per license/mo ... 25+ Licenses $59 per license/mo. |
| SP008 | Pilot | Pilot home page | We handle bookkeeping, tax, and advisory work for startups and small businesses. It's 50/50 people and software. |
| SP009 | Pilot | Pilot pricing page | Get expert bookkeeping, tax, and CFO services with Pilot, starting at $99/month. |
| SP010 | Vic.ai | Vic.ai home page | Vic.ai delivers high-fidelity AP data, reducing errors, accelerating approvals, and optimizing financial operations at scale. |
| SP011 | Vic.ai | Vic.ai accounts payable page | Vic.ai’s AI-first platform automates your entire AP process—eliminating inefficiencies, reducing errors, and delivering real-time insights. |
| SP012 | Vic.ai | Vic.ai Trust Center | Vic.ai holds SOC 1 Type II and SOC 2 Type II certifications, renewed annually and audited by third-party assessors. |
| SP013 | Digits | Digits home page | Digits is the first AI-native general ledger for business owners and accountants. |
| SP014 | Digits | Digits pricing page | Essentials $65/mo ... Core $100/mo ... Pro $250/mo. |
| SP015 | Digits | Digits for Firms page | The only general ledger with the Agentic Close, trained to serve your clients without scaling headcount. |
| SP016 | Truewind | Truewind home page | Keep Sage or QBO as your source of truth. Truewind turns bank statements and workpapers into GL-ready journal entries, reconciliations, and SOPs you can review. |
| SP017 | QuickBooks | Automate your path to growth with Intuit Assist for QuickBooks | Intuit Assist and certain other AI features and functionalities are currently available at no additional cost to certain QuickBooks users. |
| SP018 | Microsoft | Microsoft 365 Copilot home page | Add agents to Copilot that automate common tasks or work on your behalf. |
| SP019 | Microsoft | Microsoft 365 Copilot for Business | Copilot Chat is available at no additional cost for all Microsoft Entra account users with an eligible Microsoft 365 subscription. |
| SP020 | BILL | BILL accounts payable page | Automate your accounts payable process with the best accounts payable software to pay your business invoices online. |
| SP021 | Thomson Reuters Institute | 2026 AI in Professional Services Report | Today marks the strategic phase of AI, in which organizations redefine workflows, reshape value, and build AI directly into the foundation of their business strategy. |
| SP022 | PCAOB | Staff Update on Outreach Activities Related to the Integration of Generative Artificial Intelligence in Audits and Financial Reporting | The current integration of GenAI in audits ... appears to be focused primarily on administrative and research activities. |
| SP023 | NIST | AI Risk Management Framework | The NIST AI Risk Management Framework is intended ... to improve the ability to incorporate trustworthiness considerations into the design, development, use, and evaluation of AI products, services, and systems. |
| SP024 | Mordor Intelligence | AI in Accounting Market Analysis | Industry Report, Size & Forecast Insights | The AI In Accounting Market worth USD 10.87 billion in 2026 is growing at a CAGR of 44.6% to reach USD 68.75 billion by 2031. |
| SP025 | Baker Tilly | Baker Tilly collaborates with Basis to expand AI-enabled automation across accounting and finance managed services | Baker Tilly ... announced a collaboration with Basis to expand AI-enabled automation across accounting and finance managed services. |
| SI001 | Basis | Basis | Firms are stretched thin, with growing client demands, tightening margins, and not enough experienced hands to go around. |
| SI002 | Basis | Basis | About | |
| SI003 | Basis | Basis | Security | |
| SI004 | Basis | Basis | Careers | We're deployed today with the top accounting firms in the country. |
| SI005 | Basis | Privacy Policy | |
| SI006 | Basis | Terms of Service | If you and Basis have executed a written agreement governing your access to and use of the Services (such as a Master Services Agreement or Order Form), the terms of that signed agreement will control. |
| SI007 | Basis | Basis raises $100M Series B at a $1.15B valuation led by Accel alongside GV (Google Ventures) | |
| SI008 | Basis | Basis raises a $34m Series A led by Khosla Ventures | |
| SI009 | PR Newswire | Basis secures $3.6 million to bring AI to accounting firms | |
| SI010 | Wiss | Wiss Announces Strategic Partnership with Accounting AI Platform Basis | Wiss began testing Basis in beta for their Client Accounting Services (CAS) practice in November 2023. The successful results have led to a full deployment across the practice. |
