Basis
深受会计师事务所采用、但收入指标未披露的智能代理 AI 会计平台。
Basis 已在会计师事务所中取得真实进展,但在缺少公开 ARR 或利润率披露的情况下,其 $1.15B 估值很难纳入投资判断。
覆盖范围与披露说明
公开证据显示公司已获得有意义的企业客户 traction,并近期完成独角兽融资,但核心财务运营指标仍未披露。
封面要素
公司概况
Basis 是一家总部位于纽约、2023 年成立的成长期私人公司,向会计师事务所销售智能代理 AI 软件,以结果导向定价自动化税务、审计和客户会计流程。公开报道显示,公司已覆盖美国前 25 大会计师事务所约 30% 的业务,同时未披露 ARR、毛利率和留存指标。
- 成立时间
- 2023-01-01
- 创始人
- Matthew Harpe, Mitchell Troyanovsky
- 创立地点
- New York City, New York, United States
- 总部
- New York, New York, United States
- 产品
- 面向 CAS、税务准备和审计测试的智能代理 AI 平台,自动化端到端会计流程。
- 客户
- 专业会计师事务所,尤其是全国性和区域性头部 CPA 事务所。
- 商业模式
- 按已完成的会计工作收取结果导向费用,而非按席位授权。
- 阶段
- Series B
- 融资情况
- 累计融资 1.38 亿美元,其中包括 2026 年 2 月完成的 1 亿美元 B 轮融资;融资后估值为 11.5 亿美元。
执行摘要
主要优势
- 头部会计师事务所早期采用强劲,已有多个具名生产环境部署。
- 产品切中了 CPA 人手短缺和工作流瓶颈,产品市场匹配度清晰。
- 一线投资人阵容和快速融资显示出市场信心。
主要风险
- 公开 ARR、毛利率和留存数据缺失,估值难以核验。
- 受监管的会计工作流带来合规和责任风险,仍需人工复核。
- 依赖第三方模型供应商以及事务所层面的治理,可能限制规模化。
未决问题
- 经审计的 ARR、收入运行率、毛利率和现金消耗披露。
- 具名会计师事务所客户的续约、留存和合同扩张数据。
- 除头部事务所渗透率这一亮点外,生产环境部署的深度。
目录
01公司概览
1.1 公司身份、运营使命与核心商业模式
Basis 是一家总部位于纽约、由 Matthew Harpe 和 Mitchell Troyanovsky 于 2023 年创立的企业软件公司。公司处于成长期,开发面向专业服务机构的智能代理人工智能系统,自动化端到端会计流程。Basis 不做对话式聊天界面或单点任务工具,而是部署能连续执行数小时流程的自主代理,覆盖客户会计服务(CAS)、税务准备和审计测试。公司的核心使命是缓解会计行业的结构性人力短缺:退休注册会计师人数远超新入行者。Basis 希望把会计师从手工编制者变成复核者,扩大事务所的收入承载能力,并按已交付工作而不是传统席位授权收费。[CO001, CO002, CO013]
| 指标 | 数值 / 状态 | 日期 | 置信度 | 尽调缺口 |
|---|---|---|---|---|
| 估值 | $1,150M ($1.15B) | 2026-02 | 高 | 私募投后估值;二级市场定价和清算优先权尚未核验。 |
| 累计融资 | $138M | 2026-02 | 中 | 包括 $3.6M 种子轮、$34M A 轮和 $100M B 轮的累计融资。 |
| 最新轮次 | $100M B 轮 | 2026-02 | 高 | Accel 领投,GV、Khosla Ventures 和 Lloyd Blankfein 参与。 |
| ARR / 运行率 | 2026-10 | 低 | 没有公开收入披露;ARR 和毛利率仍属非公开信息。 | |
| 前 25 家事务所渗透率 | 30%(约 7 / 前 25 家) | 2026-02 | 高 | 已确认的客户关系包括 Boulay、Clark Nuber、MarksNelson、Pinion 和 UHY。 |
| 员工人数 | 51–100 名员工 | 2026-03 | 中 | 团队规模为估算值,工程和机器学习人员主要集中在纽约市。 |
快照指标反映 2026 年 2 月公开的 B 轮公告;私营财务表现指标(ARR、毛利率、现金消耗)仍未披露。
[CO003, CO004, CO005, CO007, CO014, CO015]1.2 高管领导、治理结构与人才策略
Basis 由联合创始人兼 CEO Matthew Harpe 和技术联合创始人 Mitchell Troyanovsky 共同领导。Harpe 曾在 Boston Consulting Group 和 SoftBank 从事经济分析与医疗运营,熟悉复杂计费和合规流程。Troyanovsky 负责技术架构,专注概率编程和自主多智能体编排。Basis 于 2026 年 2 月完成 B 轮融资后,任命 Accel 的 Miles Clements 和 Khosla Ventures 的 Keith Rabois 加入董事会,扩充公司治理。公司预计有 51 至 100 名员工,核心研究与机器学习工程团队集中在纽约市。Basis 大量从顶尖研究机构招聘,并设有名为 Atlas 的内部运营团队,在公司各项运营中部署自主代理。[CO001, CO006, CO012, CO015]
| 人员 | 职务 | 背景 | 创始人与市场匹配 / 职能覆盖 | 关键人物依赖 |
|---|---|---|---|---|
| Matthew Harpe | 联合创始人兼 CEO | Boston Consulting Group 和 SoftBank 前高管,具备医疗和经济分析背景 | 负责商业战略、与前 25 家事务所的企业客户合作及品类定位 | 高;企业事务所关系的主要发言人和负责人 |
| Mitchell Troyanovsky | 联合创始人 | 技术联合创始人,深耕概率编程和多代理系统架构 | 负责核心 AI 架构、OpenAI 合作集成及原生代理 LLMOps 基础设施 | 高;自主多代理会计引擎的首席技术架构师 |
| Atlas Operational Team | 内部代理运营 | 专门的内部工程、机器学习研究和运营团队,负责在公司内部部署代理 | 构建行业工作流,并管理代理在销售和人才运营中的部署 | 中;由 Columbia、MIT 和 Harvard 人才组成的专业技术执行团队 |
根据已披露的联合创始人和关键运营领导单元,管理层清单已完整 截至 2026 年 2 月 B 轮融资,管理层清单已完整。
[CO001, CO006, CO012]1.3 资本化历史、B 轮估值与机构投资者基础
2026 年 2 月,Basis 以 1 亿美元 B 轮融资、融资后估值 11.5 亿美元完成交易,跻身独角兽。该轮融资使公司累计股权融资约 1.38 亿美元,此前公司于 2024 年 12 月完成 360 万美元种子轮和 3400 万美元 A 轮。B 轮由 Accel 领投,GV(前身为 Google Ventures)、老股东 Khosla Ventures 以及前 Goldman Sachs 董事长 Lloyd Blankfein 大额参与。融资还吸引了 Box 的 Aaron Levie、Quora 的 Adam D’Angelo、Replit 的 Amjad Masad 和 Nat Friedman 等知名科技高管与天使投资者。Basis 快速积累资本,反映投资者越来越相信:高风险、受监管知识工作中的垂直 AI 代理能够交付可防御的企业软件价值。[CO003, CO004, CO005]
| 利益相关方 | 角色 | 控制权或经济重要性 | 尽调要求 |
|---|---|---|---|
| Accel | B 轮领投方 | 2026 年 2 月领投 $100M B 轮;合伙人 Miles Clements 加入董事会 | 核验董事会投票权、清算优先权顺位和保护性条款 |
| Khosla Ventures | A 轮领投方兼 B 轮参与者 | 2024 年 12 月领投 $34M A 轮,并在 B 轮追加投资;合伙人 Keith Rabois 加入董事会 | 评估持股比例、按比例认购权及与 Accel 的治理互动 |
| GV (Google Ventures) | 主要机构参与者 | 参与 $100M B 轮融资 | 确认支票金额、信息权以及战略云或基础设施协同 |
| Lloyd Blankfein | 战略天使投资人 | Goldman Sachs Group 前 CEO 兼董事长;参与 B 轮 | 评估顾问参与度、华尔街网络获取能力及机构背书价值 |
| OpenAI | 战略技术合作伙伴 | 核心模型提供商(GPT-4.1、GPT-5、o3、o3-Pro);围绕代理测试开展紧密工程协作 | 检查商业 API 定价协议、延迟 SLA 及对上游模型变化的暴露 |
| Tech Executive Syndicate | 天使投资人 | 参与者包括 Nat Friedman、Daniel Gross、Aaron Levie (Box)、Adam D'Angelo (Quora) 和 Amjad Masad (Replit) | 审查股权结构表集中度及软件生态领袖的顾问承诺 |
机构及主要战略利益相关方清单来自公开披露和融资公告 机构及主要战略利益相关方清单来自与 A 轮和 B 轮相关的公开披露及融资公告。
[CO003, CO005, CO006, CO008]1.4 运营里程碑、技术执行与市场 traction
Basis 已在美国会计市场建立快速商业 traction,覆盖前 25 大会计师事务所约 30%,其中包括 Boulay、Clark Nuber、MarksNelson、Pinion 和 UHY 等知名事务所。采用公司的事务所报告称,在 CAS、税务和审计项目中,运营效率提升 20% 至 50%,事务所无需同比例招聘就能扩大服务能力。技术上,Basis 协调由合作方 OpenAI 基础模型驱动的专业子代理,包括 GPT-4.1、GPT-5、o3 和 o3-Pro。2026 年初,公司展示了能自主完成端到端 Form 1065 合伙企业纳税申报的 AI 代理,完成了高复杂度税务合规自主执行的概念验证(PoC)。[CO007, CO008, CO009, CO010]
Basis 自 2023 年成立至 2026 年取得的关键公司、融资、技术和商业里程碑。
里程碑日期来自公司公告、新闻稿和技术案例研究。
公司成立 [CO001]
种子轮融资 [CO004]
早期事务所部署 [CO002]
A 轮融资 [CO004]
OpenAI 合作架构 [CO008]
自主完成 1065 表申报演示 [CO009]
B 轮独角兽融资 [CO003]
1.5 行业质疑、实施阻力与信息缺口
尽管估值快速上升,Basis 所处行业有软件失败的记录,也长期面对专业人士的怀疑。企业会计科技过去常见的情况是:资金充足的创业公司过度承诺自动化能力,最终却被遗留总账、客户资料不一致和零容错合规要求拖住。作为私人公司,Basis 也没有公开披露 ARR、毛利率、净收入留存率(NRR)或客户流失等关键财务基本面。对潜在合作方和投资者而言,Basis 11.5 亿美元估值究竟代表扎实的经常性软件单位经济,还是投机性的创业投资热情,仍是首要尽调难题。[CO011, CO014]
1.6 展示项
02市场分析
2.1 市场边界:智能代理会计工作流与现状软件
Basis 的可服务市场由 CAS、企业税务合规和审计准备中的复杂会计流程端到端执行构成。QuickBooks、Xero 或 NetSuite 等横向总账和单点记账工具仍要求会计师手动录入交易、核对报表;Basis 则自主完成多步骤工作流。平台面向希望在不增加员工的情况下扩大客户量的中端和大型公共会计事务所及企业财务部门。可计入的支出包括工作流自动化、税务工作底稿准备和试算平衡表映射,不包括通用 ERP 授权和核心数据库托管。因此,Basis 直接与人工工时竞争,而不是与既有软件席位预算竞争。[CM001, CM008, CM012]
| 业务分部 | 纳入支出范围 | 排除支出范围 | 主要买方 / 付款方 | 战略相关性 |
|---|---|---|---|---|
| 客户会计服务 (CAS) | 持续交易编码、多实体月末结账、总账清理、报表对账 | 总账核心托管、支付处理费、工资发放 | CAS 业务负责人 / 董事总经理 | 高频重复的月度结账;初期采用速度最快 |
| 税务业务工作底稿与合规 | 试算平衡表税务映射、多层 K-1 提取、M-1 调整、1065 申报表编制 | 政府税务申报费、本地化物业税评估软件 | 税务业务管理合伙人 | 缓解 Q1/Q3 申报截止期的季节性产能瓶颈 |
| 审计与鉴证工作底稿 | 实质性测试执行、证据引用、工作底稿起草、例外记录 | 法定鉴证意见签署、合伙人签字责任、审计委员会顾问服务 | 审计质量负责人 / 管理合伙人 | 责任敏感度最高;需要透明的引用链和人工复核 |
| 企业财务与财务控制团队 | 内部差异分析、管理报告、13 周现金流预测 | ERP 数据库基础设施、企业采购、资金管理 | 财务副总裁 / 企业财务总监 | CPA 事务所初步验证后的第二扩张分部 |
对代理会计平台可触达的软件和自动化支出与排除在外的横向 ERP 系统进行分类。
[CM001, CM008, CM012]2.2 可服务市场规模:跨业务线的多视角测算
智能代理会计平台不能用单一自上而下的软件数字来衡量,因为这项技术替代的是人工专业服务工时。美国约有 3m 名会计师,创造超过 $150b 的专业服务总收入。可服务市场集中在会计师事务所:日常合规、月结和税务准备最多占专业员工工时的 70%。这部分劳动力对应的软硬件及外部自动化预算,构成每年 $15b 至 $20b 的可服务软件市场。Basis 聚焦前 150 大会计师事务所和区域性事务所;前 25 大事务所的早期部署,已验证市场短期付费购买可扩展产能的意愿。2026 年 2 月,Accel 和 GV 领投的 $100m B 轮融资、$1.15b 估值进一步增强了投资者信心。[CM002, CM004, CM006, CM009, CM013]
| 测算视角 | 地理与年份 | 估算市场 / 支出 | 增长 / CAGR | 估算方法 | 置信度 | 分析局限 |
|---|---|---|---|---|---|---|
| 美国会计服务宏观收入 (TAM) | 美国(2026) | $150B+ 年度总收入 | 4% - 6% 年度行业增长 | 汇总约 3M 名专业会计师及 CPA 事务所的总开票额 | 中 | 捕捉的是总开票额,而非可触达的软件 / 自动化支出 |
| 可触达的垂直会计软件支出 (SAM) | 美国(2026) | $15B - $20B 年度 IT 支出 | 12% - 15% 垂直 SaaS 增长 | 将估算的 10%-12% 软件和 IT 支出占比应用于可触达事务所开票额 | 中 | 除 AI 自动化外,还包括横向工具和传统合规软件 |
| 可服务的代理会计工作流 (SOM) | 美国前 150 家及中型事务所(2026) | $3B - $5B 可触达年度支出 | 25%+ 代理工作流 CAGR | 自下而上估算中大型事务所 CAS、税务准备和审计测试的支出能力 | 中 | 取决于合伙人是否愿意用结果导向软件替代可计费人工工时 |
| 前 25 家会计师事务所企业空间 | 美国前 25 家事务所(2026) | $800M - $1.2B 企业支出池 | 30%+ 高增长浪潮 | 针对美国 25 大会计网络的企业支出 | 中 | 私下合同金额和试点转生产转化率仍未披露 |
综合专业会计劳动力池、软件自动化支出和事务所采用上限,形成多视角市场规模测算。
[CM002, CM004, CM009]从美国会计劳动力支出宏观规模到代理工作流签约采用的分层市场规模。
各层宽度仅表示层级,不代表数值比例。
数值综合公开行业劳动力总量、典型垂直软件支出占比和据报道的事务所渗透率;公司 ARR 仍未披露。
[CM002, CM004, CM006, CM013]2.3 买方画像、业务分层与采购机制
不同规模事务所和业务组的采用逻辑差异明显。在前 25 大和大型全国性会计师事务所,经济买方通常是管理合伙人、业务负责人和全国管理合伙人,他们负责 CAS、税务或咨询业务线。由于替代性执业模式把持牌鉴证合伙企业与咨询公司主体分开,采购往往需要双重治理。在中端和区域性事务所,管理合伙人优先释放产能,避免在报税高峰拒绝客户项目。Basis 按已完成的会计流程和交付物收费,而非按席位授权,使商业模式贴合事务所经济,也减轻了软件闲置浪费的担忧。[CM005, CM007, CM010]
| 事务所分部 | 经济买方 | 主要用户 | 付款实体 | 核心目标工作流 | 预算负责人 | 采用触发因素 |
|---|---|---|---|---|---|---|
| 全国前 25 家 CPA 事务所 | 全国管理合伙人 / 创新负责人 | 高级经理、复核 CPA、项目负责人 | 咨询实体(替代性业务结构) | 多办公室 CAS 结账、50 页合伙企业税务工作簿 | 事务所创新与业务线预算 | 员工严重流失,无法完成企业咨询任务 |
| 第 26 - 150 名区域会计师事务所 | 管理合伙人 / 业务组负责人 | 会计员工、税务高级人员、CAS 经理 | 运营 CPA 合伙企业 / 咨询 LLC | 端到端月末结账、M-1 账税调整 | 业务组运营预算 | 产能瓶颈阻碍新增商业客户账户上线 |
| 中型独立 CPA 事务所(10-50 名员工) | 管理合伙人 / 创始人 | 综合 CPA、簿记员、税务准备人员 | 运营 CPA 事务所 | 标准化 350+ 项程序结账和总账清理 | 通用技术预算 | 不堪重负的合伙人周末做准备工作,而不是复核 |
| 企业财务与内部会计 | 财务副总裁 / 首席会计官 | 高级会计师、财务分析师 | 企业实体 | 持续银行和信用卡编码、差异报告 | 企业 IT / G&A 财务预算 | 随交易量扩大仍保持精简会计团队的愿望 |
描述各层级会计师事务所及业务领域的买方画像和采购动态。
[CM005, CM007, CM010]2.4 采用驱动与约束:人口逆风对监管验证
智能代理会计采用的首要增长动力,是整个行业严重且结构性的人员危机。约 30 万名会计师将在两年内退休,预计未来十年有 75% 的从业者退休,事务所无法招聘足够 CPA 满足合规需求。能执行多步骤月结和纳税申报的自主代理,可节省最多 30% 的时间,释放关键运营产能。但严格的专业责任、AICPA 标准和交叉复核要求限制了采用。CPA 需在法律责任下签署纳税申报表和鉴证意见,因此 AI 平台必须提供透明审计轨迹、来源引用和置信度评分,而不能只是黑箱输出。关于客户留存和毛利率的经审计商业披露也仍然缺失。[CM003, CM011, CM013, CM014]
2.5 展示项
03竞争格局
3.1 竞争格局与替代路径
Basis 面对五类替代方案:传统会计软件和 ERP 套件、新兴垂直会计 AI 代理、通用基础模型实验室、离岸人力网络,以及根深蒂固的人工流程。Intuit、Thomson Reuters、Wolters Kluwer 和大型企业 ERP 厂商掌握既有记录系统和客户会计数据,但其架构仍以确定性规则引擎、表单处理和工作流管理为中心,而不是自主代理执行。Thomson Reuters Institute 研究显示,专业服务业已从探索式生成式试验进入战略部署阶段,事务所要求深度工作流自动化直接嵌入运营战略。与此同时,OpenAI 等通用基础模型厂商继续推进原始测试时计算、推理模型和工具调用。但专业会计工作流需要接入多种总账、对照一手工作底稿进行多步骤核验,并适配本地事务所惯例。面对会计师严重短缺和合伙人职业倦怠,事务所在专用代理平台、内部封装前沿模型的脚本,以及离岸团队支持的人工准备之间做选择。 [CP001, CP003, CP004, CP009]
| 替代类别 | 代表性参与者 | 核心价值主张 | 主要差异化 | 主要局限与风险 |
|---|---|---|---|---|
| 垂直会计 AI 代理 | Basis | 面向税务、审计和 CAS 的自主多步骤代理执行 | 350+ 项预构建会计程序、多模型路由和一手来源引用 | 定制定价不透明、税务覆盖以美国为中心、演示流程封闭 |
| 传统会计软件与 ERP | Thomson Reuters、Wolters Kluwer、Intuit | 成熟的记录系统、合规引擎和申报基础设施 | 根深蒂固的客户关系、广泛分销和认证税务引擎 | 僵化的规则架构以及大量人工录入要求 |
| 前沿基础模型实验室 | OpenAI、Anthropic、Google | 原始推理能力、测试时计算扩展和多模态工具使用 | 最先进的通用智能、前沿模型和多工具编排 | 缺乏深度垂直集成、领域审计轨迹和总账连接 |
| 现状人工与离岸团队 | 内部初级会计师、BPO 离岸人员供应商 | 灵活的人工作判断、成熟的员工分工和客户熟悉度 | 没有软件集成复杂度,合伙人熟悉直接委派 | 严重劳动力供给短缺、高流失、季节性过劳和扩张限制 |
