初创公司尽调
尽调报告 AI research / collaboration software Seed 2026-10-16

humans&

种子轮前沿 AI 实验室,创始人履历顶尖但商业披露稀少

humans& 的创始团队与前沿 AI 平台高度匹配,资金储备也很充足;但产品、客户和财务证据仍落后于估值叙事,基于公开资料目前只能保持继续研究。

覆盖范围与披露说明

融资和创始人质量的公开证据仍然扎实,但产品、客户、定价和治理披露依旧有限。

封面要素

最近融资 01
480 USDM [CO002]
估值 02
4480 USDM [CO002]

公司概况

humans& 是一家种子轮前沿 AI 实验室,公开将业务定位为促进人与人协作,而非单纯自动化。现有记录支持其具备顶尖的创始人—市场匹配度,并完成了规模异常大的种子轮融资;但具体已上线产品、客户基础、定价和治理条款仍基本未披露。

官网
humansand.ai
创始人
Andi Peng, Georges Harik, Eric Zelikman, Yuchen He, Noah Goodman
创立地点
Redwood City, California, USA
总部
San Francisco Bay Area, California, USA
产品
以人机协作为核心的 AI 模型和产品,公开强调协作、记忆、长时段规划和以用户为中心的交互。
阶段
Seed
融资情况
种子轮融资,披露金额为 $480 million,估值 $4.48 billion。
[CO001, CO002, CO003, CO004, CO005, CO006, CO009, CO013, CO014]

执行摘要

主要优势

  • 创始团队出身顶尖前沿 AI 实验室,并具备扎实的学术研究背景。
  • 已披露的大额种子轮融资提供了充足 runway,也释放出战略信号。
  • “人与 AI 协作”的核心论点清晰,且在公开来源中前后一致。

主要风险

  • 具体产品形态、包装方式和发布状态大多仍未披露。
  • 公开资料仍没有客户、收入、定价和留存证据。
  • 治理、持股和清算条款均未见公开信息。

未决问题

  • 已交付产品的具体形态、路线图和架构,包括目前是否已经存在可用的协作产品或 API。
  • 具名客户、生产环境部署、留存信号,以及按细分市场划分的买方证据。
  • 收入模式、定价、烧钱速度、runway、毛利率,以及其他核心财务评估输入。
  • 董事会构成、估值依据、股权稀释、清算优先权,以及种子轮的其他治理条款。

目录

Chapter 01

01公司概览

1.1 身份与阶段

humans& 将自己定义为一家“以人为中心的前沿 AI 实验室”,打造帮助人们相互理解、协作和共同工作的系统,而不是简单取代劳动。2026 年 1 月的独立报道与这一定位一致,也显示公司以种子阶段品牌却配以晚期资本强度,异常早地进入公众视野:多家媒体报道其完成 $480 million 种子轮融资,估值 $4.48 billion。它的法律足迹比运营足迹更具体:与 California 相关的备案记录显示,Humans& Ai, Inc. 于 2025-10-27 成立,目前仍为存续状态,地址位于 Redwood City;媒体通常将其归入更广义的 San Francisco / Bay Area 生态。[CO001, CO002, CO004, CO014]

快照 KPI 表
指标数值 / 状态截至日期 / 来源窗口置信度缺口 / 尽调备注
身份humans& 将自己描述为以人为中心的前沿 AI 实验室。2026 年官方主页high独立来源对使命陈述一致,但确切的现有产品名称仍未公开。
法律实体状态Humans& Ai, Inc.;备案有效。2025-10-27 已备案;2026-10-16 查阅medium确认运营实体或母公司结构自备案以来是否发生变化。
总部信号州备案地址为 601 Marshall Street, Redwood City, CA;媒体通常将公司归入 San Francisco / Bay Area。2025-10-27 备案及 2026 年 1 月媒体报道medium需要管理层确认当前运营总部及是否有多地点布局。
阶段Seed2026 年 1 月融资披露high指定来源池没有披露后续轮次。
已募资本已披露 $480 million 种子轮融资January 2026high获取交割文件及任何分批交割结构。
估值已披露 $4.48 billion 估值January 2026high公开来源没有拆分证券类型、稀释或优先股堆叠。
创始团队独立发布报道始终列出五位创始人。January 2026high除备案外,当前高管职务图仍然稀疏。
产品 / 商业模式人机协作 AI 理念已公开;确切的已上线产品形态仍未披露。2026 年官方站点和 2026 年 1 月报道medium需要产品演示、客户工作流和产品包装细节。
员工人数TechCrunch 报道发布时约有 20 名员工;该来源集未披露当前人数。2026-01-20medium索取按职能划分的当前员工人数和招聘计划。
收入 / 定价 / 客户已审阅的当前来源未公开披露。已审阅至 2026-10-16low需要客户名单、定价模式、ARR/收入及留存证据,以完成投资判断。

综合公司陈述、2026 年 1 月媒体报道和备案衍生注册页面;类似 null 的状态表示公开未披露,而不是数值为零。

[CO001, CO002, CO004, CO005, CO006, CO009, CO013]

1.2 创始人、领导层与治理

独立的公司发布报道始终列出五位创始人:Andi Peng、Georges Harik、Eric Zelikman、Yuchen He 和 Noah Goodman。官方定位强调团队横跨 xAI、Anthropic、Google DeepMind、OpenAI、Meta、AI2、Stanford 和 MIT 的经历,而没有给出详细组织架构;这支持较强的创始人—市场匹配度,却让当前汇报关系仍然模糊。备案衍生的公开高管记录高度集中:Eric Zelikman 同时担任 chief executive officer、chief financial officer 和 secretary。Noah Goodman 已发表的 STaR 研究记录增加了可见的学术可信度,但指定来源没有披露董事会名单、分工后的财务控制或其他能够降低关键人风险的治理细节。[CO005, CO007, CO008, CO011, CO014]

领导层与创始人表
人员角色 / 公开职务背景证据创始人—市场匹配度 / 职能覆盖关键人依赖
Andi Peng联合创始人TechCrunch 和 TechFundingNews 将 Peng 描述为前 Anthropic 研究员,曾从事 Claude 模型的强化学习和后训练。覆盖与公司理念相关的后训练、对齐和模型行为专业能力。在披露更完整的运营团队之前为 High。
Georges Harik联合创始人;Google employee #7独立报道将 Harik 与 Google 最早期的广告系统联系起来;公司则表示种子轮由 SV Angel 和联合创始人 Georges Harik 领投。增加早期互联网运营经验及投资者网络可信度。Medium-high,因为公开来源没有显示已分工的商业领导层。
Eric Zelikman联合创始人;备案列名 CEO/CFO/SecretaryTechFundingNews 称 Zelikman 是前 xAI 研究员;备案衍生注册页面显示他同时担任三项高管职务。将研究履历与最清晰的公开运营权力结合起来。Very high,因为多项已披露的高管职能集中在一位创始人身上。
Yuchen He联合创始人TechCrunch 和 TechFundingNews 都将 He 列为创始团队中的前 xAI 研究员。扩展创始团队的前沿模型工程深度。Medium,因为没有披露独立的公开运营职责。
Noah Goodman联合创始人;Stanford 教授TechCrunch 称 Goodman 是 Stanford 教授,Google Research 收录的 2022 年 STaR 论文则列 Noah D. Goodman 为共同作者。为实验室叙事增加已发表的认知与推理研究可信度。Medium,因为治理和产品归属仍未披露。

行覆盖 2026 年 1 月发布报道中反复出现的五位创始人;公开董事席位和更广泛的高管团队仍未披露。

[CO005, CO007, CO008, CO011]

1.3 资本基础与利益相关方地图

公司官网和第三方融资报道相互印证:humans& 完成了一轮规模很大的种子融资,由 SV Angel 和联合创始人 Georges Harik 领投,Nvidia、Jeff Bezos、GV、Emerson Collective 及更广泛的投资者联合体出资。这个投资者组合的意义不止于标题式信号 信号,因为它可能带来后续资本,以及对人才、基础设施和分销有价值的关系。但公开记录仍停留在出资方名单,没有披露经济条款:已审阅来源没有给出持股比例、董事席位、清算优先权或其他控制条款,因此利益相关方格局方向上很强,却还不足以仅凭治理条款完成投资判断。[CO002, CO003, CO011, CO014]

利益相关方或投资者地图
利益相关方角色控制权 / 经济重要性证据 / 备注尽调要求
SV Angel 与 Georges Harik种子轮领投方在已披露的 $480 million 融资中传递领投方信号。公司表示种子轮由 SV Angel 和联合创始人 Georges Harik 领投。索取确切的领投经济条款、持股和任何治理权利。
Nvidia战略投资者表明其与算力生态相关,可能获得硬件网络接入。公司和多家新闻媒体均将其列为投资者;未披露商业协议。询问这笔投资是否附带算力供应、额度或 GTM 关系。
Jeff Bezos个人投资者高信号外部资本,可能提高后续融资议价能力。官方和独立融资报道均提及;资本之外的角色未披露。澄清信息权和可能的后续参与。
GV机构投资者将公司连接到大型 AI 生态和分发网络。出现在官方和媒体投资者名单中,但公开资料没有董事席位。确认治理权、pro-rata 权利和任何战略合作条款。
Emerson Collective机构投资者让投资者联合体超越纯基础设施或 AI 专业资本。官方发布材料和融资回顾中将其列为投资者。确认支票金额、权利,以及该投资者是否拥有董事会观察员权利。
Eric Zelikman登记在案的高管申报文件将 CEO、CFO 和秘书职务合并列出,因此这是公开披露中最集中的内部控制节点。这是当前公开记录中最明确的具名运营负责人。索取最新组织架构、董事会审批矩阵和财务授权安排。

这是公开具名利益相关方地图,不是完整股权结构表;经济所有权、清算优先权和董事会代表权仍属非公开信息。

[CO002, CO003, CO011, CO014]

1.4 里程碑与待解决尽调问题

有日期的公开轨迹虽然短,却很能说明问题:2022 年的一篇研究论文将创始人 Noah Goodman 与已发表的推理研究联系起来,公司备案出现在 2025 年 10 月,到 2026 年 1 月,humans& 已经宣布 AI 领域规模最大的种子轮之一,同时将产品愿景聚焦于协作和富记忆交互。仍然缺失的信息与已经存在的信息同样重要。该来源池中的当前资料仍未披露收入、定价、现有客户或具名董事会,唯一的员工人数数据也来自发布初期。前沿模型公司长期受算力、电力和数据中心约束,这使得披露缺口更加关键:任何试图把创始人履历和融资转化为持久产品业务的研究型实验室,都面临更高的执行门槛。[CO009, CO010, CO012, CO013, CO014]

FO001: 公司里程碑时间线

公开记录从创始人在公司成立前的研究经历,一路延伸到公司申报、启动融资,以及尚未解决的商业化问题。

当来源粒度不同时,时间线混合使用精确日期、来源作者给出的时期和年度级里程碑。

Noah Goodman 合著 STaR 论文 [CO008]

Humans& Ai, Inc. 申报日期 [CO004]

创始团队组建 [CO005]

种子轮融资公开 [CO002]

发布报道披露早期规模 [CO006, CO009]

资金用途叙事明确 [CO003, CO009]

首款产品计划推出 [CO010, CO013]