| SI011 | Baker Tilly | Baker Tilly collaborates with Basis to expand AI-enabled automation across accounting and finance managed services | |
| SI012 | Business Wire | Basis Raises $100M at a $1.15B Valuation as Accounting Firms Adopt End-to-End Agents Across Accounting, Tax, and Audit | Basis is now working with ~30% of the Top 25 accounting firms. |
| SI013 | CPA Practice Advisor | Basis Raises $100 Million to Deploy AI Agents for Accounting Firms | |
| SI014 | Built In NYC | Basis Raises $100M Series B | |
| SI015 | PYMNTS | Basis Raises $100 Million to Build Up AI in Accounting | |
| SI016 | Crowdfund Insider | Basis announces $100M in new funding at $1.15B valuation to enable AI-driven accounting automation | |
| SI017 | AI2.work | Basis Raises $100M at $1.15B Valuation to Automate Accounting (2026) | |
| SI018 | OpenAI | Basis | Firms using Basis report 30% time savings on average. |
| SI019 | U.S. Bureau of Labor Statistics | Accountants and Auditors | |
| SI020 | Thomson Reuters | 2026 AI in Professional Services Report | |
| SI021 | Journal of Accountancy | AI risks CPAs should know | Approach generative AI responses with the same skepticism you would apply to any tax or audit situation. |
| SI022 | National Institute of Standards and Technology | AI Risk Management Framework | |
| SI023 | Securities and Exchange Commission | BILL HOLDINGS, INC. 10-K | |
| SI024 | Securities and Exchange Commission | Intuit Inc. 10-K | |
| SI025 | Securities and Exchange Commission | Workiva Inc. 10-K | |
| SI026 | Washington Society of CPAs / CAMICO | Generative Artificial Intelligence: FAQ on CAMICO's Advisory Hotline | |
| SI027 | Uniqus | Internal Controls Over Generative AI: Reassessing SOX Compliance, ITGC, Access Controls, and Segregation of Duties in the Age of AI | |
| SE001 | Basis | Basis homepage | Basis runs in the background, executing accounting workflows end-to-end and updating you at key stages. |
| SE002 | Basis | Basis About | We build agents that do real accounting work, end to end. |
| SE003 | Basis | Basis Security | Every agent action is logged and auditable. |
| SE004 | Basis | Basis Privacy Policy | This Privacy Policy applies only to personal information that we collect when you access or use our Website. It does not apply to personal information that we receive when you use our enterprise products and services. |
| SE005 | Basis | Basis Careers | We're deployed today with the top accounting firms in the country, doing real, economically productive work for our customers. |
| SE006 | Basis | Seven Questions with Annette Garcia, Accounting Product Operations Lead | There is no reason why someone cannot go to a regular chatbot and ask for accounting work to be done. But accountants like things done in a particular way. |
| SE007 | Basis | Basis Scales Accounting by Turning OpenAI Model Progress into Trusted Agents | Each task begins with a supervising agent ... which coordinates the full process—routing steps to specialized sub-agents based on task, complexity, latency needs, and input type. |
| SE008 | Basis | Basis Blog & News | Basis Updates |
| SE009 | Basis | Building a Company for the AGI Era | At Basis we’ve started to treat our context with the same rigor we treat code: version control, clear content ownership, automated reviews, deployment pipelines etc. |
| SE010 | Basis | Your team needs a unified MCP. Here’s a recipe. | Centralized policy gives us an enforceable security posture instead of a hopeful one. |
| SE011 | Basis | How We Made Our Monorepo Ergonomic for Agents | Interoperability. No layer binds the team to a single vendor. |
| SE012 | Basis | We're Hiring Agent Managers | We’re moving from getting a model to do a task, to managing a team of models to run a department. |
| SE013 | Business Wire | Basis Raises $100M at a $1.15B Valuation as Accounting Firms Adopt End-to-End Agents Across Accounting, Tax and Audit | Basis’ approach entails building “long-horizon” agents that autonomously work on complex accounting workflows over many hours. |
| SE014 | CPA Practice Advisor | Basis Raises $100 Million to Deploy AI Agents for Accounting Firms | These agents operate continuously in the background, coordinating tasks across accounting processes and returning completed deliverables for review. |
| SE015 | Wiss | Wiss announces strategic partnership with accounting AI platform Basis | We are a Microsoft firm – we have a private Wiss ChatGPT instance with firm data and have found Copilot useful – but Basis’ capability for integrated, end-to-end automation is transformative. |