根据对厂商披露、行业报告和专家目录评估的尽调审阅进行比较分类;定价和性能指标反映截至 2026 年的 企业尽调基准。
[CP001, CP003, CP007, CP009]按工作流自主性(X 轴)和会计领域专业化程度(Y 轴)对会计解决方案进行序数定位。
| 条目 | X (2–8.5) | Y (2.5–9) | 说明 |
|---|---|---|---|
| 1. Basis | 8.5 | 9 | |
| 2. 老牌套件(Thomson Reuters / Intuit) | 3.5 | 8 | |
| 3. 基础模型实验室(OpenAI / Anthropic) | 6 | 2.5 | |
| 4. 现状人工编制 | 2 | 6.5 | |
| 5. 通用企业工作流工具 | 4.5 | 3.5 | |
| 6. 离岸 BPO 会计团队 | 3 | 6 |
坐标根据 2026 年尽调评估的技术架构、领域专属功能深度和自主工作流能力,在 1–10 归一化序数尺度上定位。
[CP001, CP004, CP007, CP009, CP011, CP012]3.2 竞争者画像、技术能力与护城河耐久性
对替代路径的尽调显示,技术架构、专业化程度和防御性存在明显分化。OpenAI o3、o3-Pro、GPT-4.1 和 GPT-5 等通用模型实验室提供前沿推理能力,Basis 自身也将其作为模块化底层组件,而不是基础模型供应商。Basis 认为,前沿实验室会自然商业化搜索、文档生成和基础表格操作等广泛知识任务,复杂的领域金融控制与会计流程仍能由垂直专家守住。传统套件提供确定性计算和合规申报模块,却需要人工持续输入,才能打通非结构化文档、试算平衡表和账期调整。相比之下,Basis 部署监督代理,在专业子代理之间编排多智能体委派;每项数字交付物都引用一手来源文档,并提供 350 多个由内部驻场会计师设计的预构建会计 Procedures。该架构帮助 Basis 覆盖美国前 25 大会计师事务所约 30%,包括与 Baker Tilly 和 Sikich 的合作,并覆盖前 150 大事务所约 25%。不过,如果基础模型厂商增强原生记录系统工具,或传统供应商把代理能力深度集成进现有事务所管理软件,长期护城河仍会承压。 [CP002, CP005, CP006, CP007, CP008, CP010, CP011, CP012]
| 能力维度 | Basis 平台 | 传统税务与审计套件 | 通用 LLM 聊天机器人 | 尽调状态与缺口 |
|---|---|---|---|---|
| 端到端工作流执行 | 可运行数小时的自主多步骤代理 | 模板计算引擎支持人工任务执行 | 对话式文本提示,无法自主闭环工作流 | 已通过厂商发布和合作伙伴公告核验 |
| 工作底稿与来源引用 | 每个计算数字和调整均引用一手文件 | 由员工编制人员手动添加超链接或静态 PDF 附件 | 综合文本回答缺乏确定性工作底稿引用 | 已记录于产品架构和评测画像 |
| 会计程序库 | 350+ 项预构建专家程序,加上可定制的事务所规则 | 标准税表、固定附表和模板检查清单 | 不使用定制提示时没有内置标准操作程序 | 已通过官方小型事务所解决方案文档确认 |
| 底层模型策略 | 多模型路由,优化延迟、任务复杂度和推理 | 带有初步 GenAI 封装的专有传统规则算法 | 绑定厂商生态的单一基础模型系列 | 已在技术架构分析和实验室基准中核验 |
矩阵细节来自产品文档、技术拆解和平台架构分析;能力反映截至 2026 年可用的商业版本。
[CP002, CP005, CP007, CP008]3.3 切换成本、锁定与竞争不利风险
会计行业的切换成本和采购机制既构成进入壁垒,也带来独特采用风险。监管责任、合伙人签字要求和严格客户保密义务,使事务所高度厌恶风险。一旦专业事务所把定制 Procedures、历史试算平衡表映射和客户上下文嵌入代理平台,流程锁定会随时间加深。Basis 还通过 SOC 2 Type II、ISO 27001 和 ISO 42001 认证、美国基础设施托管,以及不使用客户数据训练模型的明确政策,落实企业合规。不过,行业监测机构 RECATOOLS 的独立评估指出,Basis 没有透明的公开价格表,访问受企业演示候补名单限制,基础设施严格位于美国区域,公开案例主要围绕美国税务和审计惯例,如 Form 1065 合伙企业申报和 K-1 汇总。对国际事务所和多司法辖区业务而言,这构成采用障碍。对专有基础模型 API 的依赖还带来平台依赖与上游模型成本变化的双重风险,结果导向定价则仍需谈判并接受企业合同审查。 [CP003, CP006, CP013, CP014, CP015, CP016]
| 厂商 / 范式 | 定价机制 | 合同结构 | 企业导入 | 战略买方权衡 |
|---|---|---|---|---|
| Basis | 与已完成会计工作流匹配的结果导向价值定价 | 年度定制企业协议,未公布公开目录价 | 受限的演示筛选和企业候补名单部署 | 价值匹配度高,但中型事务所的价格透明度有限 |
| 传统会计软件 | 按席位用户许可,加年度软件维护模块 | 带分层用量计划的多年企业订阅 | 直接与销售代表签约,或由标准化授权合作伙伴转售 | 预算可预测,但生产率不提升时席位许可成本持续上升 |
| 基础模型 API | 按 token 计费,根据输入 / 输出量收取计算消耗费 | 即用即付的实用额度计费或企业计算承诺 | 自助 API 开发者密钥或云厂商企业控制台 | 边际单位成本低,但内部工程和编排开销高 |
定价机制反映厂商公开说法和行业评估;Basis 目录价仍未披露,需通过企业报价协商。
[CP006, CP010, CP014, CP015]3.4 展示项
04财务
4.1 收入来源与定价架构
Basis 拒绝按用户人数或原始计算 token 收取费用,转而按 CAS、税务合规和审计工作流中经核验完成的会计工作收费。该模式直接对应专业服务事务所的结构性经济:客户计费取决于交付物,而非软件席位使用率。平台打包了 350 多个由内部驻场会计师开发的预构建会计流程,使整个业务线都能采用。早期商业化需要克制:Basis 接受较慢的初期收入增长,以准确归因每项交付物,并确保商业激励始终与会计吞吐量紧密一致。[CI001, CI002, CI003, CI013]
| 收入流 | 变现机制 | 价值单位 | 当前状态 | 收入质量 | 尽调要求 |
|---|---|---|---|---|---|
| 客户会计服务 (CAS) 自动化 | 结果导向工作流计费 | 已处理交易与月末结账对账 | 已在生产环境商业化 | 与事务所月度周期挂钩,重复收入潜力高 | 合同最低承诺及超额用量费率表 |
| 税务业务工作底稿与合规 | 按完成的申报或申报表交付件收费 | 已完成的 Form 1065、K-1 套件和税务工作底稿 | 已在生产环境商业化 | 收入季节性集中于 Q1/Q3 申报截止期 | 季节性收入波动及年度续约结构 |
| 审计业务准备与测试 | 按每项审计测试程序交付件收费 | 已完成的审计工作底稿和实质性测试计划 | 已在合作事务所中积极部署 | 与审计周期挂钩的年度重复项目 | 按审计项目归因收入并明确责任条款 |
| 全所业务赋能与定制程序 | 含基础容量的分层平台费 | 在会计师事务所全业务部署 | 标准商业部署模式 | 可预测的合同基础收入及扩张上行空间 | 合同平台承诺与可变用量对照表 |
收入流按会计业务功能领域和商业化机制分类;具体美元收入拆分仍是公司非公开数据。
[CI001, CI002, CI003]| 工作流 / 模块 | 合同与定价模式 | 标价与实际条款 | 折扣与商业未知项 | 证据来源 |
|---|---|---|---|---|
| 客户会计服务(CAS) | 按完成的会计程序计费,以交付结果为基础 | 按事务所定制、谈判企业合同 | 批量折扣门槛和月度最低消费未知 | ZenML LLMOps Database & Official Solutions |
| 合伙企业税务工作底稿与 Form 1065 | 按每份税表工作底稿交付计费 | 完成全流程申报后计费 | 异常处理的重新运行计费政策未披露 | PYMNTS & Built In NYC |
| 事务所全所许可 | 全所部署模式(无单席位许可) | 覆盖多客户环境的全所承诺 | 多年锁价承诺和续约上限未披露 | Accounting Today Executive Interview |
| 已部署智能系统上线 | 打包实施与工作流定制 | 纳入全所企业合同 | 独立专业服务费结构未披露 | Accounting Today & ZenML Report |
Basis 已发布的商业模式优先考虑交付结果;确切费率层级和折扣安排仍需查阅保密合同。
[CI001, CI002, CI003, CI013]4.2 GTM 推进与工作流经济转化
Basis 的 GTM 通过与美国大型会计事务所建立自上而下的合作,面向承受严重人才短缺和利润压力的领导团队。商业部署管线借助多智能体架构,把非结构化客户财务原始记录转成最终会计底稿和申报文件。采用事务所报告称,日常任务平均节省 30% 时间,专业员工因此可转向利润更高的咨询服务,业务产能也能在不同比例增加员工的情况下扩张。系统内的自主代理可独立运行数小时,处理试算平衡表映射、多实体 K-1 文件提取、复杂银行调节和 Form 1065 合伙企业纳税申报,并保留透明来源引用。[CI001, CI003, CI005, CI006]
4.3 成本结构与单位经济驱动因素
由于基础模型推理和专业工程薪酬成本高,Basis 的成本结构显著不同于传统软件公司。Basis 将复杂会计工作流的每一步路由到 OpenAI 前沿模型套件中的最优模型,包括 o3、o3-Pro、GPT-4.1 和 GPT-5,因此销售成本直接跟测试时计算扩展和推理深度相关。约 120 名员工位于纽约市,人员成本占运营支出很大比重,公司在那里与顶尖 AI 研究实验室争夺机器学习人才。未来毛利率高度取决于推理单位成本、每项交付物的 token 用量,以及内部 Deployed Intelligence 团队为验证客户准确性和建立事务所信任而提供的实施服务。[CI004, CI009, CI014]
| 指标 / 构成 | 观察值或估计值 | 置信度 | 战略重要性 | 承保判断尽调问题 |
|---|---|---|---|---|
| 客户时间节省 | 事务所平均节省 30% 时间 | 中 | ROI 和事务所付费意愿的核心驱动 | 客户层面的时间追踪审计和工作流使用数据 |
| 毛利率(COGS) | 私有数据未披露 | 低 | 决定长期软件业务与服务业务的利润率结构 | 经审计的托管、OpenAI 推理 API 计费和毛利率账簿 |
| 研发 / 工程团队人数 | 纽约市 120 名员工 | 中 | 与 AI 实验室竞争时的主要运营烧钱项 | 当前工资表、部门拆分和招聘计划 |
| 客户渗透率 | 美国前 25 大会计事务所的 30% 至 40% | 高 | 验证企业 GTM 牵引力和账户扩张空间 | 命名客户清单、每账户 ARR 和客户 logo 留存率 |
单位经济指标综合自高管披露和行业合作伙伴案例研究;毛利率与销售成本仍是关键的私有尽调缺口。
[CI004, CI005, CI009, CI014]4.4 资本充足性与财务投资判断
Basis 在 2026 年 2 月完成 Accel 和 Google Ventures 领投的 $100m B 轮融资后,资本充足性很强;融资把融资后估值推至 $1.15b,累计风险投资约 $138m。 资产负债表扩张让公司有数年现金 runway,可吸收研发、算力和薪酬高成本,并把覆盖范围从当前前 25 大美国会计师事务所约 40%、大致三成至四成的 footprint 继续扩大。Top 50 会计师事务所 Sikich 已在多个业务线部署 Basis 等机构合作,验证了强劲企业需求。但从投资判断看,$1.15b 估值仍有显著执行风险:会计科技历史上自动化失败频发,合同 ARR 未披露,实际毛利率未经验证,客户留存 cohort 也不透明。[CI007, CI008, CI010, CI011, CI012]
4.5 展示项
05产品与技术
5.1 核心产品架构与智能代理运营模式
Basis 为注册会计师事务所和专业会计团队提供专用智能代理人工智能平台。Basis 不做需要用户持续提示的对话式聊天机器人,也不只是简单的文本输入输出工具,而是把会计运营设计成自主多步骤工作流。平台部署目标导向的 AI 代理,摄取财务原始记录,解读试算平衡表,归类交易,准备工作底稿,并生成可直接申报的纳税申报表。每个多步骤流程都由会计师监督,确保专业人员在过账到客户总账前,始终能完整查看并控制所有计算、映射和待入账分录。平台由 Matt Harp 和 Mitchell Troyanovsky 于 2023 年创立,并在 2026 年 2 月完成 1 亿美元 B 轮融资、达到 11.5 亿美元估值,用于把多智能体平台拓展到专业服务事务所。[CE001, CE002, CE003, CE004, CE005]
5.2 会计工作流与预构建流程模块
平台解决 CAS、税务合规和审计支持中的关键运营瓶颈。在 CAS 中,Basis 直接连接客户总账,自动完成账务整理、清理混乱账目,并以科目代码、描述和关联来源文件准备待入账分录。对税务合规业务,平台可大规模提取复杂财务信息,例如合伙企业 Schedule K-1 脚注以及主动与被动分配,再把账面试算平衡表映射到税务项目并起草 Form 1065 合伙企业申报表。Basis 提供 350 多个由内部驻场会计师编写的预构建会计 Procedures,事务所可以把复核偏好、重要性阈值和客户特有习惯固化为可重复执行的工作流。专业人员在结构化阶段复核建议分录,减少繁琐手工录入,同时保留机构复核标准。[CE001, CE006, CE007, CE008]
| 模块 / 能力 | 目标用户 | 成熟度 / 状态 | 核心差异化 | 尽调缺口 |
|---|---|---|---|---|
| 客户会计服务(CAS) | 初级与高级会计师 | 生产环境 GA | 自动清理总账、编制调整分录,并生成待审核的记账草稿 | 不同总账系统中的真实错误率和人工覆盖率 |
| 税务准备与质量控制 | 税务编制人员与复核人员 | 生产环境 GA(已演示 1065) | 文档提取(K-1)、账簿到税务映射、M-1 工作底稿、来源引用 | 多层合伙企业脚注和罕见州代码的边界情况处理 |
| 审计工作底稿与审计轨迹支持 | 审计人员与经理 | 生产环境 GA | 结构化 / 非结构化数据提取、不可变操作日志、保留 7 年 | 中型事务所环境中的正式 SOC 合规验证 |
| 定制程序构建器 | 事务所合伙人与质量负责人 | 生产环境 GA(预置 350+ 项) | 记录事务所特定的复核规则、客户惯例和复核阈值 | 缺少公开自助文档和第三方开发者扩展能力 |
模块成熟度与能力汇总自 Basis 官方解决方案文档、合作伙伴案例研究和独立技术评测。所有模块在过账前都要求会计师复核。
[CE001, CE006, CE007, CE008]| 用户任务 | 当前人工工作流 | Basis 解决方案 | 可量化收益 | 技术 / 运营限制 |
|---|---|---|---|---|
| 月末对账 | 在 Excel 中手工匹配银行对账单与总账交易 | 监督代理匹配交易、标记差异并起草调整分录 | 事务所最多节省 30% 时间,周转时间从数小时压缩至数分钟 | 需要启用 ERP / 总账集成;未经复核无法自主过账 |
| 税务工作底稿与申报表质量控制 | 手工逐项核对来源文件与税表草稿 | 自动摄取文件、映射试算平衡表行项目,并附引用标记差异 | 每个数字都有来源文件引用;每位合伙人每周节省 10–20 小时 | 在异常规则下,自主执行税务高度依赖前沿模型准确性 |
| K-1 合伙企业信息提取 | 手工录入数十个实体的脚注及被动 / 非被动分配 | 批量摄取合伙企业 K-1 文件,自动提取并附引用 | 消除复杂多实体投资结构中的手工重复录入 | 非结构化脚注的非标准披露仍需人工介入 |
| 财务报告与交付物 | 手工汇总差异报告、现金流预测和管理层材料包 | 自动起草报告、预测 13 周现金流并解释差异 | 释放从事客户咨询服务的专业人员产能,而非只做汇编 | 依赖历史总账的整洁度;杂乱的客户账簿需要先清理 |
工作流收益和限制综合自官方产品描述、执业人员案例引语和第三方架构分析。
[CE001, CE003, CE005, CE006]5.3 模型路由、本地上下文学习与工具调用
Basis 的底层技术架构把 OpenAI 前沿推理模型与确定性会计逻辑、实体特定的本地上下文结合起来。任务从监督代理开始,该代理最初基于 OpenAI o3 构建、后迁移至 GPT-5,负责评估新任务并拆解成协调的子代理工作流。编排器依据延迟和复杂度要求,把专业任务分派给不同前沿模型:用 GPT-4.1 处理对速度敏感的交互循环,用 GPT-5 处理深度多步推理、交易歧义消解和财务调节。在启用代码解释器和网页搜索的并行工具执行内部基准中,GPT-5 在协调结构化会计操作上达到 100% 成功率。Basis 不只依赖会忽略客户细节的全局模型,而是沉淀实体规则、科目映射和业务惯例,随着事务所扩大使用,系统精度持续提升。[CE002, CE009, CE010, CE011, CE012]
| 层级 / 组件 | 角色与功能 | 核心依赖 | 技术风险与尽调重点 |
|---|---|---|---|
| 监督代理编排 | 拆解高层会计目标,将任务路由给子代理,并执行工作流关卡 | OpenAI GPT-5 和 o3/o3-Pro 推理模型;内部基准评估套件 | 高度依赖前沿模型推理稳定性和 OpenAI API 可用性 |
| 专用执行子代理 | 执行交易分类、数据提取和分录起草等定向会计步骤 | 多模型集成(GPT-5 用于深度推理,GPT-4.1 用于速度关键型交互) | 上游 API 版本更新造成模型漂移,改变分类一致性 |
| 会计规则与约束引擎 | 对 LLM 输出执行确定性会计逻辑、税法规则和平衡校验 | 专有会计规则集、350+ 项预置程序、事务所复核政策 | 新法规或司法辖区特定法规中的规则集缺口可能导致错误未被标记 |
| 集成与账本连接层 | 连接事务所文件系统、客户账本和 ERP,且不存储客户凭证 | 事务所直部署集成;账本 / ERP API 与文档存储连接器 | 缺少公开自助 API 界面,限制第三方审计和事务所定制工具 |
架构组件依据 OpenAI 技术案例、ZenML LLMOps Database 资料和 TeachUsTechnology 技术概览。
[CE002, CE003, CE009, CE011]六层架构栈展示 Basis 的多代理编排、模型路由、确定性规则和企业集成边界。
- 用户交互与监督层
- CPA 使用的网页工作区,可分配目标、检查带引用的工作底稿并批准分阶段记账分录
- 模块
- 目标分配界面
- 工作底稿检查与复核控制台
- 审计日志与置信度评分查看器
- 监督编排层
- 目标分解引擎,协调跨专业子代理的多步骤会计工作流
- 模块
- 目标分解器与规划器
- 并行工具调用协调器
- 复核阈值调度器
- 专业子代理层
- 执行离散会计、税务和审计流程的领域专属自主代理
- 模块
- CAS 与对账代理
- 税务提取与 1065 申报代理
- 审计轨迹与验证代理
- 模型路由与推理层
- 根据任务复杂度和延迟,为专业任务分配不同前沿 LLM 的多模型路由
- 模块
- 用于深度推理的 OpenAI GPT-5
- 用于低延迟交互的 OpenAI GPT-4.1
- 内部基准路由控制器
- 会计规则与本地上下文层
- 确定性护栏结合 350+ 项会计流程,以及事务所和客户特定的学习结果
- 模块
- 350+ 项预构建会计流程
- 本地事务所上下文固化
- 基于规则的验证约束
- 数据与企业集成层
- 安全的分类账连接器和文档摄取管道,无需存储客户凭证即可运行
- 模块
- ERP 与总账连接器
- 文档解析器(K-1、报表)
- SOC 2 Type II 安全存储
架构层概念化呈现 Basis 的技术实现,依据包括官方产品披露、OpenAI 合作技术细节和 LLMOps 分析。
[CE002, CE009, CE010, CE012]5.4 可审计性、确定性护栏与风险控制
在受监管的会计和审计环境中,准确性、可审计性和数据安全是不可妥协的运营要求。Basis 在基础模型推理之外施加严格的规则会计约束,确保代理不能脱离既定税法、借贷平衡约束和事务所审批参数自由推理。平台生成的每个数字、调整分录和税务项目,都直接链接到底层来源文件,并留有不可篡改的操作日志和置信度评分。云部署可接入事务所文件存储和总账,且不把客户凭据存放在 Basis 服务器上。不过,由于 Basis 采用自主模式而非辅助模式,平台可靠性根本上取决于上游基础模型精度;同时,缺少公开自助 API 也使部署支持集中在直接企业合作上。[CE003, CE004, CE011, CE013, CE014]