基础设施稀缺仍是背景 [CO012]

[CO004, CO006, CO008, CO010, CO012]

1.5 展示项

Chapter 02

02市场分析

2.1 市场边界与买方逻辑

当前公开记录显示,humans& 所处的细分市场比“数万亿美元 AI 市场”这一标题式市场口径 更窄。独立摘要一致将公司描述为打造用于协作、沟通、搜索和协调的以人为中心的软件,而不是销售原始算力或通用基础设施层。这一点很重要,因为它最接近的商业边界是知识工作的应用和工作流软件,尽管底层模型与基础设施生态大得多。买方路径似乎也在分阶段形成,而非已经确定:公开描述同时指向企业和消费者可能性,但可观察到的预算归属仍最符合 CIO、CTO 或职能软件负责人,他们可以在更广泛的平台标准出现前,为协作工具提供预算支持。[CM001, CM002, CM003, CM015, CM016]

市场定义表
细分市场 / 类别纳入支出排除支出买方 / 付款方相关性
前沿 AI 协作软件面向团队的 AI 辅助沟通、协作、搜索、记忆和工作流支持模型训练资本开支、原始算力转售和通用基础设施合同CIO / CTO 加各职能软件预算最接近 humans& 所述定位的公开类别
AI 模型层模型使用、编排和模型周边工具实体数据中心建设和非 AI SaaS 席位支出中央 AI 平台和工程预算重要的赋能层,但不等于应用收入
AI 基础设施AI 优化 IaaS、服务器、网络架构、半导体和设备协作软件席位和服务交付预算超大规模云厂商、云供应商和基础设施运营商主要设定供给侧约束,而不是 humans& 的直接需求
企业 AI 试验项目使用 GenAI 和智能体的试点及增量生产力项目尚未进入预算的长期转型价值CIO 和业务职能负责人描述规模化平台预算形成前的可能采用路径

边界依据公开产品描述和外部市场分类;它将广义 AI 资本开支与 humans& 仍需争取的更窄应用预算分开。

[CM001, CM002, CM003, CM008, CM013, CM014]
细分市场 / 买方地图
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
企业知识工作团队CIO / CTO 与业务职能负责人跨职能团队中央 IT 加各部门软件预算协作、搜索和任务协调CIO / CTO不替代员工也能带来明确生产力提升
内部 AI 平台团队AI 平台负责人或工程副总裁模型构建者和内部工具链负责人研发或平台预算记忆、编排和多智能体工作流工具CTO / AI 负责人为重复的前沿模型工作流提供可复用基础设施
消费者 / 专业消费者应用产品负责人或个人终端用户个人用户订阅或捆绑产品支出个人协作或生产力辅助产品 GM / 增长负责人日常使用价值足够强,能支撑直接付费入口

买方地图受公开描述限制,尚未得到具名客户、采购记录或定价披露的佐证。

[CM002, CM003, CM007, CM014, CM016]

2.2 市场规模视角及其实际衡量内容

市场规模只能通过宽泛的外部口径估算,而不能通过公司特定的 SAM 或 SOM 得出。Gartner 的 2026 年预测显示,全球 AI 市场约为 $2.6 trillion,但这一总盘子主要由基础设施、服务和模型支出构成,而不是直接用于协作应用的支出。更窄的软件和模型数字与 humans& 更相关,但它们仍是品类视角,并非有来源支持的、属于该公司的可触达收入池。因此,证据支持的是“市场很大,但可变现切片尚未厘清”,而不是一套精确的 TAM/SAM/SOM。下图保留了已发布的数字口径,同时明确这些层级只是分析桶,并非计算出的转化漏斗。[CM004, CM005, CM006, CM009, CM011]

TAM / SAM / SOM 或规模测算视角表
发布方年份地理范围数值CAGR / 增长方法置信度局限
Gartner2026全球$2.595667T AI 总支出+47% YoY覆盖各 AI 类别的自上而下市场预测中这一宽泛口径主要由基础设施和供应商产能驱动,而非协作软件
Gartner2026全球$453.209B AI 软件支出点估计更广泛 AI 市场内的细分预测中口径仍过宽,不能等同于 humans& 的可服务市场
Gartner2026全球$32.604B AI 模型支出2026 年增长 +110%子细分市场预测中模型层支出不等于应用收入
IEEE Spectrum 援引 Stanford AI Index / Quid2025全球 / 美国全球私人 AI 投资 $581B;美国投资 $344B相较 2024 年为创纪录年份汇总投资活动中已部署资本不等于客户需求或经常性软件预算
Colliers2025全球 / 北美$580B+ 数据中心投资;北美吸纳量 15.6 GW投资同比 +27%;吸纳量为 2024 年的 2x基础设施市场快照中衡量的是产能建设经济性,而非应用 TAM

由于公开证据没有披露有来源支撑的 humans& SAM 或 SOM,表中保留自上而下的测算视角;这些数值不是可互换的收入池。

[CM004, CM005, CM006, CM009, CM010, CM011]
FM001: 市场规模视角

从宽到窄的已发布市场口径显示,最大的 AI 支出池仍高度偏向基础设施;最接近 humans& 公开定位的细分市场规模更小,也还不是公司特定的 SAM。

各层宽度仅表示层级,不代表数值比例。

所有数值均为发布方以十亿美元计的点估计;层级顺序用于分析,并非计算得出的漏斗,相邻层级也不代表转化率。

[CM004, CM005, CM006, CM013]

2.3 采用驱动、约束与剩余尽调工作

短期采用逻辑可信,但有条件。Deloitte 预测,2025 年四分之一使用 GenAI 的企业将部署 AI agents,说明将 AI 工作流投入实际运营的意愿正在上升;IEEE 的投资和算力数据也显示,资本仍在涌入这一领域。不过,Gartner 表示企业仍偏好战术性效率提升,CIO 也难以证明业务结果,因此预算释放仍取决于可量化的工作流价值、信任和集成,而不只是前沿模型的新颖性。供给侧条件同样重要:Gartner 的结构数据和 Colliers 的数据中心数据都显示,电力、建设成本和许可已经成为前沿 AI 产能的现实约束。对 humans& 而言,最大的投资判断缺口不是 AI 是否是大市场,而是公司能否在基础设施密集型竞争重置预期前,在这个市场里证明明确的买方、产品形态和定价路径。[CM007, CM008, CM010, CM011, CM012, CM013, CM014]

Chapter 03

03竞争者

3.1 竞争格局不止一个创业公司群体

指定来源池不支持把 humans& 简化为只与一两家种子轮实验室竞争。Tracxn 描述了 1,339 家活跃竞争者,并明确将 Thinking Machines Lab、Google DeepMind 和 Recursive 列为主要同业;AI Market Watch 则将 OpenAI、Anthropic 和 xAI 列为主要竞争者。这一组合很重要,因为它同时把 humans& 放进至少四种买方比较:直接的前沿研究创业公司、成熟的基础模型实验室、特定工作流 AI 应用,以及企业内部自建加现有软件内嵌 AI 的现状。结果是,humans& 不仅要守住差异化论点,还要面对产品更成熟、安装基础更广,或两者兼具的品类。[CP001, CP002, CP003, CP004, CP005, CP013]

竞争对手概况表
替代方案类别公开规模 / 融资信号目标细分市场 / 买方差异化 / 战略方向指定证据中的关键局限
humans&直接前沿实验室初创公司$480M 种子轮,估值 $4.48B知识工作者;企业研究团队;创意专业人士协作式推理、EQ 和以人为中心的工作流指定证据中没有公开定价、具名客户或实时产品访问
Thinking Machines Lab直接前沿实验室初创公司融资约 $2B;上次估值 $12B;据报道正在讨论约 $50B 的新一轮融资模型构建者和微调用户可定制多模态 AI 系统;Tinker 微调产品已发布指定证据未披露定价和客户采用情况
Google DeepMind在位前沿模型实验室Tracxn 竞争对手排名第 #2;摘录未披露融资情况大型生态内的 AI 研究服务买方拥有现成生态邻接能力的在位研究组织指定证据池没有提供定价或买方层面的证明
Recursive新兴直接同业总融资 $650M;Series A 轮寻求递归自我改进系统的 AI 研究用户递归自我改进 AI 研究系统指定证据没有定价或客户证明
Sierra / 工作流智能体供应商邻近工作流替代方案Sierra 采用按结果定价;Decagon 在 2025 年签署了 100+ 个企业交易客户服务和支持团队围绕可量化解决结果销售实时工作流自动化用例比 humans& 的通用协作论点更窄

指定证据池未披露的公开定价和客户证明仍标为未知;Tracxn 排名和评分是目录输出,不是产品表现基准。

[CP001, CP003, CP004, CP005, CP011, CP012, CP013]
FP001: 竞争定位图

已分配证据显示,humans& 的叙事差异化接近前沿实验室同业;商业化和分销则落后于既有厂商及已交付的工作流产品。

X: 商业化成熟度 (1-5 序数)
low: 概念 / 未披露
high: 已交付并部署
description: Ordinal score from the assigned evidence: 1 means pre-product or undisclosed commercialization, while 5 means public product availability with deployment evidence.
Y: 分销 / 采购杠杆 (1-5 序数)
low: 有限
high: 嵌入式渠道或活跃企业采用
description: Ordinal score from the assigned evidence: 1 means no disclosed installed base, while 5 means large embedded distribution or active enterprise deal flow.
条目商业化成熟度 (1-5 序数) (1–5)分销 / 采购杠杆 (1-5 序数) (1–5)说明
1. humans& 1 2 协作推理主张和大额种子轮融资已公开,但已分配证据仍缺少公开定价、具名客户或实时产品访问。
2. Thinking Machines Lab 3 2 已融资约 $2B,最近一次估值为 $12B,并于 10 月推出 Tinker;在已分配来源池中,更广泛的分销证明仍然有限。
3. Google DeepMind 4 5 Tracxn 将 Google DeepMind 列为主要竞争对手,humans& 正在与绑定现有软件和模型生态的替代方案竞争。
4. OpenAI / Anthropic 4 4 AI Market Watch 将这些实验室列为主要竞争者;协作主张是相对于基础模型既有厂商定义的,而不是在其生态之外。
5. Sierra / 工作流智能体 4 3 工作流厂商可以把结果导向的自动化卖给具体职能,而不必让买方为前沿实验室的上行选择权买单。
6. 内部自建 / 现状 3 4 企业已经在现有软件中扩展嵌入式 GenAI 和智能体工作流,但许多企业仍难以证明 AI 带来的实际价值。

坐标是基于已分配来源池编制、并由证据支持的序数评分,不是实测市场份额或基准测试结果。

humans& [CP001, CP004, CP012]

Thinking Machines Lab [CP005]

Google DeepMind [CP003, CP007]

OpenAI / Anthropic [CP002, CP006]

Sierra / 工作流智能体 [CP010, CP011]

内部自建 / 现状 [CP008, CP009]

[CP001, CP005, CP007, CP008, CP009, CP011, CP012]