| SE016 | Baker Tilly | Baker Tilly collaborates with Basis | Baker Tilly collaborates with Basis |
| SE017 | OpenAI | Basis | Basis treats accounting as a system of workflows, each with its own context and complexity. |
| SE018 | OpenAI | Building self-improving tax agents with Codex | The product has to capture the full path from source material, to extracted fields and provenance, to downstream submission and expert correction. |
| SE019 | Intuit Developer | QuickBooks Online Accounting API overview | The QuickBooks Online Accounting API lets your apps utilize key features and data in QuickBooks Online. |
| SE020 | Xero Developer | OAuth 2.0 — Xero Developer | OAuth 2.0 — Xero Developer |
| SE021 | QuickBooks | Automate your path to growth with Intuit Assist for QuickBooks | Intuit Assist will help your business thrive by automating tasks and workflows ... and answering your pressing financial questions on the spot. |
| SE022 | Microsoft | Microsoft 365 Copilot | Add agents to Copilot that automate common tasks or work on your behalf. |
| SE023 | Botkeeper | Botkeeper | Bookkeeping for Accounting Firms | Our AI posts directly to the GL only when confidence is high—delivering 97% accuracy on those entries. |
| SE024 | Vic.ai | AP Automation Software | AI-First Invoice Processing | Vic.ai | Vic.ai launches VicPay, Vendor Portal, and VicAgents, introducing agentic AI partners for enterprise finance. |
| SE025 | Digits | Digits - AI-Native Accounting Software | Digits is the first AI-native general ledger for business owners and accountants. |
| SE026 | National Institute of Standards and Technology | AI Risk Management Framework | The NIST AI Risk Management Framework is intended ... to improve the ability to incorporate trustworthiness considerations into the design, development, use, and evaluation of AI products, services, and systems. |
| SE027 | Securities and Exchange Commission | SEC adopts rules on cybersecurity risk management, strategy, governance, and incident disclosure | The new rules will require registrants to disclose their processes, if any, for assessing, identifying, and managing material risks from cybersecurity threats. |
| SE028 | Public Company Accounting Oversight Board | PCAOB staff shares observations from outreach on use of generative artificial intelligence in audits and financial reporting | Most audit firms interviewed noted the potential for using GenAI in certain aspects of planning and performing the audit. |
| SE029 | AICPA & CIMA | Artificial intelligence resources | Explore how artificial intelligence is transforming accounting and finance and how AICPA and CIMA resources help accounting and finance professionals build confidence, competence, and credibility in an AI-enabled profession. |
| SE030 | Internal Revenue Service | Due diligence requirements for tax preparers | Your tax preparation software helps, but it cannot replace your professional judgment or responsibility under the law. |
| SE031 | Thomson Reuters Institute | 2026 AI in Professional Services Report | Today marks the strategic phase of AI, in which organizations redefine workflows, reshape value, and build AI directly into the foundation of their business strategy. |
| SU001 | Basis | Basis | |
| SU002 | Basis | Basis | About | |
| SU003 | Basis | Basis | Security | |
| SU004 | Basis | Privacy Policy | |
| SU005 | OpenAI | Basis scales accounting by turning OpenAI model progress into trusted agents | |
| SU006 | Wiss | Wiss Announces Strategic Partnership with Accounting AI Platform Basis | |
| SU007 | Baker Tilly | Baker Tilly collaborates with Basis to expand AI-enabled automation across accounting and finance managed services | |
| SU008 | UHY | Client Accounting Services | UHY | |
| SU009 | Business Wire | Basis Raises $100M at a $1.15B Valuation as Accounting Firms Adopt End-to-End Agents Across Accounting, Tax, and Audit | |
| SU010 | CPA Practice Advisor | Basis Raises $100 Million to Deploy AI Agents for Accounting Firms | |
| SU011 | Our Accounting World | Basis Raises $100M to Expand AI Accounting Platform, Reaches $1.15B Valuation | |
| SU012 | Financial IT | Basis Raises $3.6M with an AI Platform for Accounting Firms | |
| SU013 | Business Wire | Wiss Enters 2026 With Unprecedented Momentum Across Growth, Innovation, and Talent | |
| SU014 | CPA Practice Advisor | AI Adoption in CPA Firms: Benefits, Challenges and Best Practices for Leaders | |
| SU015 | Thomson Reuters | Challenges of adopting AI in accounting firms | |
| SU016 | ACCA Global | Key areas of risk when adopting AI | |