5.5 展示项
06客户
6.1 会计师事务所基础分层与买方经济
Basis 的 GTM 专门面向专业会计师事务所,而不是企业应付账款部门或通用内部企业财务团队。在会计行业内,公司把可服务客户分成三层:前 25 大全国性会计师事务所、前 150 大区域及中型事务所,以及精品小型会计师事务所。所有层级的核心经济买方都是事务所管理合伙人或业务负责人,日常主要使用者则是资深 CPA 和普通会计师。付款预算来自事务所运营和技术预算,背后是美国结构性人才危机:退休 CPA 多于新入行者。Basis 同时覆盖 CAS、纳税申报准备和审计测试,在统一平台中拿下多个业务线预算,而不是作为单点工具竞争。公司有意排除企业 AP 部门,以避开商品化 ERP OCR 竞争,让平台专注外部客户会计项目的流程。[CU001, CU002, CU006, CU010]
| 细分市场 | 买方 / 用户 / 付款方 | 核心使用场景 | 估计规模 | 战略与收入价值 | 尽调缺口 |
|---|---|---|---|---|---|
| 前 25 大会计事务所 | 管理合伙人和业务负责人 / 审计与税务 CPA / 事务所运营预算 | 复杂的多司法辖区税表准备、审计测试程序、企业 CAS 结账 | 约 7 家事务所(30% 渗透率) | 锚定品牌验证、大额合同价值、多办公室扩张 | 全所部署与单办公室部门部署的范围未知 |
| 前 150 家区域 CPA 事务所 | 管理合伙人和 CAS 业务负责人 / 高级会计师 / 事务所技术预算 | 客户会计服务编制、月末对账、1065 合伙企业申报 | 约 35–40 家事务所(25% 渗透率) | 高运营杠杆、严重的区域人才短缺驱动、更快销售周期 | 合同期限和跨业务扩张速度未披露 |
| 小型与精品事务所 | 个体执业者和管理合伙人 / CPA 与簿记员 / 业务所有者 | K-1 汇总、试算平衡表映射、记账草稿暂存、客户咨询 | 通过候补名单覆盖数百家精品事务所 | 广泛的程序测试场、社区倡导、实施摩擦低 | 合伙人自报的节省工时缺乏独立验证 |
| 企业级应付账款(非目标市场) | 首席会计官 / 应付账款员工 / 企业财务 | 供应商发票处理、基础应付账款匹配、总账编码 | 排除在当前 GTM 重点之外 | 为避免与通用 ERP 自动化商品化竞争,降低优先级 | 企业财务与事务所 CAS 工作流之间的边界尚未绘制 |
客户细分源自公开客户公告、供应商定位披露和合作伙伴案例研究;市场规模反映美国会计行业结构。
[CU001, CU002, CU006, CU010]6.2 采用轨迹、产能扩张与效率证明
自 2023 年成立以来,Basis 的采用速度快速提升,受全行业 CPA 人手短缺和繁忙季工作压缩推动。Reuters 独立报道及风险投资披露确认,Basis 当前服务美国前 25 大会计师事务所约 30%,约相当于 7 家全国性事务所;同时覆盖前 150 大区域性事务所约 25%。采用背后是核心合规流程的显著效率提升:合伙人研究和客户报告显示,日常任务平均节省最多 30% 时间,各事务所整体工作流效率提升 20% 至 50%。在较小事务所,独立执业者和管理合伙人称每周可找回 10 至 20 小时,把数小时复核任务压缩到 30 分钟。为加快部署、减少初期实施摩擦,Basis 配备 350 多个由内部驻场会计师编写的预构建会计流程,事务所无需定制软件工程即可马上进行试算平衡表映射和 K-1 脚注提取。[CU001, CU003, CU004, CU005, CU008]
| 采用指标 | 报告值 | 日期 | 来源 | 置信度 | 战略含义 | 缺失分母 |
|---|---|---|---|---|---|---|
| 前 25 大 CPA 事务所渗透率 | 约 30%(约 7 家事务所) | 2026-02 | SiliconANGLE & Reuters | 高 | 确立其在会计自动化企业品类中的领导地位 | 事务所总席位数和活跃项目量未披露 |
| 前 150 大 CPA 事务所渗透率 | 前 150 家事务所中约 25% | 2026-02 | ZenML LLMOps Database | 中 | 证明商业版图已扩展至区域会计事务所 | 已部署许可数占全所专业人员总数的比例未知 |
| 平均工作流时间节省 | 最多节省 30% 时间(20%–50% 区间) | 2025-08 | OpenAI & Accountio | 中 | 在繁忙季直接缓解 CPA 人才短缺 | 缺少经审计的计费工时替代与非计费任务吸收数据 |
| 预置会计程序 | 350+ 项标准化程序 | 2026-02 | Basis Official Surface | 高 | 无需定制工作流工程即可快速让小型事务所上线 | 各程序在真实客户账户中的使用频率未经验证 |
轨迹指标依赖公司公告、创投报道和合作伙伴研究;没有第三方审计或 SEC 文件佐证当前客户数量。
[CU001, CU003, CU004, CU008]6.3 生产环境部署与具名事务所案例
公开证据确认,Basis 已从初步概念验证(PoC)试验走向多家知名会计师事务所的生产部署。已确认的企业和中端客户包括 UHY LLC(美国前 25 大会计师事务所)、Boulay PLLP、Clark Nuber PS、MarksNelson LLC、Pinion LLC 和 Wiss & Company,以及 Whitcomb & Sheridan LLP 等精品事务所。客户部署覆盖复杂生产流程,而非孤立营销试点:Wiss & Company 的会计团队称,平台让专业员工摆脱繁琐手工调节,专注利润更高的咨询关系,改变了公司发展轨迹。在 Pinion、Boulay 等中型事务所,代理执行多步骤税务准备、审计测试和客户账务整理,直接连接客户总账、QuickBooks、Xero 和企业资源规划系统。CPA 仍在主动监督闭环中,复核并批准代理准备的工作底稿和会计分录。但由于客户账户受双方保密协议约束,头部客户的席位数量、业务组和活跃客户文件分布仍部分未披露。[CU001, CU002, CU005, CU007, CU009]
| 客户 | 事务所细分市场 | 部署与使用场景 | 生产环境还是试点 | 报告结果 | 限制 / 风险 |
|---|---|---|---|---|---|
| Wiss & Company | 前 150 大 CPA 事务所 | 全所 CAS、税务和咨询代理工作流 | 生产环境部署 | 效率提升足以定义增长轨迹,让员工专注高价值客户咨询 | 运营和财务影响为自报,未经第三方审计 |
| UHY LLC | 前 25 大 CPA 事务所 | 覆盖 CAS、税务和审计的多业务代理部署 | 生产环境部署 | 无需同比例招聘即可扩大各办公室的项目承接能力 | 参与办公室和业务组的确切数量未披露 |
| Boulay PLLP | 区域 CPA 事务所 | 税务准备、工作底稿生成和账户对账 | 生产环境部署 | 自动处理劳动密集型文档复核和试算平衡表映射 | 合同期限、年度支出和席位数未披露 |
| Clark Nuber PS | 区域 CPA 事务所 | 审计测试程序和复杂税务工作流执行 | 生产环境部署 | 压缩客户会计业务的繁忙季复核周期 | 生产环境部署时长和续约历史未公开 |
| MarksNelson LLC | 中型 CPA 事务所 | 客户会计服务和自动分录暂存 | 生产环境部署 | 缓解人员瓶颈,加快月末结账交付 | 留存历史和税季后的软件使用情况未经验证 |
| Pinion LLC | 全国农业 / 食品 CPA 事务所 | 多业务会计和税务合规自动化 | 生产环境部署 | 尽管区域 CPA 招聘严重短缺,仍扩大事务所服务能力 | 软件净支出和多年合同承诺未披露 |
| Whitcomb & Sheridan LLP | 精品 CPA 事务所 | 税务试算平衡表映射和 K-1 合伙企业汇总 | 生产环境部署 | 管理合伙人称每周节省 10 至 20 小时 | 单一合伙人证言,难以外推至大型审计事务所 |
枚举覆盖 Series B 报道和供应商材料中识别出的所有公开命名客户会计事务所;保密的前 25 大客户仍未披露。
[CU001, CU002, CU005, CU009]6.4 客户旅程、扩张动力与留存风险
Basis 的长期商业耐久性取决于其在事务所各业务线的 land-and-expand 轨迹,以及结果导向定价模式能否持续。事务所通常先在 CAS 的月结调节中启动合作,再扩展到季节性纳税申报准备和审计测试。与按席位收费或加价转售原始模型 token 的主流软件供应商不同,Basis 按完成的会计结果定价,让客户支出直接对应事务所交付收入。但客户集中度和长期留存仍有重大尽调风险。会计合伙人要为已提交的纳税申报和认证审计底稿承担个人监管及法律责任,任何代理幻觉或流程错误都会带来软件代码生成工具不会面对的严重责任风险。此外,Basis 既未披露 ARR,也未披露 NRR 或总体续约率等客户留存指标,因此前 25 大事务所的早期渗透究竟代表可持续的多年承诺,还是波动的季节性试用软件使用,仍没有答案。[CU007, CU011, CU012, CU013, CU014]
从发现 CPA 人员约束,到全所流程固化和成果续约的六阶段采用生命周期。
[CU003, CU007, CU011]6.5 展示项
07风险
7.1 监管与专业执业责任
在持牌会计、税务和审计业务中部署自主 AI 代理,会引入复杂的法定责任和专业治理层。在美国,AICPA、州会计委员会和 IRS Circular 230 规范的专业标准,把不可转让的法律责任放在持牌专业人员身上。Basis 可自动创建工作底稿、映射试算平衡表和准备 Form 1065,但软件供应商不能承担专业鉴证或申报责任。企业会计事务所合伙人要求严格的人在回路复核政策,因为监管框架和客户都要求对自动化输出保留书面监督。Sikich 等事务所公开免责声明也明确排除因第三方依赖 AI 生成内容而产生的责任。合规还需要严格的数据治理;Basis 通过 SOC 2 Type II 认证及合同保证,确保客户财务记录留在美国区域且不用于训练基础模型。[CR001, CR002, CR003, CR004]
| 规则 / 许可 / 标准 | 司法辖区 | 状态 | 可能性 | 严重程度 | 缓释机制 | 残余风险 | 尽调路径 |
|---|---|---|---|---|---|---|---|
| IRS Circular 230 税务编制人员标准 | 美国(联邦) | 有效法定要求 | 高 | 关键 | 代理起草工作底稿和 Form 1065/1040,但提交前必须由持证执业人员签字。 | 签字合伙人仍需对自动计算异常引发的法定编制人员处罚负责。 | 审查事务所层面的质量控制政策,以及税务申报的签字审计轨迹样本。 |
| AICPA / PCAOB 审计文档标准 | 美国(鉴证) | 有效专业标准 | 中 | 高 | 每项输出都将数字引用直接链接至来源文件,并配有不可变审计日志。 | 如果工作底稿证据缺乏充分的人类监督记录,将受到同行复核审查。 | 检查工作底稿导出是否符合 AICPA AU-C Section 230 文档要求。 |
| 州会计委员会执业法 | 美国多州许可 | 有效州法规 | 低 | 高 | 软件定位为支持持证 CPA 的生产力平台,而非无证执业。 | 州层面开始调查无证实体提供自主鉴证交付物的问题。 | 跟踪 NASBA 模范规则工作组关于专业鉴证服务中人工智能的更新。 |
| SOC 2 Type II 安全与客户数据保密 | 美国(商业) | 已通过认证 | 低 | 中 | 每年进行第三方 SOC 2 Type II 审计,美国区域数据隔离,不使用客户数据训练模型。 | 如果外部模型 API 端点遭到未授权暴露,子处理方可能发生数据泄漏。 | 核实最新 SOC 2 Type II 审计报告、渗透测试结果和子处理方 DPA 条款。 |
事实型风险登记表,梳理美国会计、税务和隐私法规下的监管边界与专业责任边界。
[CR001, CR002, CR003, CR004]7.2 运营执行、模型路由与验证控制
会计自动化的运营可靠性与消费软件或开发者软件根本不同,因为财务计算容不得无声错误传播。Basis 设计了能在会计系统中持续运行数小时的长链路自主代理,包括处理需要复杂 K-1 脚注提取和附表调节的 Form 1065 合伙企业申报。为减少推理幻觉,架构加入多智能体监督层,把来源引用追溯到客户文档。不过,行业独立报道指出,会计师事务所扩大 AI 软件预算的速度明显快于建立正式监督框架的速度。在异构企业 ERP 环境中执行复杂多步骤结账时,如果事务所复核制度没有强制二次验证,未经核验的边界分类或差异异常可能一路传导进最终总账。[CR005, CR006, CR007, CR008]
| 失效模式 | 可能性 | 严重程度 | 缓释成熟度 | 残余风险 | 未解决的尽调缺口 |
|---|---|---|---|---|---|
| 多步骤计算静默漂移 | 中 | 高 | 中 | 跨总账和税务表的复杂数小时代理工作流,可能让细小分类错误层层累积。 | 未公开自动差异捕获率与人工高级复核人员对比的基准数据。 |
| ERP 架构不一致与字段错位 | 高 | 中 | 高 | 定制化 ERP 科目表和非标准试算表结构需要手动调校流程。 | 需要量化每个新接入客户 ERP 的初始配置工时。 |
| 繁忙报税季的模型 API 延迟峰值 | 中 | 高 | 低 | 季度结账高峰带来并发代理执行负载,而负载取决于上游基础模型 API 的吞吐能力。 | 需要在全国报税截止期提供峰值负载 SLA 保证和备用推理端点。 |
| 企业治理与监督不足 | 高 | 中 | 低 | 会计师事务所扩充 AI 软件预算的速度快于其建立审查流程和审阅者培训的速度。 | 需要客户提供代理执行后复核拒绝率和人工调整频率等标准化指标。 |
自主代理失效模式、多步骤错误传播和企业安全边界分析。
[CR005, CR006, CR007, CR008]展示核心技术、监管、合作伙伴和商业风险类别的剩余风险敞口与缓解成熟度矩阵。
矩阵评分综合经核验的架构披露和监管框架中的定性尽调发现。
[CR002, CR005, CR009, CR014]7.3 上游模型集中度与会计事务所依赖
Basis 的架构高度依赖第三方基础模型基础设施,主要是 OpenAI。平台依靠从 OpenAI o3 迁移到 GPT-5 的监督代理,指挥端到端任务委派并评估推理一致性。轻量子代理使用 GPT-4.1 处理低延迟查询,而月末调节等高上下文任务则回到 GPT-5。这样的紧密结合让 Basis 能快速吸收前沿推理进展,却也使平台在报税高峰期暴露于 OpenAI 服务中断、API 速率限制和模型行为漂移。客户侧采用同样高度集中在头部事务所:Basis 报告称已覆盖美国前 25 大事务所约 40%,Sikich 等关键客户在数千名团队成员中部署工具;但独立评估指出,事务所层面的渗透不等于所有业务线都普遍使用席位。[CR009, CR010, CR011, CR012]
| 依赖方 / 交易对手 | 交易对手角色 | 集中度 | 失败情景 | 严重程度 | 缓解机制 | 剩余风险敞口 |
|---|---|---|---|---|---|---|
| OpenAI(GPT-5、o3 系列、GPT-4.1) | 核心推理基础模型 API 提供商 | 关键集中 | API 服务中断、代币价格大幅上涨,或针对金融工作流的政策限制变化。 | 关键 | 多模型路由架构已部署在轻量级子代理之间,但监督层仍依赖 GPT-5。 | 监督推理代理在不同模型家族之间迁移,需要大量提示词和基准测试重构。 |
| 会计 ERP 平台(Intuit、Sage、NetSuite) | 目标客户的交易及总账系统记录 | 高度集中 | 老牌会计软件供应商限制 API 访问、施加激进的速率限制,或构建原生竞品代理。 | 高 | 标准化流程集成,以及浏览器 / 系统代理自动化层。 | 突然修改合作伙伴 API 政策,或防御性提高平台费用的风险。 |
| 顶级会计师事务所合作伙伴(Sikich 与 Top 25) | 企业早期采用者和 CAS 收入支柱 | 高度集中 | 顶级事务所取消试点、高管更迭,或无法从试点团队扩展到全所部署。 | 中 | 已覆盖 Top 25 事务所中约 40%,并向 Top 150 区域性事务所拓展。 | 使用强度数据仍未披露;如果部署集中在孤立的业务组,风险敞口会随之上升。 |
| 云基础设施托管商(美国数据区域) | 计算基础设施和企业存储托管商 | 中度集中 | 季末或月末企业结账周期发生区域性云故障。 | 中 | 多区域美国托管配置,以及经 SOC 2 Type II 审计的灾难恢复流程。 | 月末结账同步高峰期间可能暂时积压工作流执行。 |
评估对基础模型 API 的依赖、企业 ERP 集成脆弱性和客户集中度。
[CR009, CR010, CR011, CR012]7.4 商业模式可持续性与尽调否决条件
从商业模式看,Basis 按已完成会计交付物收费,而不是采用传统席位授权或原始 token 转嫁。这种方式让客户激励与可量化的自动化价值一致,但也增加执行摩擦:核验交付物完成情况需要准确归因客户,并保持代理完成率稳定。纯 token 消耗模式会随着模型 API 价格下降而压缩利润;结果导向合同则要求技术表现持续达标,才能支撑溢价。投资者在 Basis 完成 1 亿美元 B 轮、估值 11.5 亿美元后评估公司时,应跟踪明确的论点失效触发器,包括强制性报税季审计后企业合同流失、合伙人反对 AI 错误赔偿条款,以及大型会计软件厂商设置防御性 API 门槛或推出原生代理竞争。[CR013, CR014, CR015]
7.5 展示项
08估值
8.1 投资正论与反论
Basis 的看多投资逻辑建立在专业会计的结构性长期顺风上:合格注册会计师严重短缺,同时 CAS、税务合规和审计准备存在大量重复人工合规工作。Basis 构建了专门执行端到端多步骤流程的自主智能代理架构,而非通用对话式聊天机器人,并利用 OpenAI GPT-5、o3、o3-Pro 和 GPT-4.1 等领先基础模型。公司按已完成交付物收费,而非按席位授权或 token 用量计费,使软件支出直接对应事务所账单收入和毛利率扩张。前 25 大美国会计师事务所约 30% 的合作关系(包括前 100 大 IPA 事务所 Sikich)为早期企业验证提供了支撑。\n+\n+\n+ 反过来,看空逻辑集中在脱离已披露经常性收入基本面的极端估值倍数,以及执行与模型责任风险。Basis 虽于 2026 年 2 月完成 1 亿美元 B 轮、估值 11.5 亿美元,但没有公开披露 ARR、总体留存、NRR 或毛利率。会计工作流还受到严格监管标准、纳税申报错误责任和复杂复核要求约束;合伙企业申报或财务报表中的任何无声幻觉或映射失败,都可能让事务所停止合作。过去五至七年中,早期风险投资支持的会计自动化创业公司曾遭遇严重客户质疑,并在实现可持续工作流委派前陷入停滞;如果采用无法扩展到事务所范围的合同承诺,估值倍数压缩风险不容忽视。 [CV001, CV002, CV003, CV004, CV005, CV006, CV007, CV008]
| 分析维度 | 投资论点(乐观视角) | 反论点(悲观视角) | 哪些事实会改变判断 |
|---|---|---|---|
| 市场需求与劳动力动态 | CPA 结构性短缺,加上近 30 万人退休,带来不可自由裁量的企业自动化需求。 | 事务所存在文化惯性,长期不信任自动化核心判断工作的软件。 | 审计确认全所采用软件,并确认 Top 100 事务所的席位数 / 工作量部署持续扩大。 |
| 产品与模型架构 | 自研多代理编排结合 OpenAI GPT-5、o3 和 o3-Pro,可使事务所效率提升 20–50%。 | 底层基础模型实验室可能吸收专用推理能力,或将工作流封装层商品化。 | 通过专有系统集成和专利 / IP 申请,证明技术防御性。 |
| 商业模式与定价 | 基于成果定价,直接捕获自动化工作的经济价值,而不是压低代币加价。 | 复杂工作流中的成果归因很难衡量;相比按席位收费,销售速度更慢。 | 审计后的净收入留存率,以及基于成果计费的多年合同续约数据。 |
| 企业牵引力 | 已部署于美国 Top 25 会计师事务所中约 30%,包括 IPA Top 100 事务所 Sikich。 | 试点合作可能范围有限、合同金额承诺不足,尚未证明能够扩展到全所。 | 已验证的年度经常性收入规模超过 $30M,以及平均合同价值数据。 |
评估推动投资决策的核心分析维度;需要对收入运行率进行事实核验。
[CV003, CV005, CV006, CV007, CV008, CV014]连接市场机会、企业级证明、估值门槛和“继续研究”建议的逻辑决策链。
| 节点 / 连接 | 节点 / 起点 | 终点 | 说明 |
|---|---|---|---|