3.2 差异化最清晰地体现在产品理念,而非商业化证明

第三方描述对 humans& 想成为什么样的公司大体一致。AI Market Watch 将公司定位为协作推理、情商以及由 AI 支持人而非取代人的工作流;AgentMarketCap 则认为 humans& 追求的是协作基础设施,而不是正面参与基于基准测试的基础模型性能竞争。这一理念在战略上清晰,但同一来源也指出,Microsoft Copilot、Google Workspace AI 和 Notion AI 等成熟产品已经让协作市场拥挤,而且它们从既有用户分发渠道起步。实际上,humans& 可以主张概念上存在空白,却仍然落后于已经掌控周边工作流入口的竞争者,开启商业化竞赛。[CP001, CP006, CP007, CP013]

功能 / 能力矩阵
购买标准humans&Thinking Machines Lab在位套件 / 模型实验室工作流替代方案内部自建 / 现状
核心论点以人为中心的协作式推理可定制多模态 AI 加微调模型能力和嵌入式套件效用特定任务解决或工作流自动化按内部流程定制工具
公开产品可用性指定证据未披露Tinker 于 10 月发布现有软件应用中的嵌入式模型已上线的客户服务 AI 部署取决于内部团队执行
公开定价可见度未找到公开定价未找到公开定价指定证据池未建立Sierra 披露按结果定价预算受基础设施、模型和人力选择影响
分销杠杆创始人和投资者网络,但未披露存量客户基础指定证据显示初创公司规模的分销能力现有用户基础和在位软件渠道企业工作流入口和活跃交易现有内部系统和采购控制
信任 / 切换约束没有客户参考或实时产品访问,尚未证明尽管产品已发布,在指定证据中仍未证明买方可以先扩展现有套件,而非切换供应商结果下滑时,糟糕的解决闭环会迅速损害信任控制力最高,但实施负担也最高

标为缺失或未披露的单元反映指定来源池的局限,并不表示该能力不存在。

[CP001, CP005, CP007, CP008, CP010, CP011, CP012]

3.3 商业替代方案已经覆盖已上线产品、结果定价和嵌入式套件

来源池中最强的商业化对比,不是 humans& 与另一家隐身实验室的比较,而是 humans& 与采购动作更具体的替代方案之间的比较。Reuters 报道 Thinking Machines Lab 已经推出用于微调语言模型的产品 Tinker;CNBC 则描述了 Sierra 和 Decagon 等已上线的客服智能体供应商,它们以结果挂钩或工作流挂钩的价值主张向企业销售。Gartner 还表示,企业正在扩大对现有软件应用中嵌入的 GenAI 模型的使用,这进一步强化了成熟套件的采购优势。相比之下,指定的 humans& 证据更强调研究方向和资金用途,而不是公开 SKU、产品包装或采用证明,因此今天的买方有更多购买 AI 的方式,却没有足够证据专门评估 humans&。[CP005, CP008, CP011, CP012]

定价 / 打包比较
替代方案公开定价信号合同 / 计费单位信号所含能力信号仍未知的内容含义
humans&指定证据未披露公开定价未披露围绕多智能体 RL、记忆和长时程规划的人机协作平台研发SKU、席位 / API 定价、实施范围、企业条款难以将实现的 ROI 与已交付替代方案比较
Thinking Machines Lab指定证据未披露公开定价未披露面向语言模型微调的 Tinker打包、基准、定价和部署证明商业化动作已经存在,但经济性仍不透明
Sierra / 工作流智能体按结果定价与解决结果挂钩的采购逻辑面向客户服务工作流的对话式 AI实际价格水平、毛利率和扩展条款买方可以购买与 KPI 挂钩的自动化,而非前沿实验室研发期权
内部自建 / 在位技术栈支出集中在模型、基础设施和现有软件预算随基础设施、模型使用和应用套件分配当前软件环境内嵌 GenAI 和智能体工作流按项目计算的总成本和人员构成大型供应商和内部团队近期可以吸收比新实验室更多的预算

本表比较的是公开定价信号,而非实际合同经济性;缺失行是证据缺口,不是隐藏折扣或同等性的证明。

[CP005, CP008, CP011, CP012]

3.4 护城河能否持久仍是未解问题

核心争论在于,humans& 是在技术栈中构建持久层,还是只是在当前前沿实验室融资周期上包装了一个吸引人的叙事。BCG 表示,只有 26% 的公司具备超越概念验证所需的能力,74% 尚未展示切实的 AI 价值;CNBC 则记录了许多 AI 界面仍把用户困在无法解决问题的循环中。这些品类信号意味着,买方会奖励可量化的解决结果、信任和工作流适配,而不只是前沿实验室履历。在 humans& 展示公开客户采用、定价或转换成本证据之前,来源池支持的判断是:其护城河有可能成立,但尚未得到验证;如果公司无法创造明显更好的人机协作体验,成熟分发渠道和定义更清楚的替代品仍可能把协作功能商品化。[CP008, CP009, CP010, CP014]

3.5 展示项

Chapter 04

04财务

4.1 资本基础与变现可见度

公开财务事实的核心是:2026 年 1 月完成 $480 million 种子轮融资,估值 $4.48 billion,投资者包括 Nvidia、Jeff Bezos、SV Angel、GV 和 Emerson Collective。这一资本基础很有分量,但现有证据仍将公司描述为尚未完成商业验证:公开报道称 humans& 仍没有产品,也没有明确说明究竟会交付什么,尽管它将目标体验描述为多人沟通或协作场景的替代方案。公司披露的资金用途指向围绕多 agent 强化学习、记忆、长时段规划和以用户为中心的协作软件开展模型与产品开发;这有助于理解成本驱动,却不足以确定收入结构、实际定价、毛利率或收入确认机制。实际而言,现有证据只显示资本流入和技术野心,尚未显示已披露的变现引擎。[CI001, CI002, CI003, CI004, CI005, CI009, CI013]

收入来源表
收入来源机制计量单位当前数值 / 状态质量尽调要求
企业协作软件用于职场沟通和决策工作流的协作产品按席位、按工作区或年度企业合同公开描述了使用场景,但未披露产品发布、客户合同或价格低置信度;当前证据是发布前定位,而非变现证明提供试点合同、打包方案和初始转化假设
消费者或家庭协作面向家庭或群组的潜在消费者协作层月度订阅、免费增值或家庭套餐提到消费者用例,但没有公开方案、价格或发布数据低置信度;用例有所暗示,但尚未公开变现提供发布路线图、价格架构和留存假设
协作平台替代路径产品可能替代 Slack、Google Docs 或 Notion 类工作流等多人协作场景工作区或按使用量计费的合同定位已公开,但收入机制和确认政策未披露低置信度;没有签约部署或收入确认证据提供目标客户、合同期限和实施服务假设
战略算力经济性Nvidia 既是投资者,也是战略基础设施信号,但经济性未公开产能预留、额度或基础设施支出承诺对毛利率和现金跑道重要,但未披露合同条款重要性置信度中等,实际经济性置信度低提供 GPU 预留条款、最低承诺和转嫁成本

公开来源描述的是可能的变现入口和成本依赖,而非已披露的当前收入结构;空白表示留存证据没有发布相关信息。

[CI002, CI003, CI004, CI005, CI009, CI010]
定价 / 变现表
方案 / 基准价格 / 单位 / 合同标价与实际价格折扣 / 未知项来源
humans& 协作产品未公开披露留存来源没有标价或实际价格证据合同期限、席位模式和折扣均未知CI004 / CI009
humans& 企业部署未公开披露未披露具名试点、客户群组或付费席位销售周期、实施范围和渠道经济性未知CI005 / CI009
humans& 消费者或家庭用例未公开披露消费者提示没有公开的免费增值、订阅或广告模式ARPU、流失率和获客成本假设未知CI005 / CI009
OpenAI ChatGPT Go 基准每月 $8 USD;计划在美国免费版和 Go 套餐中投放广告已披露公开标价;实际变现结构对外部观察者仍不透明仅作基准,不应将竞争对手经济性推定到 humans&CI015

前三行是 humans& 的披露缺口;OpenAI 行是清楚标注的前沿模型变现设计基准,不是 humans& 定价证据。

[CI004, CI005, CI009, CI015]
FI001: 收入模式桥接图

公开证据显示,大额种子资本正流向模型和产品开发,但变现和毛利阶段仍未披露。

节点 / 连接节点 / 起点终点说明
节点 1 种子资本 [capital]$480M seed at $4.48B valuation funding 产品或定价公开披露前筹集的资本
节点 2 模型和产品研发 [r-and-d]多智能体强化学习、记忆、长时程规划和以用户为中心的 AI 开发
节点 3 协作产品界面 [product]面向沟通和协作场景的消息与协调产品
节点 4 客户工作流 [workflows]企业决策、研究协作、项目管理和家庭协调是来源提到的使用场景
节点 5 变现可见度 [monetization]定价、合同形式和收入确认均未公开披露
节点 6 毛利可见度 [margin]算力成本、支持成本和利润率结构均未公开披露
连接 1 种子资本 [capital]模型和产品研发 [r-and-d]为产品和模型建设提供资金
连接 2 模型和产品研发 [r-and-d]协作产品界面 [product]技术工作旨在转化为协作产品
连接 3 协作产品界面 [product]客户工作流 [workflows]产品面向多用户协调场景
连接 4 客户工作流 [workflows]变现可见度 [monetization]公开来源未描述付费转化
连接 5 变现可见度 [monetization]毛利可见度 [margin]没有定价和算力经济性,无法评估毛利
[CI013]

4.2 需求代理指标与经营经济性

外部需求背景有利:Gartner 预测 2026 年全球 AI 支出为 $2.59 trillion,同比增长 47%;IDC 表示,仅软件和信息服务行业就将从 2024 年的 $33 billion AI 支出增至 2028 年接近 $222 billion。这支持存在庞大可触达预算池,但不能单独验证 humans& 的 GTM 动作。Deloitte 2026 年关于人机交互的研究是与公司理念更相关的商业代理指标:多数组织仍是在旧流程上叠加 AI,而只有工作设计与模型同步变化时,ROI 才会改善。因此,协作优先的 AI 产品既是实施和工作流问题,也是模型质量问题。公开层面,humans& 仍未披露可用的销售周期、CAC、回本期、转化率或客户数;公开备案中的财务职能也显得单薄,2026 年 3 月的备案摘录将 Eric Zelikman 列为 CEO、CFO 和 secretary。[CI008, CI009, CI011, CI012, CI013, CI014, CI015]

单位经济性表
指标数值 / 状态置信度重要性尽调要求
已筹集种子资本$480 million 种子轮,估值 $4.48 billion高说明初始资本规模异常,并设定现金跑道和进展预期确认现金总收入、交割时间和任何分批承诺
融资时的团队规模代理指标约 20 名员工中相对于所筹资本规模,显示运营组织仍处早期按工程、研究、产品、销售和财务提供当前员工数
财务领导层深度公开申报摘录将 Eric Zelikman 列为 CEO、CFO 和秘书中表明财务监督集中,可见财务基础设施有限提供当前财务组织架构、controller/CFO 招聘情况和报告节奏
收入 / ARR / 客户数留存来源未公开披露低核心投资评估指标缺失提供收入运行率、客户数和队列拆分
CAC / 回本周期 / 销售周期保留来源未公开披露低缺少获客和转化数据,无法评估销售效率提供销售管道、赢单率、销售周期长度、CAC 和回本假设
毛利率 / 算力 COGS / 资金续航期保留来源未公开披露低前沿模型的成本结构决定种子轮融资是否充足提供训练成本、推理成本、毛利率、月度消耗和资金续航期