| SU017 | Journal of Accountancy | AI risks CPAs should know | |
| SU018 | IBM | The Biggest AI Adoption Challenges for 2026 | |
| SU019 | Basis | Basis raises $100M Series B at a $1.15B valuation led by Accel alongside GV (Google Ventures) | |
| SU020 | Basis | Basis Raises $34m Series A Led by Khosla Ventures | |
| SU021 | Basis | Basis Raises $3.6m with an AI Platform for Accounting Firms | |
| SU022 | Basis | Seven Questions with Annette Garcia, Accounting Product Operations Lead | |
| SU023 | Basis | Wiss Announces Strategic Partnership with Accounting AI Platform Basis | |
| SU024 | Basis | Baker Tilly and Basis Announce Collaboration to Expand AI-Powered Automation Across Thousands of Accounting Clients | |
| SU025 | Basis | OpenAI (Codex) Customer Spotlight: Interview with Mitch Troyanovsky | |
| SR001 | Basis | Basis | Security | We’re SOC 2 Type II, ISO 27001, and ISO 42001 certified for Security, Availability, and Confidentiality. |
| SR002 | Basis | Privacy Policy | The privacy terms applicable to our enterprise products and services are set forth in separate written agreements between Basis and the applicable customer. |
| SR003 | Basis | Terms of Service | AI Outputs are not a substitute for professional accounting, tax, audit, legal, or financial advice. |
| SR004 | Basis | Baker Tilly and Basis Announce Collaboration to Expand AI-Powered Automation Across Thousands of Accounting Clients | |
| SR006 | Journal of Accountancy | Generative AI and risks to CPA firms | This potential for misinterpretation or misapplication of standards, laws, and regulations can have disastrous consequences. |
| SR007 | Journal of Accountancy | AI risks CPAs should know | Perhaps the most common and well-known risk is hallucinations. |
| SR008 | CPAI | Should I disclose my use of gen AI to clients? | If the disclosure/use involves tax return information, consider Regs. Sec. 301.7216-2 and assess if the disclosure/use requires consent. |
| SR009 | CPAI | Generative AI and risks to CPA firms | |
| SR011 | Internal Revenue Service | Publication 4557, Safeguarding Taxpayer Data | Protecting taxpayer data is the law. |
| SR015 | Internal Revenue Service | IRM 10.24.1 IRS Policy for Artificial Intelligence Governance | High-impact AI is AI with an output that serves as a principal basis for decisions or actions with legal, material, binding, or significant effect. |
| SR016 | PCAOB | AS 1105 Audit Evidence | The auditor must plan and perform audit procedures to obtain sufficient appropriate audit evidence. |
| SR018 | PCAOB | Auditing Standards | |
| SR019 | Securities and Exchange Commission | SEC adopts rules on cybersecurity risk management, strategy, governance, and incident disclosure | The new rules also add Regulation S-K Item 106, which will require registrants to describe their processes for assessing, identifying, and managing material risks from cybersecurity threats. |
| SR020 | Securities and Exchange Commission | Cybersecurity Risk Management, Strategy, Governance, and Incident Disclosure final rule | |
| SR025 | NIST | AI Risk Management Framework | The profile can help organizations identify unique risks posed by generative AI. |
| SR027 | NIST | Cybersecurity Framework | |
| SR028 | NIST | SP 800-53 Rev. 5 Security and Privacy Controls for Information Systems and Organizations | This publication provides a catalog of security and privacy controls for information systems and organizations. |
| SR029 | Federal Trade Commission | Artificial Intelligence | Rytr’s service generated detailed reviews that contained specific, often material details that had no relation to the user’s input. |
| SR032 | Grand View Research | ERP Software Market Summary | The finance segment accounted for a significant revenue share in 2025 and is anticipated to grow at the fastest CAGR during the forecast period. |
| SR033 | MarketsandMarkets | Cloud ERP Market Report 2024-2029 | Data security concerns remain a major barrier to cloud ERP adoption. |
| SR034 | IDC | IDC MarketScape Worldwide AI-Enabled Small Business Finance and Accounting Applications 2025-2026 | The ideal GL must act as the ultimate integration layer, embedding financial logic directly into operational front lines. |
| SR039 | Basis | Basis | Blog & News | Basis Scales Accounting by Turning OpenAI Model Progress into Trusted Agents. |
| SR040 | PCAOB | Staff Publications | Spotlight: Staff Update on Outreach Activities Related to the Integration of Generative Artificial Intelligence in Audits and Financial Reporting. |