| 节点 1 | 劳动力短缺:会计人才严重不足,创造不可自由裁量的需求 [market-tailwinds] | ||
| 节点 2 | 代理验证:多代理 GPT-5 架构已部署于约 30% 的 Top 25 CPA 事务所 [technical-proof] | ||
| 节点 3 | 成果计费:与 Sikich 等 IPA 100 领导者的部署建立企业立足点 [commercial-traction] | ||
| 节点 4 | 倍数偏高:$1.15B B 轮投后估值意味着 ARR 倍数 >50 倍 [valuation-hurdle] | ||
| 节点 5 | 执行风险:历史初创失败、未经核实的 ARR 和倍数压缩 [downside-risks] | ||
| 节点 6 | 结论:等待 ARR 和留存指标审计确认后再继续研究 [recommendation-call] | ||
| 连接 1 | 劳动力短缺:会计人才严重不足,创造不可自由裁量的需求 [market-tailwinds] | 代理验证:多代理 GPT-5 架构已部署于约 30% 的 Top 25 CPA 事务所 [technical-proof] | 推动采用 |
| 连接 2 | 代理验证:多代理 GPT-5 架构已部署于约 30% 的 Top 25 CPA 事务所 [technical-proof] | 成果计费:与 Sikich 等 IPA 100 领导者的部署建立企业立足点 [commercial-traction] | 促成变现 |
| 连接 3 | 成果计费:与 Sikich 等 IPA 100 领导者的部署建立企业立足点 [commercial-traction] | 倍数偏高:$1.15B B 轮投后估值意味着 ARR 倍数 >50 倍 [valuation-hurdle] | 激进定价 |
| 连接 4 | 倍数偏高:$1.15B B 轮投后估值意味着 ARR 倍数 >50 倍 [valuation-hurdle] | 执行风险:历史初创失败、未经核实的 ARR 和倍数压缩 [downside-risks] | 放大敞口 |
| 连接 5 | 执行风险:历史初创失败、未经核实的 ARR 和倍数压缩 [downside-risks] | 结论:等待 ARR 和留存指标审计确认后再继续研究 [recommendation-call] | 需要尽调 |
代表投资委员会的定性综合判断;不是量化财务计算。
[CV003, CV004, CV007, CV011, CV012, CV015]8.2 融资背景与资本结构
Basis 自 2023 年成立以来,通过三轮机构风险投资累计融资 $138 million。资本化历程包括 2023 年 10 月 的 $3.6 million 种子轮、2024 年 12 月的 $34 million A 轮,以及 2026 年 2 月 24 日宣布的标志性 $100 million B 轮融资。B 轮由顶级风险投资机构 Accel 领投,GV (Google Ventures)、现有领投方 Khosla Ventures 及前 Goldman Sachs 首席执行官 Lloyd Blankfein 大额跟投。本轮融资对应的投后估值 为 $1.15 billion,使公司在成立三年内跻身独角兽。 从入场纪律看,$1.15 billion 的估值基准给潜在成长期投资者设下了很高的门槛。在 ARR 未披露的情况下, 即便采用私募市场对年化软件收入 $10 million 至 $20 million 的乐观估算,入场倍数仍达 ARR 的 57.5x 至 115x。相比之下,上市垂直软件和税务合规同业的企业价值收入中位数倍数为 6x 至 12x,这一价格 溢价显著。资本表虽显示有高质量投资人组成的联合体支持、账上弹药充足,但后续清算优先权压力以及 $100 million 注资带来的稀释,要求收入高速复合增长并长期维持软件毛利率,才能避免后续融资降价轮 或结构化二级折价。 [CV001, CV002, CV004, CV009, CV010, CV011, CV012, CV013]
| 分析维度 | 状态 / 指标 | 估值背景 | 决策含义 |
|---|---|---|---|
| 建议 | 继续研究 | 投后 B 轮估值为 $1.15B;ARR 和单位经济性未披露 | 在审计数据确认 ARR 规模和留存率之前,暂缓投资 |
| 信心 | 中 | 高置信度的技术验证抵消不了不透明的财务指标 | 有足够证据识别管理层,但不足以支持估值判断 |
| 风险评级 | 高 | 隐含倍数极高(ARR 的 >50 倍),执行责任重 | 如果成果变现落后于试点扩张,存在下轮融资估值下调风险 |
| 估值立场 | 偏高 | 企业倍数显著高于垂直软件基准 | 估值已计入截至 2028 年的无瑕疵执行和超高速增长 |
总结截至 2026 年 10 月的核心建议维度、信心评级、风险评级和决策含义。
[CV001, CV011, CV012, CV015]8.3 情景估值建模与退出展望
为评估公开证据是否支持当前 $1.15 billion 估值,我们综合多空及基准运营情景,预测至 FY2028。 在乐观情景下,Basis 成功加深在美国排名前 150 家会计师事务所中的渗透,将结果导向定价从最初 的 CAS 试点拓展至全所税务工作底稿和审计部署。按此情景,Basis 到 2028 年 ARR 达到 $100 million, 软件毛利率为 75%,若上市或被 Intuit、Thomson Reuters 等财务软件龙头战略收购,按 25x 收入倍数 估值可达 $2.5 billion。 基准情景下,Basis 以稳健但克制的速度获得企业客户,到 2028 年 ARR 达到 $45 million;会计师事务所 保留人工复核环节,并将自主代理限制在结构化 CAS 对账和试算平衡表映射。按市场标准的远期 ARR 16x 倍数估值,企业价值为 $720 million,低于 2026 年 2 月的投后估值,意味着倍数明显压缩,退出估值 持平甚至下行。悲观情景下,客户持续存疑、基础模型商品化,或围绕自主报税的监管审查限制企业推广, ARR 被困在 $18 million。按 8x 倍数计算,企业价值仅 $144 million,较 B 轮峰值缩水 87%。因此, 我们仍认为估值偏高,建议继续研究,等待经过核验的商业 ARR 审计数据。 [CV005, CV006, CV007, CV011, CV012, CV013, CV014, CV015]
| 情景 | 概率 | FY2028 ARR 预估(USD M) | 隐含 ARR 倍数 | 目标估值(USD M) | 关键运营假设与下行触发因素 |
|---|---|---|---|---|---|
| 乐观情景 | 25% | 100 | 25x | 2500 | 覆盖 Top 150 CPA 事务所的全所推广;高成果变现;维持溢价倍数 |
| 基准情景 | 50% | 45 | 16x | 720 | CAS 和税务底稿业务稳步扩张;倍数压缩至 B 轮估值以下的成熟 SaaS 水平 |
| 悲观情景 | 25% | 18 | 8x | 144 | 工作流责任风险、试点流失或基础模型竞争压力导致增长停滞 |
情景模型反映前瞻性估计;基准情景估值意味着相对于当前 $1.15B 投后估值,融资下调时倍数会被压缩。
[CV001, CV007, CV011, CV012, CV013, CV014, CV015]8.4 展示材料
免责声明
本报告是基于公开证据的尽调快照,不构成投资建议。重要的财务、法律、技术和合同事实仍未公开,任何投资决策前都应直接向管理层及一手文件核实。
证据索引
| 编号 | 陈述 | 可信度 | 来源 |
|---|---|---|---|
| CO001 | Basis was founded in 2023 by Matthew Harpe and Mitchell Troyanovsky, establishing corporate headquarters in New York City. | 高 | SO003, SO004, SO007, SO009 |
| CO002 | Basis operates as a growth-stage private developer of agentic artificial intelligence systems designed to automate end-to-end accounting, tax, and audit workflows. | 高 | SO001, SO002, SO004, SO005 |
| CO003 | Basis raised $100 million in Series B financing in February 2026 at a post-money valuation of $1.15 billion. | 高 | SO003, SO004, SO008, SO009 |
| CO004 | Basis cumulative venture capital funding reached approximately $138 million across a $3.6 million seed round, a $34 million Series A, and the $100 million Series B. | 中 | SO006, SO009, SO010 |
| CO005 | The Series B round was led by Accel, with participation from GV, Khosla Ventures, and former Goldman Sachs CEO Lloyd Blankfein. | 高 | SO004, SO008, SO009, SO010 |
| CO006 | Basis appointed Miles Clements of Accel and Keith Rabois of Khosla Ventures to its board of directors following the Series B round. | 中 | SO009 |
| CO007 | Basis platform has been adopted across approximately 30% of the top 25 U.S. accounting firms, including Boulay, Clark Nuber, MarksNelson, Pinion, and UHY. | 高 | SO001, SO004, SO007, SO010 |
| CO008 | Basis coordinates specialized multi-agent architectures leveraging OpenAI models including GPT-4.1, GPT-5, o3, and o3-Pro to conduct complex multi-hour accounting procedures. | 高 | SO002, SO005 |
| CO009 | Basis demonstrated what it reported as the first AI agent capable of autonomously completing an end-to-end Form 1065 partnership tax return. | 中 | SO003, SO009, SO010 |
| CO010 | Accounting firms deploying Basis report measurable efficiency gains ranging from 20% to 50% across practice workflows. | 中 | SO001, SO005, SO006, SO009, SO010 |
| CO011 | The accounting automation sector faces historical market skepticism stemming from prior software vendors overpromising automation capabilities across complex legacy environments. | 中 | SO007 |
| CO012 | Co-founder Matthew Harpe brings a background in healthcare analytics from Boston Consulting Group and SoftBank, while Mitchell Troyanovsky leads core system engineering. | 中 | SO007, SO010 |
| CO013 | Basis commercial model uses outcomes-based pricing tied to completed accounting work rather than conventional seat-based or token-consumption pricing. | 中 | SO007 |
| CO014 | Verifiable figures for Basis annual recurring revenue (ARR), gross margin percentage, net revenue retention, and active customer account totals remain undisclosed. | 中 | SO003, SO004, SO007 |
| CO015 | Basis maintains an estimated workforce of 51 to 100 employees, with core engineering and machine learning teams concentrated in New York City. | 中 | SO006, SO007 |
| CM001 | Basis defines its market scope around agentic automation of multi-step accounting workflows across client accounting services, tax, and audit practices. | 中 | SM001, SM002 |
| CM002 | The addressable accounting market in the United States comprises approximately 3 million practicing accountants across public practice and corporate finance. | 中 | SM007 |
| CM003 | A structural labor shortage has led to roughly 300,000 accountant retirements within a two-year period, with approximately 75% of the existing workforce projected to retire over the next decade. | 中 | SM007 |
| CM004 | Basis reports penetration into approximately 30% of the top 25 US accounting firms and 25% of the top 150 firms. | 高 | SM004, SM007 |
| CM005 | Top accounting firm buyers include major mid-market and enterprise practices such as Sikich, Wiss, Citrin Cooperman, UHY, Clark Nuber, Boulay, Schneider Downs, Pinion, and Kaufman Rossin. | 中 | SM001, SM002 |
| CM006 | Accounting firms adopting autonomous workflow agents report average time savings of up to 30%, expanding their capacity to service new accounts without proportional headcount additions. | 中 | SM005 |
| CM007 | Basis employs an outcomes-based pricing structure where firms pay based on completed accounting deliverables rather than traditional per-seat or token-consumption billing. | 中 | SM007 |
| CM008 | Basis provides more than 350 pre-built accounting procedures configured by accountants in residence to standardize recurring month-end close and tax preparation tasks. | 中 | SM003 |
| CM009 | In February 2026, Basis raised $100 million in Series B financing at a $1.15 billion valuation led by Accel and GV to accelerate agentic workflow deployments. | 高 | SM004, SM006, SM008 |
| CM010 | Enterprise accounting practices frequently operate under alternative practice structures that separate licensed attest CPA partnerships from tax and advisory entities. | 中 | SM009 |
| CM011 | To satisfy strict professional verification and audit standards, AI accounting agents must attach inspectable audit logs, source citations, calculations, and explicit confidence scores to every proposed entry. | 高 | SM002, SM005 |
| CM012 | Traditional horizontal ERP and bookkeeping software vendors remain constrained by heavily manual data entry processes that reduce practitioner capacity for strategic advisory services. | 中 | SM008 |
| CM013 | Adoption of agentic platforms in complex tax workbooks requires high-reasoning models capable of navigating 50-tab workbooks and completing multi-step partnership filings. | 中 | SM001, SM006 |
| CM014 | Commercial disclosures for Basis continue to omit audited ARR, net revenue retention, and gross margins, leaving open questions regarding commercial unit economics. | 低 | |
| CP001 | Basis reports customer penetration across approximately 30 percent of the top 25 United States accounting firms across CAS, Tax, and Audit practices. | 高 | SP001, SP004 |