公开披露的数值与明确的披露缺口并存;“未公开披露”意味着保留的来源集合没有提供可支持的数字。

[CI001, CI007, CI008, CI009, CI010]

4.3 投资判断结论与尽调阻塞点

财务结论是:humans& 目前更像一项资本充足的前沿 AI 研究与产品押注,而不是一家拥有公开单位经济性的企业。报道显示,公司在融资时约有 20 名员工,同时开发新模型和产品;一篇来源明确警告,训练和扩展模型需要大量现金,并要与成熟厂商争夺算力。另一个审慎观察是,以协作为中心的增强方案仍可能遭遇企业更偏好降本而非提升协调质量的激励机制。这些问题没有否定机会,但会显著限制投资判断信心,因为公开记录仍缺少收入、客户、定价、毛利、烧钱、资金跑道、债务和算力义务披露。因此,在管理层展示上线经济性、客户采用和资本消耗纪律之前,本章只能支持“继续研究”。[CI006, CI007, CI009, CI010, CI014, CI016]

Chapter 05

05产品与技术

5.1 产品形态仍由工作流理念定义,而非已上线 SKU

保留的公开记录围绕 humans& 想改善的工作流描述公司,而不是围绕一款普遍可用的产品。独立简介和融资报道都将公司描述为打造帮助人们相互协调、而不是直接取代人的 AI;首款产品也反复被描述为消息或协作类应用。这一定位比泛泛的前沿模型叙事更具体:公司不只是说要打造强模型,而是希望模型成为群体工作中的连接组织。 但成熟度仍是问题。同一批保留来源还表示,截至 2026 年春季和年中,首款产品尚未公开上线;公司宣布种子轮时成立仅数月;已披露的产品形态仍主要停留在概念阶段。因此在尽调中,下面的模块图应视为早期运营假设,而非已经由生产环境证明的 SKU 层级。公开描述支持其发展方向,但尚无客户可见的采用、可靠性、实际定价或实施深度证据。[CE001, CE002, CE014]

产品模块 / 资产矩阵
模块 / 资产主要用户公开状态 / 成熟度差异化尽调缺口
协作应用界面知识工作者和团队上线前;公开描述为消息或协作风格人机协作的论点,而非自主替代没有公开 GA 日期、定价表或客户部署证明
内部前沿模型训练研究和模型工程团队重点投入领域;没有公开的模型发布由大额算力预算和研究导向的创始团队支持没有模型卡、基准测试包或能力披露
记忆 / 用户理解层终端用户和协作团队概念已公开描述,但没有技术文档持久化上下文和信息请求行为没有披露隐私、保留或控制细节
开发者 / 分发界面集成方、买方和技术评估者未公开披露可能将触达范围扩大到单一应用界面之外没有宣布 API、开放权重或渠道架构

各行汇总保留的公开描述,并清楚区分已披露的产品界面、推断出的层和缺失证据。

[CE001, CE002, CE003, CE004, CE013, CE014]
工作流 / 用例表
用户任务当前工作流问题humans& 解决方案定位可衡量的公开收益当前局限
协调长期群体工作团队在聊天和提示词中反复重述上下文AI 充当人与人之间的连接组织,而不是单打独斗的智能体围绕协作的战略叙事清晰没有发布部署结果或使用指标
保留共享上下文和意图一次性聊天机器人像是在不断参加第一次会议记忆和用户理解被列为明确的研究押注决策之间可能拥有更强的连续性没有公开的数据处理或隐私控制细节
处理更长、更复杂的多步骤工作孤立的任务自动化会打断复杂团队工作流长期规划和多智能体 RL 是明确的重点领域清楚解释了产品为何不只是聊天机器人没有发布基准测试或产品可靠性证明

收益依据公开陈述的战略意图,而非客户验证过的结果。

[CE001, CE003, CE004]

5.2 公开架构可以推断到层级,但实现细节尚未披露

从保留来源中浮现出的技术图景是分层结构:资本密集型的内部模型训练、面向长时段工作的推理与协调层、记忆和用户理解层,最后是面向协作的应用层。公开报道将多智能体 强化学习、记忆、长时段规划和以用户为中心的产品开发列为明确投资方向。Eric Zelikman 过往的 Quiet-STaR 工作为推理导向的模型训练增添了真实研究履历,但不能替代 humans& 自身公开的系统卡、基准测试发布、API 参考或架构文档。 因此,今天的差异化更多体现在叙事框架,而非已披露的实现。公司显然在押注不同于默认自主智能体的路线,但同行也在转向协作语言。例如 Thinking Machines Lab 表示正在打造与人协作的多模态系统,并明确承诺发布技术博客、论文和代码。与这一基准相比,humans& 的公开理念比公开产物轨迹更强。这不意味着战略无效,却使技术尽调仍必须聚焦于尚未披露的内容。[CE003, CE004, CE005, CE011, CE012, CE013]

技术 / 运营架构表
层 / 流程作用依赖风险
算力和训练基础设施为内部前沿模型实验提供资金GPU 和超大规模云厂商容量,以及持续资本产品公开验证前就需要高额支出
推理 / 模型层支持协作规划和模型质量架构、基准测试和发布路径未披露技术深度更多从团队背景推断,而非来自系统证据
记忆和用户理解层保留上下文,并帮助系统询问缺失信息数据保留、隐私和用户控制政策信任与合规尽调仍未完成
应用和部署层将研究转化为面向团队的协作界面上线执行、集成、支持和企业级包装如果企业偏好更简单的自动化工具,采用风险会上升

该架构是基于保留描述拼出的公开来源推断,并非公司发布的系统图。

[CE003, CE005, CE007, CE008, CE009, CE010, CE013]
FE001: 可从公开信息推断的产品架构图

已保留的公开证据支持分层协作技术栈,但不足以还原详细的实现蓝图。

企业团队工作流
面向用户的任务是群组协调,而不是完成单用户提示词。
条目
  • 消息与协作
  • 共享上下文与决策
应用界面
公开描述指向协作应用,而非面向开发者的平台。
条目
  • AI 辅助协作应用
  • 群组协调体验
记忆与编排
这是已保留公开描述中最明确的命名中间层。
条目
  • 持久记忆
  • 用户理解
  • 长时程规划
模型训练层
公开描述通过研究重点介绍模型栈,而不是提供架构文档。
条目
  • 多智能体强化学习
  • 内部前沿模型训练
外部依赖
产品交付既取决于基础设施可用性,也取决于企业接受度。
条目
  • 算力容量
  • 企业信任证明
  • 发布与集成执行
验证层
仍需通过客户、基准和可靠性证据补齐公开证明。
条目
  • 客户部署证明
  • 基准证据
  • 安全与合规披露

这张分层图根据已保留的公开描述和研究引用推断而来,因为已分配来源池不包含公司发布的架构图。

[CE002, CE003, CE007, CE010, CE013]

5.3 主要产品风险是算力密集型经济性下的执行,以及有限的信任证明

产品和技术风险不在于缺少野心,而在于证明必须按什么顺序到来。保留报道称,大部分种子资本将用于算力和内部模型训练;Gartner 的 2026 年支出展望则称,AI 需求仍由基础设施供应商主导,企业大多偏好战术性效率提升,而不是颠覆性工作流变革。IEEE Spectrum 对 2026 AI Index 的总结还指出,重要模型发布仍高度集中在产业界,自 2022 年以来全球 AI 算力容量一直高速增长。综合来看,humans& 必须先为前沿式技术栈投入资金,才能知道协作优先的应用能否赢得分发和信任。 这种先后顺序带来三个具体尽调问题。第一,保留的公开证据中仍没有已上线模型、公开 API、开放权重发布或基准测试包。第二,保留来源池没有提供安全文档、可靠性披露或合规资质等公开信任材料。第三,公司仍需证明,以增强为中心的协作能在一个往往奖励更快自动化叙事的市场中创造可量化的企业价值。技术押注可能自洽,但验证路径仍未走通。[CE006, CE007, CE008, CE009, CE010, CE013]

5.4 展示项

Chapter 06

06客户

6.1 公开定位指向企业协作买方,而非已证明的客户基础

保留的公司特定来源始终将 humans& 定位为服务沟通、协作和协调工作的以人为中心 AI 平台,而非消费产品或狭窄的垂直应用。这最直接指向企业知识工作买方、AI 项目负责人,以及需要 human-in-the-loop 工作流支持的研究密集型团队。然而,同一批描述产品方向的 2026 年 1 月融资报道没有公布客户名单、具名设计合作伙伴,或买方 / 用户 / 付费方拆分。2026 年 3 月的注册摘要确认公司成立时间很近,却没有增加任何商业细节,因此公开记录仍停留在定位阶段,尚未有客户证明。[CU001, CU002, CU003, CU004]

客户细分表
细分市场买方 / 用户 / 付款方用例规模收入 / 战略价值缺口
企业知识工作团队买方:CIO / COO / AI 负责人;用户:知识工作者和管理者;付款方:中央 IT 或 AI 预算以人为中心的协作,以及与 AI 系统协调工作品类需求可见;humans& 的账户数未披露如果采用率通过治理和 ROI 审查,可能成为共享工作流层指定来源池中没有公开披露 humans& 的具名账户、定价或合同案例
前沿模型和产品团队买方:研究或工程负责人;用户:模型构建者、评估者和运营人员;付款方:研发预算多智能体强化学习、记忆、规划和以用户为中心的 AI 工具产品意图已有描述;没有部署证据在形成重大收入前,设计合作伙伴的战略价值可能更重要没有公开具名的设计合作伙伴、试点或参考团队
正在试验智能体的大型企业买方:业务线和 IT 发起人;用户:跨职能运营团队;付款方:数字化转型预算现有软件环境内从试点走向生产的智能体工作流Microsoft 展示了大规模品类试验;缺少 humans& 的证明如果供应商能展示结果和控制能力,采用后扩张路径是合理的humans& 的采购路径、安全审查和结果证据未公开
渠道或生态主导的触达买方:合作伙伴或生态负责人;用户:联合客户;付款方:共享或嵌入式软件预算潜在的合作伙伴分销或工作流嵌入指定来源池没有证据支持如果真实存在,可能加速触达未披露合作伙伴渠道组合、市场存在感或经销商证据

细分市场依据公司定位和企业 AI 采用研究推断;指定来源池没有披露官方客户名单。

[CU001, CU005, CU007, CU012]

6.2 品类需求可见,但公司特定的采用证明不存在

更广泛的 2026 年市场背景偏正面。Gartner 预测 AI 支出将大幅增长,并明确表示 2026 年企业需求将出现拐点,而不只是 hyperscaler 建设加速。Microsoft 的 Work Trend Index 同样报告了 AI 给员工带来的显著价值,以及企业软件环境中活跃 agent 的快速增长。IEEE 对 2026 AI Index 的总结也显示,私人资本和模型开发仍然充足。即便如此,这些品类信号也不能验证 humans& 是否已把兴趣转化为生产部署。Gartner 提醒企业仍偏好战术性 ROI;IEEE 则指出基准测试提升未必能顺畅映射到真实工作流,因此 humans& 仍需强市场叙事之外的客户证据。[CU005, CU006, CU007, CU008, CU009, CU010]