| SR041 | Workiva | Workiva AI | Common GRC use cases include IPO & SOX Readiness. |
| SR043 | Intuit | Intuit Assist | Intuit Assist will help customers in each of our products. |
| SR046 | Workiva | Workiva Form 10-K for the year ended December 31, 2024 | As our markets expand, we expect to compete with more highly specialized software vendors, as well as larger vendors that may continue to acquire or bundle their products more effectively. |
| SR049 | Basis | Basis homepage | One platform across your firm. Core modules built for the needs of each practice area. CAS. Tax. Audit. Advisory. |
| SR050 | Thomson Reuters | Artificial intelligence at Thomson Reuters | CoCounsel Audit standardizes audit workflows with AI-powered automation that delivers verifiable, citation-backed insights. |
| SR052 | NIST | Artificial Intelligence at NIST | NIST advances a risk-based approach to maximize the benefits of AI while minimizing its potential negative consequences. |
| SR053 | Legal Information Institute | 26 U.S.C. 7216 Disclosure or use of information by preparers of returns | Any person who knowingly or recklessly discloses return-preparation information or uses it for another purpose shall be guilty of a misdemeanor. |
| SV001 | Basis | Basis raises $100M Series B at a $1.15B valuation led by Accel alongside GV (Google Ventures) | We started Basis with the core belief that agents would become an integral part of knowledge work in the near future, and that accounting would be among the first and most important domains where this happened. |
| SV002 | CPA Practice Advisor | Basis Raises $100 Million to Deploy AI Agents for Accounting Firms | AI accounting startup Basis said Feb. 24 it has raised $100 million in Series B funding ... at a $1.15 billion valuation. |
| SV003 | Crowdfund Insider | Basis Announces $100M In New Funding At $1.15B Valuation To Enable AI-Driven Accounting Automation | |
| SV004 | Business Wire | Basis Raises $100M at a $1.15B Valuation as Accounting Firms Adopt End-to-End Agents Across Accounting, Tax, and Audit | Basis is now working with ~30% of the Top 25 accounting firms. |
| SV005 | Baker Tilly | Baker Tilly collaborates with Basis to expand AI-enabled automation across accounting and finance managed services | |
| SV006 | OpenAI | Basis scales accounting by turning OpenAI model progress into trusted agents | Firms using Basis report 30% time savings on average. |
| SV007 | Tech Funding News | AI accounting startup Basis raises $100M at $1.15B valuation from Accel, GV to automate Big Four drudgery | |
| SV008 | FinTech Futures | Basis raises $100m Series B at $1.15bn valuation | |
| SV009 | Built In NYC | Basis Raises $100M to Deploy Automated Accounting Agents | |
| SV010 | PYMNTS | Basis Raises $100 Million to Build Up AI in Accounting | |
| SV011 | Numeric | Numeric raises $51M Series B, led by IVP, to expand beyond close management | |
| SV012 | TechCrunch | Exclusive: Numeric grabs $28M Series A to automate accounting using AI | |
| SV013 | CPA Practice Advisor | AI Accounting Platform Numeric Raises $51M Series B | |
| SV014 | Truewind | Announcing Truewind's Series A | |
| SV015 | CPA Practice Advisor | Truewind Accounting AI Platform Raises $13 Million in Series A Funding | |
| SV016 | TechCrunch | Ramp raises $750M at $44B valuation as investors hunger for fintechs with an AI story | |
| SV017 | PR Newswire | Ramp Raises Series F at $44 Billion Valuation | |
| SV018 | CNBC | Fintech Ramp's valuation hits $16 billion in deal led by Peter Thiel's Founders Fund | |
| SV019 | Workiva | Workiva Annual Report 2025 | |
| SV020 | Yahoo Finance | Workiva Inc. (WK) Valuation Measures & Financial Statistics | |
| SV021 | BILL Holdings | Bill Com Holdings Annual Report 2025 | |
| SV022 | Yahoo Finance | BILL Holdings, Inc. (BILL) Valuation Measures & Financial Statistics | |
| SV023 | BlackLine | BlackLine Annual Report 2025 | |
| SV024 | Yahoo Finance | BlackLine, Inc. (BL) Valuation Measures & Financial Statistics | |
| SV025 | Yahoo Finance | Intuit Inc. (INTU) Valuation Measures & Financial Statistics | |
| SV026 | Aventis Advisors | SaaS Valuation Multiples: 2015-2026 | |
| SV027 | Journal of Accountancy | COSO creates audit-ready guidance for governing generative AI | |
| SV028 | Journal of Accountancy | How accountants can balance technology and critical thinking | |
| SV029 | PCAOB | PCAOB Staff Shares Observations From Outreach on Use of Generative Artificial Intelligence in Audits and Financial Reporting | |
| SV030 | National Institute of Standards and Technology | AI Risk Management Framework |