| CP002 | Basis provides a library of over 350 pre-built accounting Procedures created by accountants in residence to automate standardized workflows. | 中 | SP002, SP003 |
| CP003 | The Basis platform maintains SOC 2 Type II, ISO 27001, and ISO 42001 certifications, US infrastructure data hosting, and zero model training on client records. | 中 | SP002, SP003 |
| CP004 | Agentic AI platforms capable of multi-step autonomous planning and execution represent one of the most sought-after categories among enterprise venture capital investors in 2026. | 中 | SP004 |
| CP005 | The Basis architecture coordinates tasks through a supervising agent that delegates to specialized sub-agents based on complexity, latency, and input modality. | 高 | SP005, SP007 |
| CP006 | Basis wants to deploy autonomous agents that work over multi-hour horizons on complex accounting tasks across CAS, tax, and audit. | 中 | SP001, SP006 |
| CP007 | Basis differentiates from foundation model labs by addressing specialized general ledger integrations and regulatory audit trails rather than general text, search, and document creation. | 中 | SP007 |
| CP008 | Basis adopted an outcomes-based pricing philosophy linking customer costs to completed accounting work rather than token consumption or user seats. | 中 | SP007 |
| CP009 | Thomson Reuters research surveying over 1,500 professionals indicates that professional services entered a strategic phase in 2026 where firms restructure workflows around generative and agentic AI. | 中 | SP009 |
| CP010 | Basis raised a $100 million Series B funding round at a $1.15 billion valuation in February 2026, bringing cumulative funding to approximately $138 million. | 高 | SP001, SP008 |
| CP011 | Basis achieves adoption across approximately 25 percent of the top 150 accounting firms in the United States according to technical teardown disclosures. | 中 | SP007 |
| CP012 | Basis demonstrated the first AI agent capable of autonomously completing an end-to-end Form 1065 partnership tax return deliverable. | 高 | SP001, SP008 |
| CP013 | Accounting firms adopting Basis capture procedural lock-in as custom firm standard operating procedures and client-specific categorization rules are codified into the system. | 中 | SP003, SP007 |
| CP014 | Third-party directory evaluations by RECATOOLS indicate that Basis maintains no public list pricing and manages enterprise onboarding exclusively via demo requests and waitlists. | 中 | SP002 |
| CP015 | RECATOOLS assessment notes that Basis hosts client data strictly within US infrastructure regions and constructs primary case studies around US-specific tax and audit structures. | 中 | SP002 |
| CP016 | Basis relies on underlying foundation models including OpenAI GPT-5 and o-series models, exposing the platform to upstream API dependency and model cost shifts. | 中 | SP005 |
| CI001 | Basis monetizes its platform through an outcomes-based pricing architecture where accounting firms pay for completed accounting deliverables rather than user seats or raw compute tokens. | 中 | SI003 |
| CI002 | Basis packages over 350 pre-built accounting procedures developed by in-house accountants in residence across tax, client accounting services, and audit workflows. | 中 | SI001 |
| CI003 | Basis charges accounting firms based on firm-wide practice deployments rather than individual per-seat licenses, aligning contract value with autonomous deliverable volume. | 中 | SI003, SI004 |
| CI004 | Basis routes workflow steps across a multi-model stack including OpenAI o3, o3-Pro, GPT-4.1, and GPT-5, creating a direct cost of goods sold linked to foundation model inference. | 高 | SI001, SI005 |
| CI005 | Deploying Basis generates an average of 30 percent time savings across repetitive reconciliations, journal entries, and financial statement reviews at adopting accounting firms. | 中 | SI005 |
| CI006 | Basis's autonomous agents can operate independently for hours on complex multi-step processes including trial balance mapping and end-to-end partnership tax returns. | 中 | SI007, SI008 |
| CI007 | Basis raised a $100 million Series B financing round in February 2026 led by Accel and Google Ventures that valued the company at a post-money valuation of $1.15 billion. | 高 | SI002, SI007, SI008 |
| CI008 | Basis has accumulated roughly $138 million in total lifetime venture capital funding across its seed, Series A, and Series B rounds. | 中 | SI006, SI009 |
| CI009 | Basis employs approximately 120 staff members concentrated entirely in its New York City offices, representing its primary operational expenditure driver. | 中 | SI004 |
| CI010 | Basis's platform has been adopted across approximately 30 percent to 40 percent of the top 25 U.S. accounting firms and roughly 25 percent of the top 150 firms. | 高 | SI002, SI003, SI004 |
| CI011 | Top 50 accounting firm Sikich publicly partnered with Basis to deploy agentic artificial intelligence technology across its accounting, tax, and audit practices. | 中 | SI004 |
| CI012 | Historical vertical accounting software vendors experienced widespread adoption friction and failed commercial automation attempts over the five to seven years preceding Basis. | 中 | SI003 |
| CI013 | Basis resisted industry pressure to adopt seat-based or token-markup pricing, accepting slower initial revenue growth to establish measurable deliverable attribution. | 中 | SI003 |
| CI014 | Basis's forward gross margins remain highly sensitive to multi-model inference expenses and customer onboarding costs associated with its Deployed Intelligence engineering team. | 中 | SI003, SI004, SI005 |
| CE001 | Basis operates an agentic AI platform that automates client accounting services, tax preparation, and audit workflows for accounting firms under practitioner supervision. | 高 | SE001, SE003, SE005 |
| CE002 | The platform core architecture routes accounting tasks across OpenAI models including GPT-5, o3, o3-Pro, and GPT-4.1 coordinated by a supervising agent. | 高 | SE002, SE003 |
| CE003 | Basis agents operate within deterministic rules-based accounting constraints rather than reasoning freely, keeping generated outputs predictable and auditable. | 中 | SE003 |
| CE004 | Basis agents generate immutable action logs and cite every figure back to its source document, requesting human review when confidence falls below configurable thresholds. | 中 | SE001, SE004 |
| CE005 | Accounting firms utilizing Basis report saving up to 30 percent of practitioner time on automated month-end close and reconciliation tasks. | 中 | SE002, SE003 |
| CE006 | The platform includes over 350 pre-built accounting Procedures designed by in-house accountants in residence to standardize recurring tax and accounting workflows. | 中 | SE001 |
| CE007 | Basis demonstrated an autonomous agent capable of preparing a complete Form 1065 partnership tax return from underlying financial records. | 中 | SE006 |
| CE008 | The platform client accounting services module stages proposed journal entries with account codes, memos, and linked source documents for accountant review prior to ledger posting. | 中 | SE001 |
| CE009 | Basis uses a supervising agent built on GPT-5 to decompose complex accounting goals into discrete sub-agent tasks based on complexity and latency requirements. | 中 | SE002 |
| CE010 | On internal tool-calling benchmarks evaluating parallel tool execution with code interpreter and web search, GPT-5 achieved a 100 percent success rate in coordinating structured actions. | 中 | SE002 |
| CE011 | Basis connects to accounting firm ledgers and ERP systems in the cloud without storing client credentials on its servers. | 中 | SE003 |
| CE012 | Basis architecture combines global foundation model reasoning with local learning that crystallizes entity-specific categorizations and accounting conventions. | 中 | SE007 |
| CE013 | Industry framework analyses highlight that agentic enterprise value depends on redesigning accounting workflows and embedding mandatory human checkpoints where accountability matters. | 中 | SE008 |
| CE014 | Basis autonomous architecture relies heavily on upstream AI model precision and lacks public self-serve APIs, making it less adaptable for non-standard tax situations. | 中 | SE003 |
| CU001 | Basis serves approximately 30% of the top 25 accounting firms and 25% of the top 150 accounting firms in the United States. | 高 | SU001, SU004, SU006 |
| CU002 | Publicly confirmed customer accounts include major accounting firms Boulay PLLP, Clark Nuber PS, MarksNelson LLC, Pinion LLC, UHY LLC, and Wiss & Company. | 中 | SU001, SU008 |
| CU003 | Accounting firms deploying Basis agents report up to 30% time savings on average across structured accounting tasks. | 高 | SU003, SU008 |
| CU004 | Reuters confirmed that Basis's enterprise client roster includes approximately seven of the top 25 US accounting firms. | 高 | SU006, SU001 |
| CU005 | Small firm managing partner Todd Whitcomb reported that Basis saves 10 to 20 hours per week by reducing 2-to-3 hour preparation tasks to 30 minutes. | 中 | SU005 |