客户增长 / 采用轨迹表
指标数值日期来源置信度含义缺失分母
全球 AI 支出预测$2.59T(2026 年);同比增长 +47%2026-05-19Gartner高企业 AI 预算正在扩大,支持漏斗顶部需求支出没有细分到 humans& 或特定协作工作流
AI 用户报告的价值66% 花更多时间做高价值工作;58% 产出一年前还做不出的工作2026-05-05Microsoft Work Trend Index中增强人类能力的 AI 工作流正在体现用户价值调查覆盖广泛的 AI 用户,而非 humans& 用户
活跃智能体增长整体同比 15 倍;大型企业同比 18 倍2026-05-05Microsoft Work Trend Index中企业正在从简单聊天转向可重复的智能体工作流未披露 humans& 的部署数、活跃企业数或使用份额
humans& 客户证明披露保留的公司专项报道中没有具名客户、客户数、部署数或留存指标2026-01Crunchbase / VentureCapital / The SaaS News中在指定证据中,公司证明落后于品类势头公开记录无法区分隐蔽试点和零个上线部署

公司相关行记录指定来源池披露与未披露的内容;不能证明保留证据之外不存在私人客户。

[CU003, CU005, CU007, CU008, CU013]
FU001: 客户旅程图

公开证据支持品类需求和产品定位,但没有证明 humans& 已有具名部署或扩张。

阶段顺序反映从品类需求到续约证明的证据推进,并非 humans& 的实际转化率。

[CU005, CU006, CU008, CU010, CU012, CU013]

6.3 持久性、扩张和集中度仍是投资判断缺口

本章的核心投资判断问题不是需求,而是持久性。指定的公司特定来源均未披露客户数、活跃部署、合同期限、续约行为,或按客户 / 渠道划分的集中度。因此 land-and-expand 仍是假设,而非已证明的动作。来源池中的反向证据也很重要:Forbes 描述了 human-data 和前沿模型生态中对项目负责人、销售人才和客户关系经理的激烈争夺,合同金额可能非常高。对 humans& 即使早期兴趣存在,扩大企业需求也可能在持久的经常性采用出现之前,先受到面向客户的执行能力限制。[CU011, CU012, CU013]

具名客户证明表
客户细分市场部署 / 用例生产环境还是试点结果局限
指定来源池中没有公开具名账户未知保留的融资报道没有公开的客户专项部署无证据没有公开结果公开来源讨论公司愿景和融资,而非客户参考案例
未披露公开试点群组未知没有列出设计合作伙伴名单、试点群组或评估项目无证据没有公开结果指定证据无法区分隐蔽试点和没有试点
未披露公开续约或扩张案例未知没有发布多团队推广、续约、追加销售或合同期限案例无证据没有公开结果采用后扩张仍是假设,而非观察到的证明

这是一张依据指定公共来源池建立的样本式负面证据表;它记录缺失的具名证明,不是断言公司没有私人客户。

[CU003, CU012, CU013]

6.4 展示项

Chapter 07

07风险

7.1 按严重程度排序的风险图景

核心投资判断问题不是隐藏丑闻,而是公开叙事仍主要由资本和履历构成。多个来源都确认 humans& 以已经隐含品类创建预期的估值完成了规模异常大的种子轮融资,但公开报道也仍将公司描述为尚未有产品或仅处于研究阶段。这一组合使最高风险很直接:如果 humans& 无法把创始人履历和技术野心转化为外部可见的产品证明,当前估值将几乎没有空间承受普通的前沿模型延迟。因此,热力图和运营风险登记表将产品前执行、算力强度、成熟套件压力和安全治理成熟度排在尚未公开出现的较窄法律事件之前。[CR001, CR002, CR007, CR014, CR015]

运营 / 质量 / 安全风险登记表
失效模式可能性严重性缓解成熟度残余风险证据未解决缺口
产品上线前执行进度落后于估值预期高严重低严重尽管种子轮规模可观,公开报道仍将 humans& 描述为产品上线前或仅做研究需要真实运行的产品、具名部署和结果证据,而不是路线图语言
训练新前沿模型期间算力和资本受挤压高高中高TechCrunch 和 Built In 都将模型路径描述为昂贵且资源密集没有公开披露算力合作伙伴、成本结构或资金续航期假设
多智能体、长期工作流引发安全 / 网络 / 隐私故障中高低高前沿 AI 文献指出,智能体系统存在自主性、集中化、滥用和网络安全风险指定证据中未见公开的评估、隐私或事件响应材料
协作论点退化为普通自动化和岗位压缩高高低高独立评论认为,企业 AI 激励仍偏向效率和人员杠杆需要客户证据证明增强人类能力的结果,而非监控或替代劳动力

残余风险反映当前公开证据支持的内容,而非管理层的私人路线图;缓解成熟度低,意味着指定来源尚未展示运营证明或已披露控制。

[CR002, CR004, CR006, CR008, CR013, CR014]
FR001: 风险热力图

最高的剩余敞口集中在产品推出前的执行、资本密集度、既有厂商套装产品压力,以及仍未披露的控制成熟度。

该序数矩阵源自所引证据;它用于确定优先级,不是量化损失模型。

[CR002, CR004, CR006, CR008, CR010, CR016]

7.2 监管、治理与安全暴露

humans& 将自己定位在用于沟通和协作的多 agent、长时段、富记忆 AI 上,这使公司面临前沿 AI 政策文献已经提出的治理问题:自主性、权力集中、滥用和网络安全。指定证据没有披露信任中心、模型评估计划、隐私架构或事件报告流程,因此公开记录尚未显示公司打算如何把这些控制落地。公开实体证据也很薄:注册衍生摘录显示 CEO 同时列为 CEO、CFO 和 Secretary。这不能证明治理薄弱,却凸显外部可见的制度支撑有多有限。[CR003, CR006, CR009]

监管 / 法律风险登记表
风险管辖区当前公开状态可能性严重性缓解措施残余风险尽调路径
智能体协作工具面临的 AI 安全、隐私和劳动法义务美国 / 欧盟 / 多管辖区指定证据中没有公开的安全、隐私或模型治理项目;因此,该风险是从产品方向推断,而非来自已披露的控制体系中高管理层将产品定位为协作而非替代,这可能收窄部分用例高在评估企业规模化投资前,要求提供模型评估、隐私、数据保留和事件响应材料
关键人物和控制权集中Delaware / California登记处证据显示,实体摘录中 Eric Zelikman 同时担任 CEO、CFO 和秘书中中投资人基础可以施加约束,但指定证据中没有公开董事会设计或授权体系文件中高获取资本表条款、董事会构成、审批阈值和关联方控制措施
人才紧缺的前沿 AI 市场中的收购、IP 和保密压力美国据报道,有意收购方曾接触公司,而大型实验室也在争夺同一人才池中中管理层称其拒绝了相关接触,希望保持独立中审查雇佣、IP 转让、保密和创始人留任条款

该风险登记表仅依据指定公开证据,按严重性部分排序;它不是完整的法律清单,应结合直接政策、合同、法院记录和合规材料扩展。

[CR003, CR006, CR009, CR011]

7.3 运营与依赖风险

运营模式加剧了治理不确定性。TechCrunch 和 Built In 都描述了一条依赖训练新模型、保留用户上下文并支持长时段交互的产品路径;与在现有模型上叠加工作流功能相比,这条路径资本更密集、技术也更脆弱。Gartner 的 2026 年市场观点增加了一个结构性风险:AI 支出仍由供应商和 hyperscaler 主导,而企业仍偏向战术性效率提升,不愿进行颠覆式工作流重构。同时,TechCrunch 指出 Anthropic、Google 和 OpenAI 已经在现有套件中推进协作和多 agent 工作流功能。因此 humans& 同时暴露在技术栈两端:上游算力集中与下游分发捆绑。[CR004, CR005, CR008, CR010, CR013, CR016]

合作伙伴 / 依赖风险登记表
依赖交易对手作用集中度失效情景严重性缓解措施残余风险
训练和推理容量超大规模云厂商 / GPU 和基础设施供应商提供训练新模型和运行富含记忆工作流所需的算力层高humans& 无法以有竞争力的成本或节奏训练或服务目标产品高大额种子轮融资为公司提供了一定采购能力高
协作分发层Anthropic、Google、OpenAI 和其他现有套件控制企业用户已经协作的场所,以及可捆绑多智能体工作流功能的场所高在 humans& 建立切入口前,现有厂商就把协作层做成捆绑功能高如果快速发布,差异化的社会智能模型仍可能切出利基市场高
创始研究人员和高管核心创始团队掌握技术路线图、招聘叙事,以及本轮融资背后的大部分外部信誉高人才流失、治理冲突或收购压力拖慢路线图,并削弱融资杠杆中高强大的投资人联合体可以帮助招聘和留住人才,但公开证据尚未显示公司有更广泛的人才梯队中高

集中度仅为方向性判断,因为指定证据没有披露这些依赖的支出承诺、合同条款或留任方案。

[CR004, CR005, CR010, CR011, CR016]

7.4 监测指标与退出标准

最有用的尽调姿态是要求具体里程碑,而不是争论“以人为中心”的 AI 哲学。据报道,公司计划在一年内推出首款产品,TechCrunch 还报道战略收购方已经表现出兴趣。这些事实形成了清晰检查点。正向路径应展示可用产品、具名生产用户、成文的隐私和模型治理控制,以及协作确实创造客户价值、而不只是又一个压缩劳动力的工具的证据。负向路径则是持续停留在研究阶段、依赖稀缺算力却没有差异化分发,或人才和收购压力打断公司的独立产品路线图。[CR011, CR012, CR015, CR016]

7.5 展示项

Chapter 08

08估值

8.1 建议与入场纪律

当前公开记录指向一个对价格敏感的“继续研究”建议,而不是明确的买入结论。Reuters 和 The SaaS News 相互印证,humans& 于 2026 年 1 月完成 $480 million 种子轮融资,估值约 $4.48 billion;即使在火热的 AI 市场,这也是异常大的种子轮估值。然而,同一批保留证据仍没有收入、ARR、客户数或留存披露;Reuters 将公司描述为预计在 2026 年初推出产品,而不是已经证明规模化商业采用。这使投资逻辑更多依赖团队质量、资本获取和市场期权,而非已验证的变现能力;因此,在没有更多尽调或明显更好入场价格的情况下,公开证据不支持按上一轮披露估值买入。[CV001, CV003, CV004, CV012, CV013]

建议摘要表
建议置信度风险评级估值立场决策含义
继续研究中高偏高在没有客户、收入和下行保护证据,或没有明显更好入场点之前,不要按 2026 年 1 月的估值进行投资判断。