| CU006 | Basis positions its autonomous agent platform specifically for professional accounting practices across tax, audit, and CAS rather than corporate enterprise accounts payable. | 中 | SU002, SU008 |
| CU007 | Basis customer deployments operate as long-horizon autonomous workflows where human accountants transition from data preparers to reviewers. | 中 | SU007, SU008 |
| CU008 | Basis provides a library of more than 350 pre-built accounting procedures covering trial balance mapping, K-1 extraction, and workpaper generation. | 中 | SU005 |
| CU009 | Accounting firm Wiss & Company described Basis as trajectory-defining in enabling teams to expand higher-value advisory services. | 中 | SU008 |
| CU010 | Pricing for Basis accounting firm deployments is enterprise-quoted and custom-structured with no published self-serve list pricing. | 中 | SU002, SU008 |
| CU011 | Basis utilizes an outcomes-based pricing model tied to completed accounting work rather than user-seat subscriptions or raw token consumption. | 中 | SU004 |
| CU012 | Reported efficiency gains of 20% to 50% across practices are self-reported by Basis and partner firms without published third-party audits. | 中 | SU008 |
| CU013 | Unlike software code generation, errors in autonomous accounting workpapers and tax returns directly create regulatory penalties and personal partner liability. | 中 | SU008 |
| CU014 | Basis has not publicly disclosed customer renewal rates, net revenue retention, contract durations, or revenue concentration figures. | 中 | SU008 |
| CR001 | Under professional accounting and tax standards, licensed practitioners remain legally liable for errors in tax returns and audit workpapers regardless of software automation. | 高 | SR002, SR008 |
| CR002 | Enterprise accounting firm partners require clear human-in-the-loop review policies because regulatory bodies and clients mandate documentation of AI verification. | 中 | SR001, SR002 |
| CR003 | Sikich expressly disclaims legal liability for loss arising from reliance on its publications or AI-generated language model content in its advisory disclosures. | 中 | SR008 |
| CR004 | Basis maintains SOC 2 Type II certification and subjects its internal security controls to annual third-party audits. | 中 | SR003 |
| CR005 | Basis does not train underlying artificial intelligence models on customer accounting records, and client data remains hosted in U.S. regions. | 中 | SR003 |
| CR006 | Basis autonomously demonstrated completion of an end-to-end Form 1065 partnership tax return, coordinating multi-step schedules, K-1 generation, and IRS rules. | 中 | SR002, SR006 |
| CR007 | Complex multi-hour autonomous accounting workflows require long-horizon agents to execute cross-system reconciliations, variance analyses, and journal entries before human handoff. | 中 | SR002, SR007 |
| CR008 | Accounting firms report increasing artificial intelligence software budgets faster than they establish formalized governance and oversight review frameworks. | 中 | SR001 |
| CR009 | Basis's core supervisory agent relies on OpenAI frontier reasoning models, migrating from OpenAI o3 to GPT-5 to orchestrate sub-agent routing. | 中 | SR005 |
| CR010 | Sub-agents in the Basis stack use OpenAI GPT-4.1 for latency-sensitive queries while delegating complex multi-step close reconciliations to GPT-5. | 中 | SR005 |
| CR011 | Sikich deploys Basis into its client accounting services delivery model across roughly 2,500 team members to automate ERP-integrated reconciliations and month-end closes. | 高 | SR001, SR008 |
| CR012 | Firm penetration figures reported as 40 percent of the top 25 firms reflect account-level adoption rather than universal day-to-day seat usage across individual practitioners. | 高 | SR001, SR004 |
| CR013 | Basis operates an outcomes-based pricing model tied to completed accounting work rather than passing through raw token-consumption or per-seat fees. | 中 | SR007 |
| CR014 | Token-consumption models expose software vendors to margin compression as foundation model API costs drop and customers demand clear ROI attribution. | 中 | SR007 |
| CR015 | Basis raised $100 million in Series B financing at a $1.15 billion post-money valuation in February 2026 led by Accel with participation from Khosla Ventures and GV. | 高 | SR002, SR004, SR006 |
| CV001 | Basis announced a $100 million Series B funding round on February 24, 2026, valuing the company at a post-money valuation of $1.15 billion. | 高 | SV002, SV004, SV009 |
| CV002 | Basis has raised a cumulative $138 million in equity financing since its inception in 2023 across Seed, Series A, and Series B rounds. | 中 | SV001, SV002 |
| CV003 | The bull investment thesis for Basis is underpinned by severe accounting talent shortages and demographic retirements across US CPA firms. | 中 | SV001, SV007, SV009 |
| CV004 | The Series B round was led by Accel with participation from GV, Khosla Ventures, and former Goldman Sachs CEO Lloyd Blankfein. | 高 | SV002, SV004, SV009 |
| CV005 | Basis operates an outcomes-based pricing model where accounting firms pay for completed accounting workflows rather than per-seat licenses. | 中 | SV007 |
| CV006 | Basis leverages foundation models including OpenAI GPT-5, o3, o3-Pro, and GPT-4.1 to orchestrate autonomous multi-agent accounting workflows. | 高 | SV001, SV005 |
| CV007 | Basis platform deployments are reported across approximately 30% of the top 25 US accounting firms and 25% of the top 150 firms. | 中 | SV001, SV002, SV007 |
| CV008 | Historical venture-backed accounting automation attempts over the prior five to seven years faced persistent customer skepticism after failing to automate complex workflows. | 中 | SV007 |
| CV009 | Basis raised a $3.6 million Seed round in October 2023 and a $34 million Series A round in December 2024 prior to its Series B. | 中 | SV001 |
| CV010 | Basis offers more than 350 pre-built accounting procedures across tax compliance, client advisory services, and general ledger operations. | 中 | SV003 |
| CV011 | The $1.15 billion valuation benchmark implies an estimated private ARR multiple exceeding 50x in the absence of verified public revenue figures. | 中 | SV002, SV004 |
| CV012 | Under a base-case scenario, Basis reaches $45 million in ARR by 2028, implying a $720 million enterprise valuation that compresses below the Series B post-money mark. | 中 | SV004, SV007 |
| CV013 | In a bear-case scenario where enterprise pilot expansion stalls, Basis generates $18 million in ARR by 2028, implying an 87% valuation impairment. | 中 | SV004, SV007 |
| CV014 | Top 100 IPA accounting firm Sikich announced a commercial partnership incorporating the Basis platform into its client accounting services delivery model in July 2026. | 中 | SV008 |
| CV015 | The valuation stance remains stretched and recommendation is research-more due to undisclosed ARR, gross margin, and renewal retention figures. | 中 | SV002, SV004, SV007 |
| 编号 | 出版方 | 标题 | 引文 |
|---|---|---|---|
| SO001 | Basis | Basis | AI Agents for Accounting | Accounting teams cannot keep absorbing more manual work. Basis takes on long-horizon, complex workflows so people can start at review. |
| SO002 | Basis | Basis | About | We build agents that do real accounting work, end to end. These agents run autonomously, sometimes for hours, collaborating with accountants at key decision points and delivering finished work for review. |
| SO003 | Built In NYC | Basis Raises $100M to Deploy Automated Accounting Agents | Built In NYC | Tax auditing software provider Basis recently announced a $100 million Series B funding round at a $1.15 billion valuation. The company will put its new capital from lead investor Accel and others toward further developing its agentic AI platform. |
| SO004 | Reuters via Yahoo Finance | AI accounting startup Basis raises $100 million at $1.15 billion valuation | venture capital firm Accel, with participation from GV, formerly known as Google Ventures, former Goldman Sachs Chief Executive Lloyd Blankfein and existing investor Khosla Ventures. |
| SO005 | OpenAI | Basis scales accounting by turning OpenAI model progress into trusted agents | Founded in 2023, Basis builds AI agents used by top accounting firms—designed to take on structured accounting work with the reliability and depth those tasks require. |
| SO006 | SHORTList | Curated Startup Jobs, Alerted ASAP | SHORTList | Basis has raised a total of $138 million in funding, including a significant $100 million Series B round in February 2026. |
| SO007 | ZenML | Basis: Agentic AI for Automated Accounting Workflows at Scale - ZenML LLMOps Database | The company did encounter skepticism, particularly given the history of accounting technology companies overpromising and underdelivering. There is a pattern of second-time founders who, after working with accounting processes in their first company, think the work should be more efficient and attempt to solve the problem, only to discover that accounting is more complicated than it appears and that startup accounting is much simpler than the accounting at more developed organizations. |