该评估针对最后披露的估值和保留的公开证据,不是泛化的公司质量评分。

[CV001, CV012, CV013, CV014]
FV001: 推荐逻辑

推荐仍是继续研究,因为前沿趋势和充裕资本无法抵消披露入场价格下缺失的证明。

节点 / 连接节点 / 起点终点说明
节点 1 $480M 种子轮,估值 $4.48B [funding]大额种子轮设定了溢价入场点。
节点 2 AI 资本仍然充裕 [market]独立行业来源仍显示支出和融资非常强劲。
节点 3 商业证明仍然有限 [proof]已保留来源没有披露收入、客户数或留存。
节点 4 下行保护不清楚 [governance]没有公开的优先股堆叠、董事会或二级定价细节。
节点 5 估值偏高 [stance]价格跑在公开证明前面。
节点 6 继续研究 [recommendation]只有证明更充分或价格更合理时才上调判断。
连接 1 $480M 种子轮,估值 $4.48B [funding]估值偏高 [stance]设定
连接 2 AI 资本仍然充裕 [market]估值偏高 [stance]支持但不足以消除疑虑
连接 3 商业证明仍然有限 [proof]估值偏高 [stance]削弱
连接 4 下行保护不清楚 [governance]估值偏高 [stance]增加下行不确定性
连接 5 估值偏高 [stance]继续研究 [recommendation]推动
[CV001, CV007, CV008, CV012, CV013, CV014]

8.2 融资背景与可比对象

融资背景解释了 humans& 为什么能获得前沿实验室式的种子轮价格,但本身不能证明价格有吸引力。Gartner 预测 2026 年全球 AI 支出将达到 $2.595667 trillion,同比增长 47%,同时警告支出仍由供应商和 hyperscaler 主导,企业尚未完全释放需求。IEEE Spectrum 另报道称,2025 年 AI 投资超过 $581 billion,超过 2024 年两倍;Feed the AI 则称,2026 年融资正集中到少数前沿和基础设施赢家。这些条件有助于解释为什么一支由前 OpenAI/xAI/Google 人员组成的精英团队能够快速融资,但也意味着市场情绪正在显著参与估值。作为粗略的公开参照,CoreWeave 2026 年 10 月市值 $38.43 billion,约为 humans& 上次披露私人估值的 八点六倍,既凸显上行野心,也凸显上市 AI 公司与尚未规模化的私营实验室之间巨大的证明缺口。[CV005, CV006, CV007, CV008, CV009, CV010, CV013]

正论 / 反论表
立场论点哪些变化会改变判断
正论创始团队和投资人联合体足够强,可以在市场仍奖励前沿 AI 稀缺性的情况下争取更多时间和资本。如果后续证据显示客户采用,或后续融资以运营证明而非声誉定价,应上调判断。
正论$480 million seed 融资让 humans& 获得了不同寻常的资产负债表空间,可以招聘、训练模型,并迭代出有防守力的平台。如果管理层展示算力接入、招聘杠杆和里程碑节奏,证明资本能够转化为已发布产品,这一论点会更强。
正向论点行业资本高度集中;如果协作平台成为品类赢家,即使估值倍数偏高,也仍可能有其合理性。如果公司能展示买方、使用深度或经基准测试的产品表现,投资逻辑会更扎实。
反向论点公开证据仍未披露收入、ARR、客户数或留存,因此当前估值没有建立在可观察的商业证明之上。如果投资人拿到经过核实的经营指标,且这些指标与前沿级估值相符,这项担忧会减轻。
反向论点独立行业来源称,AI 支出确实存在,但主导者仍是模型厂商和超大规模云厂商;这可能让应用层采用落后于叙事型估值。如果企业需求转化为明确客户部署,并在 humans& 身上带来重复扩张,这项风险会下降。
反向论点公开来源没有披露董事会、优先股堆叠、稀释情况或任何二级定价,因此入场时的下行保护并不清楚。如果股权表条款和后续估值显示投资人受到保护,且不存在隐藏的重置条款,判断会改善。

这些论点对估值高度敏感,并假设除已保留的公开记录外,不存在未披露的商业或治理支持。

[CV001, CV003, CV007, CV008, CV009, CV010, CV012, CV013, CV014]

8.3 情景、估值立场与最终尽调要求

乐观情景要求 humans& 把前沿人才和规模很大的种子轮余额转化为已上线产品、可信的企业采用,以及一轮凭运营证明而非单纯叙事达到或超过当前估值的后续融资。基准情景更窄:公司继续吸引人才和资本,但外部投资者仍看不到客户增长、重复使用、单位经济性和下行保护条款的公开证据,因此当前估值显得偏高,却还不能说已经失效。悲观情景是商业化延迟,同时 AI 倍数收缩、大客户放慢实际部署,最终出现 down-round,或公司在很长一段时间里具有战略价值,却无法进行财务投资判断。在不再“继续研究”之前,投资者应索取具名生产客户、cohort 或 pipeline 证据、收入和毛利细节、董事会与优先权层级条款,以及任何能进一步刻画下行风险的后续估值或二级交易。[CV004, CV005, CV008, CV010, CV011, CV012, CV013, CV014]

乐观 / 基准 / 悲观情景表
情景假设估值 / 回报逻辑主要风险概率信号
乐观产品上线、明确的企业采用出现,且下一轮融资或战略估值建立在清晰的商业里程碑之上。如果 humans& 成为稀缺的前沿平台,而不是收入证明有限的研究实验室,$4.48B 的种子轮入场估值仍可能成立。必须在市场情绪降温前完成市场拓展和模型差异化。需要新的外部证据,证明客户、定价和产品使用情况。
基准团队质量和资本仍是优势,但在下一轮尽调周期内,公开商业化证明仍然有限。当前估值仍然偏高;只有证据或价格改善后,才适合继续跟踪或谈判。机会成本、稀释和优先权悬置压力比上行选择权更重要。这是目前与已保留来源集合最匹配的情景。
悲观商业发布延期,企业需求慢于 AI 融资新闻所暗示的速度,后续融资重新给公司定价。由于入场逻辑已经假设未来会取得显著成功,降价轮或平价轮都会损害回报。倍数压缩、客户证明延迟以及隐藏的条款清单保护,都可能伤害新投资人。这一判断得到 KPI 未披露以及行业需求证据参差不齐的支持。

这些情景是定性的,因为已保留的公开来源没有披露收入、利润率、客户群组或优先权条款。

[CV004, CV006, CV008, CV010, CV012, CV013, CV014]

8.4 展示项

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查阅一手文件核实。

证据索引

结论
编号陈述可信度来源
CO001 humans& describes itself as a human-centric frontier AI lab focused on strengthening collaboration, organizations, and communities. 高 SO001, SO003
CO002 Independent January 2026 coverage reported that humans& raised a $480 million seed round at a $4.48 billion valuation. 高 SO002, SO005, SO006
CO003 The company says its seed round was led by SV Angel and co-founder Georges Harik and included Nvidia, Jeff Bezos, GV, Emerson Collective, and a broader syndicate. 高 SO001, SO002, SO005
CO004 A filing-derived registry page says Humans& Ai, Inc. was filed on 2025-10-27, is active, and lists a principal and mailing address at 601 Marshall Street in Redwood City, California. 中 SO004
CO005 Independent launch coverage consistently names Andi Peng, Georges Harik, Eric Zelikman, Yuchen He, and Noah Goodman as humans&'s founders. 高 SO002, SO005, SO006
CO006 TechCrunch reported that humans& had about 20 employees at launch, with team members coming from OpenAI, Meta, Reflection, AI2, and MIT. 中 SO002
CO007 The official launch materials say the founding team has experience across xAI, Anthropic, Google DeepMind, OpenAI, Meta, AI2, Stanford, and MIT. 高 SO001, SO003
CO008 Google Research hosts the 2022 STaR paper listing Noah D. Goodman as a coauthor, supporting a visible pre-company research record in reasoning methods. 中 SO007
CO009 Public descriptions portray humans& as building software and training methods that help people collaborate with each other, including memory and multi-agent reinforcement-learning capabilities. 高 SO001, SO002, SO005
CO010 Tech Funding News reported that humans& planned to launch its first product early in 2026. 中 SO006
CO011 The filing-derived public officer record concentrates the clearest disclosed operating authority in Eric Zelikman, who is listed as chief executive officer, chief financial officer, and secretary. 中 SO004, SO006
CO012 J.P. Morgan says AI supply shortages span chips, power, and data-center infrastructure, creating a tougher execution backdrop for capital-intensive frontier AI labs. 高 SO008, SO001
CO013 The reviewed current sources disclose the financing headline and a launch-era employee count, but they do not disclose current revenue, pricing, or customer metrics. 中 SO001, SO002, SO005
CO014 The reviewed source set does not disclose a board roster, cap-table economics, or other detailed control terms beyond the named investors and filing-listed officer roles. 中 SO001, SO002, SO004, SO005
CM001 Humans& is publicly framed as a human-centric AI company that aims to strengthen collaboration between people rather than replace workers. 中 SM004, SM005, SM009
CM002 Independent descriptions place humans&'s first product closer to communication, collaboration, search, and coordination workflows than to a generic stand-alone chatbot. 中 SM005, SM006, SM009
CM003 TipRanks places humans& in Enterprise AI and says it spans enterprise and consumer applications, so the public market boundary is broader than one vertical workflow. 低 SM007
CM004 Gartner forecasts worldwide AI spending of $2.595667 trillion in 2026, up 47% year over year. 中 SM001
CM005 Gartner forecasts AI software spending of $453.209 billion in 2026. 中 SM001
CM006 Gartner forecasts AI models spending of $32.604 billion in 2026 and says that segment will grow 110% in that year. 中 SM001
CM007 Deloitte forecasts that 25% of enterprises already using GenAI will deploy AI agents in 2025 and that the share will rise to 50% by 2027. 中 SM003
CM008 Gartner says enterprises still favor tactical AI initiatives with incremental efficiency and productivity gains and that CIOs face challenges proving tangible business outcomes. 中 SM001
CM009 IEEE Spectrum reports that private AI investment exceeded $581 billion globally in 2025 and that more than $344 billion of that total was invested in the United States. 中 SM002
CM010 IEEE Spectrum reports that world AI compute capacity has increased more than threefold each year since 2022 and 30-fold since 2021. 中 SM002
CM011 Colliers says 2025 global data-center investment exceeded $580 billion, build costs rose 47% year over year, and 40-50% of total project cost is now attributable to power infrastructure. 中 SM008
CM012 Colliers says more than $64 billion of U.S. data-center projects have been delayed or canceled since 2023 because community and regulatory constraints increasingly affect feasibility. 中 SM008
CM013 AI infrastructure and power availability are immediate commercial constraints for frontier-model companies because Gartner's spending mix and Colliers's development data both show capacity build-out absorbing capital faster than application budgets are proven. 高 SM001, SM008
CM014 Enterprise willingness to experiment with agentic AI is rising, but both Gartner and Deloitte show that measurable ROI and deployment maturity still govern budget release. 高 SM001, SM003
CM015 Humans& had not shipped its first product when its funding was announced in January 2026. 中 SM005
CM016 TipRanks listed humans& with 32 employees as of October 12, 2026, which still implies an early commercialization stage relative to the breadth of the AI market it references. 低 SM007
CP001 Current third-party profiles position humans& as a human-centric, collaborative-reasoning AI company aimed at knowledge-work and enterprise-research workflows rather than straightforward task replacement. 中 SP002, SP007
CP002 AI Market Watch names OpenAI, Anthropic, and xAI as humans&'s main competitors. 中 SP002
CP003 Tracxn says humans& has 1,339 active competitors, including 35 funded and 13 exited, and specifically lists Thinking Machines Lab, Google DeepMind, and Recursive among the top competitors. 中 SP001
CP004 humans& raised a $480M seed round at a $4.48B valuation in January 2026. 中 SP001, SP003, SP007
CP005 Thinking Machines Lab is a better-capitalized direct peer in the assigned evidence, with about $2B raised, a last-reported $12B valuation, and a launched fine-tuning product called Tinker. 高 SP001, SP004
CP006 Third-party commentary frames humans& as collaboration infrastructure between humans and agents rather than another benchmark-led foundation-model lab. 中 SP002, SP003
CP007 The assigned evidence describes the collaboration market as crowded with incumbents such as Microsoft Copilot, Google Workspace AI, and Notion AI, giving humans& immediate distribution-power competition. 中 SP003
CP008 Current enterprise AI spending still favors incumbents and embedded suites: Gartner forecasts $2.59 trillion of 2026 AI spend with more than 45% in infrastructure, while enterprises remain focused on incremental rather than disruptive change. 高 SP005, SP008
CP009 BCG reports that only 26% of companies have the capabilities to move beyond AI proofs of concept and 74% have yet to show tangible value, which raises adoption friction for new collaboration platforms. 中 SP005
CP010 Category-level adverse evidence is real: buyers and users still punish AI products that fail to resolve problems or show value, with CNBC describing non-resolving chatbot loops and BCG documenting weak value realization across most companies. 高 SP005, SP006
CP011 Workflow substitutes can offer more concrete commercial logic than frontier-lab startups: CNBC reports Sierra uses outcomes-based pricing, tying purchase value to successful issue resolution. 中 SP006
CP012 In the assigned evidence, humans& discloses fundraising and R&D priorities but not a public price list, enterprise package, or named-customer proof. 中 SP002, SP003, SP007
CP013 The Tracxn peer set mixes frontier startups, incumbent labs, and research institutes, implying that humans& competes across direct, adjacent, and substitute categories rather than within one narrow application segment. 中 SP001
CP014 The assigned evidence does not yet prove durable switching costs or lock-in for humans&, so moat quality still depends more on founder pedigree and thesis coherence than on public customer or pricing proof. 中 SP002, SP003, SP005, SP006
CI001 humans& raised a $480 million seed round in January 2026 at a $4.48 billion valuation. 高 SI001, SI002, SI003, SI004
CI002 Public coverage names Nvidia, Jeff Bezos, SV Angel, GV, and Emerson Collective among the investors in the seed round. 中 SI003, SI004, SI001
CI003 The disclosed use of funds is to advance an AI platform for human collaboration with investment in multi-agent reinforcement learning, memory, long-horizon planning, and user-centered product development. 中 SI004
CI004 As of the retained January-February 2026 coverage, humans& still had no publicly launched product and no public pricing disclosure. 高 SI001, SI002
CI005 Public descriptions frame the initial product as a collaboration layer for multiplayer communication or workspace contexts, with both enterprise and consumer applications hinted. 中 SI002
CI006 AI Funding Tracker says the planned launch window was early 2026 and the cited use cases included enterprise decision-making, research collaboration, project management, and consumer family coordination. 低 SI001
CI007 AI Funding Tracker described the broader team at funding time as approximately 20 employees. 低 SI001
CI008 A March 2026 filing extract lists Eric Zelikman as humans&'s CEO, CFO, and secretary, indicating a concentrated visible finance and governance function. 中 SI005
CI009 None of the retained public sources disclose humans& revenue, ARR, customer count, list pricing, gross margin, monthly burn, or runway. 中 SI001, SI002, SI003, SI004, SI009
CI010 The public record points to front-loaded capital intensity because humans& is training and scaling a new model while coverage warns that doing so will require large sums of cash and scarce compute access. 中 SI001, SI002, SI004
CI011 Gartner forecasts worldwide AI spending of $2.59 trillion in 2026, up 47% year over year. 中 SI006
CI012 IDC says software and information services should spend about $33 billion on AI in 2024 and nearly $222 billion by 2028, implying a growing budget pool for enterprise AI tooling. 中 SI007
CI013 Deloitte says collaboration-oriented AI adoption depends on redesigning human-machine workflows, with most organizations still taking a tech-first approach and only 14% of leaders adept at shaping those interactions. 中 SI008
CI014 Deloitte reports that organizations are twice as likely to exceed AI ROI expectations when they prioritize work design, and one cited telecom rollout improved from a 5% to a 30% productivity gain after redesigning human-AI workflows. 中 SI008
CI015 OpenAI's public monetization model pairs an $8 per month ChatGPT Go plan with planned advertising tests for free and Go tiers, showing that frontier-model businesses can mix subscription and ad revenue to widen access and support infrastructure costs. 中 SI009
CI016 AI Funding Tracker argues that collaboration-oriented augmentation still faces enterprise incentives that often favor cost reduction and headcount elimination, which is a direct commercial risk to humans&'s thesis. 低 SI001
CE001 Retained public coverage says humans& is building AI to strengthen human collaboration rather than replace workers with autonomous agents. 中 SE001, SE002, SE006
CE002 Retained public descriptions frame the first humans& product as a messaging- or collaboration-style application, and they say it had not launched publicly as of spring and mid-2026. 中 SE001, SE002
CE003 Public sources name multi-agent reinforcement learning, memory, long-horizon planning, user understanding, and user-centered product development as core humans& technical focus areas. 中 SE001, SE003, SE006
CE004 The retained public profiles still do not disclose a humans& model architecture, capability target, or product release timeline in enough detail to evaluate implementation maturity. 中 SE002, SE003
CE005 Eric Zelikman's Quiet-STaR paper presents token-level rationale generation for future-text prediction and reports zero-shot gains from 5.9% to 10.9% on GSM8K and from 36.3% to 47.2% on CommonsenseQA after continued pretraining. 中 SE004
CE006 The retained public record supports humans& more through founder pedigree and strategic framing than through released benchmarks, deployed customer evidence, or publicly inspectable product artifacts. 中 SE002, SE003, SE004
CE007 Coverage of the seed round says most of the capital is being allocated to compute and in-house model training rather than to a visibly launched product surface. 中 SE001, SE002, SE006
CE008 Gartner says 2026 AI spending is dominated by vendors and hyperscalers and that organizations still favor tactical efficiency and productivity initiatives over disruptive enterprise change. 中 SE007
CE009 IEEE Spectrum's summary of the 2026 AI Index says industry released 87 notable models in 2025 and that global AI compute capacity has increased more than threefold annually since 2022. 中 SE008
CE010 Taken together, the company-specific funding disclosures and 2026 AI infrastructure market signals imply that humans& must absorb frontier-model infrastructure cost before public product validation is visible. 中 SE002, SE007, SE008
CE011 Thinking Machines Lab says it is building multimodal systems that work with people collaboratively and that it plans to publish technical blog posts, papers, and code. 中 SE005
CE012 Compared with that peer benchmark, the retained humans& public record offers a clearer collaboration thesis than a comparable public trail of technical artifacts. 中 SE002, SE003, SE005
CE013 One retained profile says humans& had not announced cloud-partner channels, API access, or open-weights release plans. 中 SE002
CE014 The retained funding coverage describes humans& as roughly a 20-person company that had operated in stealth for about three months at the time of the seed announcement. 中 SE001, SE002, SE006
CU001 humans& is publicly positioned in retained January 2026 coverage as a human-centric AI platform for communication, collaboration, and coordinated work rather than as a consumer app or a narrow vertical tool. 中 SU001, SU002, SU005
CU002 Retained January 2026 coverage says the seed capital is being used for compute and product-building areas such as multi-agent reinforcement learning, memory, long-horizon planning, and user-centered development, implying product build-out still precedes public deployment proof. 中 SU001, SU005
CU003 None of the retained company-specific January 2026 reports names a customer, production deployment, pilot cohort, customer count, or retention metric for humans&. 中 SU001, SU002, SU005
CU004 A March 2026 registry summary identifies Humans& Ai, Inc. as an October 27, 2025 filing and explicitly warns that the extracted filing data can be out of date or incomplete, so the registry record does not close the company's customer-proof gap. 低 SU008
CU005 The broader 2026 enterprise backdrop for collaborative and agentic AI is expansionary rather than stagnant. 高 SU003, SU004
CU006 Gartner says enterprises still favor tactical AI initiatives with incremental efficiency gains, so customer procurement is likely to demand measurable near-term outcomes rather than frontier-AI aspiration alone. 中 SU003
CU007 Microsoft's 2026 Work Trend research reports user-level value from AI, supporting category-level buyer appetite for human-augmenting workflows. 中 SU004