| SO008 | PYMNTS | Basis Raises $100 Million to Build Up AI in Accounting | PYMNTS.com | The company’s Series B round, announced Tuesday (Feb. 24), was led by Accel and Google Ventures and values Basis at $1.15 billion. |
| SO009 | Crowdfund Insider | Basis Announces $100M In New Funding At $1.15B Valuation To Enable AI-Driven Accounting Automation | Crowdfund Insider | This brings Basis’s cumulative funding to roughly $138 million, building on a $3.6 million seed and a $34 million Series A completed in December 2024. |
| SO010 | Accountio | Basis AI raises $100m Series B at $1.15bn valuation - Accountio | Matt Harpe, the company’s CEO and co-founder, said the capital will be used to accelerate platform development, with a particular emphasis on building more advanced agents capable of handling increasingly complex workflows across tax, audit and advisory services. |
| SM001 | Agentic AI Accounting | Basis — Agentic AI Accounting | Basis targets mid-market and enterprise accounting practices seeking to scale operations through AI-powered automation while maintaining human oversight and verification. |
| SM002 | Agentic Index | Basis Features & Pricing: 7.5/14 Score | Agentic Index | Basis builds AI agents that do accounting work end to end for accounting firms and corporate finance teams, across client accounting services, tax and audit, so that accountants start at review instead of preparation. |
| SM003 | Basis | Basis for Small Firms | Built for the firm that does everything | Choose from 350+ pre-built accounting Procedures, built by our accountants in residence. |
| SM004 | Reuters | AI accounting startup Basis raises $100 million at $1.15 billion valuation | Agentic AI, which can plan, decide and act autonomously rather than simply respond to a prompt, has become one of the most sought-after categories in venture capital as investors focus on professional services. |
| SM005 | OpenAI | Basis scales accounting by turning OpenAI model progress into trusted agents | Firms using Basis report 30% time savings on average, and continue expanding agent responsibilities as trust grows. |
| SM006 | Built In NYC | Basis Raises $100M to Deploy Automated Accounting Agents | Built In NYC | The agents will complete end-to-end tasks for firms’ client accounting services, tax and audit practices. |
| SM007 | ZenML | Basis: Agentic AI for Automated Accounting Workflows at Scale - ZenML LLMOps Database | With approximately 300,000 accountants retiring in a two-year period and roughly 75% expected to retire in the next decade, there is an urgent need for AI to augment the remaining workforce. |
| SM008 | PYMNTS | Basis Raises $100 Million to Build Up AI in Accounting | PYMNTS.com | The accounting and bookkeeping industry is at a turning point... Many professionals are overworked and under-resourced, facing increasing client demand and rising pressure from evolving regulatory and compliance requirements. |
| SM009 | Baker Tilly | Baker Tilly collaborates with Basis to expand AI-enabled automation across accounting and finance managed services | Baker Tilly | Baker Tilly US, LLP is a licensed independent CPA firm that provides attest services to clients. Baker Tilly Advisory Group, LP and its subsidiary entities provide tax and business advisory services to their clients. |
| SP001 | 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 to address these challenges by deploying agents that complete complex accounting workflows end-to-end across their CAS, Tax, and Audit practices. |
| SP002 | RECATOOLS | Basis AI — AI Accounting Agents for Firms — Practical Tools | The caveats: there is no public pricing, access is by demo or waitlist, client data is hosted in US regions, and the case studies are built on US tax and audit work. |
| SP003 | Basis | Basis for Small Firms | Built for the firm that does everything | Give Basis the work you can’t delegate and watch it excel. Choose from 350+ pre-built accounting Procedures, built by our accountants in residence. |
| SP004 | Reuters via Yahoo Finance | AI accounting startup Basis raises $100 million at $1.15 billion valuation | Agentic AI, which can plan, decide and act autonomously rather than simply respond to a prompt, has become one of the most sought-after categories in venture capital as investors focus on professional services. |
| SP005 | OpenAI | Basis scales accounting by turning OpenAI model progress into trusted agents | Each task begins with a supervising agent, originally built on OpenAI o3 and now migrated to GPT‑5, which coordinates the full process—routing steps to specialized sub-agents based on task, complexity, latency needs, and input type. |
| SP006 | Built In NYC | Basis Raises $100M to Deploy Automated Accounting Agents | Basis wants to create AI agents that spend hours at a time working on complex accounting workflows and deliver accurate results. |
| SP007 | ZenML | Basis: Agentic AI for Automated Accounting Workflows at Scale | Regarding competition from large labs, Basis argues that the labs naturally focus on areas one degree removed from their core competencies. Labs have successfully attacked products that are proximal to what they already do well: helping generate and edit text in documents, performing internet research, building spreadsheets and presentations. |
| SP008 | Crowdfund Insider | Basis Announces $100M In New Funding At $1.15B Valuation To Enable AI-Driven Accounting Automation | This brings Basis’s cumulative funding to roughly $138 million, building on a $3.6 million seed and a $34 million Series A completed in December 2024. |
| SP009 | Thomson Reuters Institute | 2026 AI in Professional Services Report | Drawing on perspectives from more than 1,500 professionals, this report highlights that the era of early AI adoption has passed. 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. |
| SI001 | Basis | Basis for Small Firms | Built for the firm that does everything | Choose from 350+ pre-built accounting Procedures, built by our accountants in residence. |
| SI002 | Reuters | AI accounting startup Basis raises $100 million at $1.15 billion valuation | Agentic AI, which can plan, decide and act autonomously rather than simply respond to a prompt, has become one of the most sought-after categories in venture capital as investors focus on professional services. |
| SI003 | ZenML | Basis: Agentic AI for Automated Accounting Workflows at Scale - ZenML LLMOps Database | The company took a disciplined approach to commercialization, refusing to launch with a seat-based pricing model or token-consumption model despite pressure to do so. Instead, they held out until they had developed an outcomes-based pricing model where customers pay based on the actual accounting work completed rather than based on the number of users or the computational resources consumed. |
| SI004 | Accounting Today | Basis offers AI agents, love letters and an ode to accounting | The company has all 120 of its employees working in its New York offices, where it competes with AI giants like OpenAI and Anthropic for talent. |
| SI005 | 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. Firms using Basis report 30% time savings on average, and continue expanding agent responsibilities as trust grows. |
| SI006 | StartupHub.ai | Basis - $138M Raised, Reviews & Alternatives | Basis has raised a total of $138M in funding. The most recent round on record is Series B. |
| SI007 | Built In NYC | Basis Raises $100M to Deploy Automated Accounting Agents | Built In NYC | Tax auditing software provider Basis recently announced a $100 million Series B funding round at a $1.15 billion valuation. The company will put its new capital from lead investor Accel and others toward further developing its agentic AI platform. It will also expand its team across engineering and machine learning. |
| SI008 | PYMNTS.com | Basis Raises $100 Million to Build Up AI in Accounting | PYMNTS.com | The company’s Series B round, announced Tuesday (Feb. 24), was led by Accel and Google Ventures and values Basis at $1.15 billion. |
| SI009 | Crowdfund Insider | Basis Announces $100M In New Funding At $1.15B Valuation To Enable AI-Driven Accounting Automation | Crowdfund Insider | This brings Basis’s cumulative funding to roughly $138 million, building on a $3.6 million seed and a $34 million Series A completed in December 2024. |
| SE001 | Basis | Basis for Small Firms | Built for the firm that does everything | Choose from 350+ pre-built accounting Procedures, built by our accountants in residence. |
| SE002 | OpenAI | Basis scales accounting by turning OpenAI model progress into trusted agents | To support that, the team built a multi-agent architecture that assigns the best-fit OpenAI model to the right job. |
| SE003 | TeachUsTechnology | Basis: Agentic AI Accounting for CPA Firms | Each agent operates within a fixed set of rules-based constraints rather than reasoning freely, combining large language models with accounting logic to keep output predictable and auditable. |
| SE004 | Digital Applied | Basis AI $100M: Agentic Accounting Tax and Audit | A Basis agent receives a goal... and autonomously determines the sequence of tasks needed to achieve it, requesting human review only when confidence falls below configurable thresholds. |
| SE005 | Reuters via Yahoo Finance | AI accounting startup Basis raises $100 million at $1.15 billion valuation | Agentic AI, which can plan, decide and act autonomously rather than simply respond to a prompt, has become one of the most sought-after categories in venture capital as investors focus on professional services. |
| SE006 | Built In NYC | Basis Raises $100M to Deploy Automated Accounting Agents | The agents will complete end-to-end tasks for firms’ client accounting services, tax and audit practices. Basis recently demonstrated an AI agent that autonomously completed a 1065 tax return. |
| SE007 | ZenML | Basis: Agentic AI for Automated Accounting Workflows at Scale - ZenML LLMOps Database | Their approach leverages large language models as a foundation but then applies them in local contexts, allowing the systems to crystallize learnings specific to individual entities. |
| SE008 | Cognizant | Modular Framework for Agentic AI - BASIS | Agentic systems change that economic logic: when software can plan and act, the highest returns come from redesigning the workflow itself... and embedding the right human checkpoints where judgment, accountability, and risk truly matter. |
| SU001 | SiliconANGLE | AI accounting startup Basis secures $100M at $1.15B valuation as firms adopt agent-based workflows | Basis counts among its customers approximately 30% of the Top 25 accounting firms that are deploying agents that complete complex accounting workflows end-to-end across their client accounting services, tax and audit practices. Notable customers include Boulay PLLP, Clark Nuber PS, MarksNelson LLC, Pinion LLC and UHY LLC. |
| SU002 | Agentic AI List | Basis – Agentic AI List | Enterprise-oriented, custom-quoted pricing for accounting firms. No self-serve or published list pricing. |
| SU003 | 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. Firms using Basis report 30% time savings on average, and continue expanding agent responsibilities as trust grows. |
| SU004 | ZenML | Basis: Agentic AI for Automated Accounting Workflows at Scale - ZenML LLMOps Database | The company recently raised $100 million at a $1.15 billion valuation and serves accounting teams across approximately 30% of the top 25 accounting firms and 25% of the top 150 firms in the United States. |
| SU005 | Basis | Basis for Small Firms | Built for the firm that does everything | Give Basis the work you can’t delegate and watch it excel. Choose from 350+ pre-built accounting Procedures, built by our accountants in residence. |
| SU006 | Reuters | AI accounting startup Basis raises $100 million at $1.15 billion valuation | The company said it now serves about seven of the top 25 accounting firms in the U.S. |
| SU007 | Built In NYC | Basis Raises $100M to Deploy Automated Accounting Agents | Built In NYC | Basis wants to create AI agents that spend hours at a time working on complex accounting workflows and deliver accurate results. The agents will complete end-to-end tasks for firms’ client accounting services, tax and audit practices. |
| SU008 | Accountio | Basis AI: $1.15B valuation, funding and customers | Named clients include Wiss, UHY, Boulay, Clark Nuber, MarksNelson, and Pinion. The company states efficiency gains of 20% to 50% across practices. Those figures are self-reported and should be treated accordingly. No third-party benchmarks or audited performance data have been published. |
| SR001 | Accountsphere | Basis AI Agents Now Run Inside 40 Percent of the Top 25 Accounting Firms | Accounting Today's reporting does not include usage-intensity data, seat counts, or how many workflows within each firm actually route through Basis agents versus traditional software. |
| SR002 | ai2.work | Basis Raises $100M at $1.15B Valuation to Automate Accounting (2026) | Set clear policies for AI agent oversight-define which outputs require senior review, how audit trails are maintained, and what quality assurance checkpoints exist. |
| SR003 | Basis | Basis for Small Firms | Built for the firm that does everything | Security controls undergo an annual audit. |
| SR004 | Reuters via Yahoo Finance | AI accounting startup Basis raises $100 million at $1.15 billion valuation | Basis runs an AI agent platform for accountants and says its systems learn client-specific needs and work independently across multi-step accounting tasks. The company said it now serves about seven of the top 25 accounting firms in the U.S. |
| SR005 | OpenAI | Basis scales accounting by turning OpenAI model progress into customer value | Each task begins with a supervising agent, originally built on OpenAI o3 and now migrated to GPT‑5, which coordinates the full process—routing steps to specialized sub-agents based on task, complexity, latency needs, and input type. GPT‑5 is the strongest model Basis has evaluated to date in reasoning, consistency, and explainability, making it well-suited to guide agents across high-context workflows with minimal oversight. |
| SR006 | Built In NYC | Basis Raises $100M to Deploy Automated Accounting Agents | Tax auditing software provider Basis recently announced a $100 million Series B funding round at a $1.15 billion valuation. The company will put its new capital from lead investor Accel and others toward further developing its agentic AI platform. |
| SR007 | ZenML | Basis: Agentic AI for Automated Accounting Workflows at Scale | The outcomes-based pricing model, while difficult to implement, creates strong alignment and forces the technology to deliver measurable value rather than just plausible activity. |
| SR008 | Sikich | Sikich partners with Basis to advance AI-enabled accounting services | Sikich is incorporating the Basis platform into its client accounting services (CAS) delivery model. The platform integrates with existing enterprise resource planning (ERP) systems and automates key components of month-end closing, transaction processing and reconciliation workflows, helping reduce manual effort and improve consistency. |
| SV001 | AI Market Watch | Basis - AI Startup Profile | AI Market Watch | Delivers 20-50% efficiency improvements by automating complex accounting workflows end-to-end, addressing the critical accounting talent shortage. |
| SV002 | CPA Practice Advisor | Basis Raises $100 Million to Deploy AI Agents for Accounting Firms | Founded by Matthew Harpe and Mitchell Troyanovsky, the New York City-based startup has raised $138 million thus far and said its platform is used by approximately 30% of the top 25 accounting firms, deploying agents that complete complex accounting workflows end to end across their client advisory services, tax, and audit practices. |
| SV003 | Basis | Basis for Small Firms | Built for the firm that does everything | Give Basis the work you can’t delegate and watch it excel. Choose from 350+ pre-built accounting Procedures, built by our accountants in residence. |
| SV004 | Reuters / Yahoo Finance | AI accounting startup Basis raises $100 million at $1.15 billion valuation | Feb 24 (Reuters) - AI accounting startup Basis has raised $100 million in a Series B round valuing the company at $1.15 billion, as investors pour money into agentic AI systems designed to operate autonomously on complex tasks. |
| SV005 | OpenAI | Basis scales accounting by turning OpenAI model progress into trusted agents | The team uses OpenAI o3, o3‑Pro, GPT‑4.1, and GPT‑5 to power AI agents that help accounting firms automate repetitive tasks like reconciliations, journal entries, and financial summaries while giving accountants full visibility into how decisions are made and control over the process. |
| SV006 | Built In NYC | Basis Raises $100M to Deploy Automated Accounting Agents | Built In NYC | Tax auditing software provider Basis recently announced a $100 million Series B funding round at a $1.15 billion valuation. The company will put its new capital from lead investor Accel and others toward further developing its agentic AI platform. It will also expand its team across engineering and machine learning. |
| SV007 | ZenML | Basis: Agentic AI for Automated Accounting Workflows at Scale - ZenML LLMOps Database | Multiple well-funded companies in the five to seven years before Basis had failed to make significant progress in automating accounting workflows, creating skepticism that persisted when Basis entered the market. |
| SV008 | Inside Public Accounting | Sikich Partners with Basis to Advance AI-enabled Services | Chicago-based IPA 100 firm Sikich (FY24 net revenue of $438.9 million) announced a partnership with Basis, a platform that helps modernize accounting services through AI-enabled automation. |
| SV009 | PYMNTS | Basis Raises $100 Million to Build Up AI in Accounting | PYMNTS.com | The company’s Series B round, announced Tuesday (Feb. 24), was led by Accel and Google Ventures and values Basis at $1.15 billion. |