CU008 Microsoft's reported 15x year-over-year growth in active agents suggests enterprises are moving beyond simple chat into repeatable agent workflows, which could support a land-and-expand path for vendors that can clear governance and ROI review. 中 SU004
CU009 IEEE Spectrum's summary of the 2026 AI Index shows that frontier AI development and investment remain heavily funded, confirming a strong supply-side environment around the category. 中 SU007
CU010 Category excitement is not equivalent to production customer value, because Gartner says CIOs still struggle to prove outcomes and IEEE quotes Stanford AI Index leadership cautioning that benchmark scores may say little about fit in real workflows. 高 SU003, SU007
CU011 Forbes reports that human-data and frontier-model suppliers compete aggressively for project leads, sales staff, and relationship managers who can influence contracts worth up to hundreds of millions, highlighting commercialization risk around scarce customer-facing talent. 中 SU006
CU012 The assigned source pool does not disclose NRR, GRR, churn, renewal rates, contract duration, top-customer mix, or channel-partner contribution for humans&, so durability and concentration remain unverified. 中 SU001, SU002, SU005, SU008
CU013 On the assigned evidence alone, humans& looks like a category-aligned but still pre-proof commercial story, with strong external demand indicators but no retained company-specific adoption or expansion evidence. 中 SU001, SU003, SU004, SU005
CR001 Humans& raised a $480 million seed round at roughly a $4.48 billion valuation in January 2026. 高 SR001, SR002, SR007
CR002 Public coverage in early 2026 still described humans& as pre-product or research-only rather than shipping a live external product. 高 SR001, SR007
CR003 Humans& is trying to pair a collaboration product with a new model emphasizing multi-agent reinforcement learning, memory, and long-horizon planning. 高 SR001, SR004, SR008
CR004 Training and scaling that model is expected to require large ongoing cash and compute, putting humans& into direct resource competition with larger AI labs. 高 SR001, SR004
CR005 Gartner reported that 2026 AI spending remains vendor- and hyperscaler-led and that enterprises still prefer tactical efficiency gains over disruptive workflow change. 中 SR006
CR006 The International AI Safety Report 2026 says agentic general-purpose AI brings emerging risks around human autonomy, concentration of power, misuse, and cybersecurity. 中 SR005
CR007 Independent commentary characterized humans&'s seed valuation as unusually large relative to the company's opacity and lack of demonstrated market fit. 中 SR002
CR008 Enterprise AI deployments often drift toward cost reduction and role compression, which cuts against a collaboration-first product story unless customers show augmentation outcomes. 高 SR002, SR006
CR009 Registry-derived public filing data lists Eric Zelikman as CEO, CFO, and Secretary on the available California entity extract. 低 SR003
CR010 Major incumbents are already embedding collaboration or multi-agent workflow features into existing platforms, increasing distribution and bundling pressure on humans&. 中 SR001
CR011 Interested acquirers had already approached humans&, according to TechCrunch, even though management said it wanted to remain independent. 中 SR001
CR012 Daily with AI reported that humans& planned to launch its first product within the year, creating a near-term execution checkpoint. 低 SR009
CR013 Built In said the large raise was driven by the cost of training interactive systems that ask users questions and store their responses. 中 SR004
CR014 NeoLabs summarized humans& as a human-centric frontier lab but also said it had not released anything outsiders could use yet. 中 SR007
CR015 If humans& does not ship an externally usable product or show named production use before the next financing milestone, the current valuation story remains unverified. 中 SR001, SR007, SR009
CR016 If AI infrastructure spending stays vendor-led and collaboration features remain bundled inside incumbent suites, humans& faces both compute scarcity and distribution squeeze rather than a clean greenfield market. 中 SR001, SR006
CV001 humans& announced a $480 million seed financing at a $4.48 billion valuation on 2026-01-20. 高 SV003, SV007
CV002 Reuters reported that the round was led by SV Angel and Georges Harik, with Nvidia, Jeff Bezos, and GV among participating investors. 中 SV003
CV003 Recent Funding still lists only one sourced humans& round on record, the January 2026 seed financing. 高 SV001, SV003
CV004 Reuters said humans& expected to launch a product early in 2026, which means the disclosed valuation arrived before retained public sources showed scaled commercial adoption. 中 SV003
CV005 The SaaS News said humans& planned to use the funding for multi-agent reinforcement learning, memory, long-horizon planning, and user-centered AI product development. 中 SV007
CV006 CoreWeave's market capitalization was $38.43 billion on 2026-10-15, about 8.6 times humans&'s last disclosed $4.48 billion private valuation. 中 SV002, SV003
CV007 Gartner forecast worldwide AI spending of $2.595667 trillion in 2026, up 47 percent year over year. 中 SV005
CV008 Gartner said AI spending is still dominated by vendors and hyperscalers and that enterprises have yet to flex their spending potential. 中 SV005
CV009 IEEE Spectrum reported that 2025 AI investment exceeded $581 billion, more than double 2024's $253 billion, and was led by private investment in AI companies. 中 SV006
CV010 Feed the AI described 2026 fundraising as concentrating into frontier labs and infrastructure-scale financings rather than normal venture rounds. 中 SV004
CV011 Bizprofile's filing summary lists Humans& Ai, Inc. as active and identifies Eric Zelikman as CEO, CFO, and secretary in its extracted public filing data. 低 SV008
CV012 None of the retained public valuation-context sources disclose humans& revenue, ARR, customer count, retention, or unit economics, so the current underwriting case rests more on team and market narrative than on commercial proof. 中 SV001, SV003, SV007
CV013 The combination of a premium seed valuation, concentrated AI capital, and limited public commercialization proof supports a stretched valuation stance and a research-more recommendation at the last disclosed mark. 高 SV003, SV004, SV005, SV006
CV014 Retained public sources still do not disclose humans& board composition, preference stack, dilution terms, or any secondary pricing that would clarify downside protection for new investors. 中 SV001, SV003, SV007, SV008
来源
编号出版方标题引文
SO001 humans& AI for humans | humans& Today we introduce humans&, a human-centric frontier AI lab.
SO002 TechCrunch Humans&, a "human-centric" AI startup founded by Anthropic, xAI, Google alums raised $480M seed round Humans&, a startup with a philosophy that AI should empower people rather than replace them, has raised $480 million in seed funding at a $4.48 billion valuation, reports The New York Times.
SO003 StartupHub.ai DeepMind, OpenAI Vets Launch humans& for Human-Centered AI The startup is founded by a supergroup of researchers and builders who previously led major efforts at xAI, Anthropic, Google DeepMind, and OpenAI, and aims to fundamentally pivot the industry’s focus away from autonomous agents toward truly human-centered AI.
SO004 Bizprofile Humans& Ai, Inc Redwood City, CA - filing information Officially filed on October 27, 2025, this corporation is recognized under the document number B20250359156.
SO005 The SaaS News humans& Raises $480M Seed at $4.48B Valuation humans&, a San Francisco, CA–based startup focused on AI that empowers people rather than replaces them, has raised $480 million in Seed funding at a $4.48 billion valuation.
SO006 Tech Funding News Ex-OpenAI researchers’ Humans& raises $480M: Can it beat Thinking Machines Lab? Humans& plans to launch its first product early this year.
SO007 Google Research STaR: Self-Taught Reasoner Bootstrapping Reasoning With Reasoning
SO008 J.P. Morgan Asset Management Is AI running out of compute? Many AI services today are priced well below cost, subsidized by companies incentivizing adoption and competing for market share, but rising compute expenses could force a rethink.
SM001 Gartner Gartner Forecasts Worldwide AI Spending to Grow 47% in 2026 Worldwide spending on AI is forecast to total $2.59 trillion in 2026, a 47% increase year-over-year, according to Gartner, Inc. a business and technology insights company.
SM002 IEEE Spectrum 12 Graphs That Explain the State of AI in 2026 Title: Stanford's AI Index for 2026 Shows the State of AI
SM003 Deloitte Deloitte Global's 2025 predictions report Deloitte predicts that 25% of enterprises using GenAI are expected to deploy AI agents in 2025, growing to 50% by 2027.
SM004 Altis Research humans — Frontier AI Labs | Altis Research The bet is that every frontier lab has optimized its models to work alone, and that the larger prize belongs to whoever builds for groups of people working together.
SM005 TechStartups AI startup Humans& raises $480M seed at $4.48B valuation as former OpenAI and Google researchers launch frontier AI lab Humans& says it is building human-centric AI tools aimed at communication and collaboration.
SM006 Dealroom News Backed by Nvidia and Bezos, AI startup humans& raises $480m seed at $4.48b valuation The company is working on software that it describes as an AI-enabled communication tool, closer to a messaging platform than a traditional chatbot, designed to ask clarifying questions, retain context, and assist groups over time.
SM007 TipRanks humans& Leadership, Clients & Company Overview humans& is a San Francisco–based AI startup focused on developing AI systems that empower people rather than replace them. The company builds human-centric AI tools and platforms to augment worker capabilities across enterprise and consumer applications.
SM008 Colliers 2026 U.S. Data Center Marketplace Community opposition, zoning constraints, and environmental review processes are exerting increasing influence on development timelines and feasibility, with more than $64 billion in U.S. projects delayed or canceled since 2023.
SM009 The SaaS News Humans& Raises $480M Seed At $4.48B Valuation The company will use the funding to advance its AI platform for human collaboration, focusing on innovations in multi-agent reinforcement learning, memory, long-horizon planning, and user-centered AI product development.
SP001 Tracxn humans& - 2026 Company Profile, Team, Funding & Competitors ... The company has 1339 active competitors, including 35 funded and 13 that have exited .
SP002 AI Market Watch Humans& - AI Startup Profile | AI Market Watch Focuses on 'Collaborative Reasoning' and Emotional Intelligence (EQ) rather than just task automation. Utilizes proprietary 'Quiet-STAR' algorithms to give AI an internal monologue, allowing it to reason before speaking, which contrasts with the standard next-token prediction models of competitors.
SP003 AgentMarketCap Humans& Raises $480M at $4.48B: Why Cross-Lab Founders Are ... The collaboration market is crowded. Microsoft (Copilot), Google (Workspace AI), and Notion (AI) have distribution advantages that a three-month-old startup cannot easily replicate. Humans& will need to win on architectural differentiation, not feature parity.
SP004 Reuters Mira Murati's Thinking Machines seeks $50 billion valuation in funding talks, Bloomberg News reports (Reuters) -Thinking Machines Lab, the artificial intelligence startup ​founded by former OpenAI executive Mira ‌Murati, is in early talks to raise a ‌new funding round at a roughly $50 billion valuation, Bloomberg News reported on Thursday.
SP005 Boston Consulting Group AI Adoption in 2024: 74% of Companies Struggle to Achieve and Scale Value New BCG Research on AI Adoption Identifies Fintech, Software, and Banking as the Sectors with the Highest Concentration of AI Leaders
SP006 CNBC 'I hate customer-service chatbots': The consumer-AI refund relationship is off to a rocky start Title: 'I hate customer-service chatbots': The consumer-AI refund relationship is off to a rocky start
SP007 The SaaS News humans& Raises $480M Seed at $4.48B Valuation humans&, a San Francisco, CA–based startup focused on AI that empowers people rather than replaces them, has raised $480 million in Seed funding at a $4.48 billion valuation. ... The company will use the funding to advance its AI platform for human collaboration, focusing on innovations in multi-agent reinforcement learning, memory, long-horizon planning, and user-centered AI product development.
SP008 Gartner Gartner Forecasts Worldwide AI Spending to Grow 47% in 2026 STAMFORD, Conn., May 19, 2026
SI001 AI Funding Tracker Humans& Raises $480M Seed Round (2026) On January 20, 2026, a three-month-old AI startup with no product announced a $480 million seed round at a $4.48 billion valuation.
SI002 Yahoo Tech / TechCrunch Humans& thinks coordination is the next frontier for AI, and they’re building a model to prove it Humans& still doesn't have a product, nor has it been clear about what exactly it might be, though the team said it could be a replacement for multiplayer or multi-user contexts like communication platforms (think Slack) or collaboration platforms (think Google Docs and Notion).
SI003 Forbes Humans& Raises $480 Million, Eleven Labs Launches AI Music, Xreal Sues Viture Three month old AI startup Humans& raises a $480 million seed round at a $4.48 billion valuation, one of the largest seed financings on record.
SI004 The SaaS News humans& Raises $480M Seed at $4.48B Valuation The company will use the funding to advance its AI platform for human collaboration, focusing on innovations in multi-agent reinforcement learning, memory, long-horizon planning, and user-centered AI product development.
SI005 Bizprofile Humans& Ai, Inc Redwood City, CA - filing information The corporation is led by a team of key individuals: Eric Zelikman from Redwood City CA, holding the position of Chief Executive Officer; Eric Zelikman from Redwood City CA, serving as the Chief Financial Officer; Eric Zelikman from Redwood City CA, serving as the Secretary.
SI006 Gartner Gartner Forecasts Worldwide AI Spending to Grow 47% in 2026 Worldwide spending on AI is forecast to total $2.59 trillion in 2026, a 47% increase year-over-year, according to Gartner, Inc.
SI007 IDC IDC’s Worldwide AI and Generative AI Spending – Industry Outlook Combined, these sectors are projected to allocate approximately $89.6 billion towards AI in 2024, representing 38% of the global AI market. With an impressive five-year Compound Annual Growth Rate (CAGR) of 27%, their collective investment is anticipated to surge to nearly $222 billion by 2028.
SI008 Deloitte Getting human and machine relationships right Deloitte research shows that organizations are twice as likely to exceed their return on investment expectations for AI when they prioritize work design, thoughtfully redesigning human and machine interactions and roles.
SI009 OpenAI Our approach to advertising and expanding access to ChatGPT We’ve been working to make powerful AI accessible to everyone through our free product and low-cost subscription tier, ChatGPT Go, which has launched in 171 countries since August.
SE001 yespress humans& - AI That Connects People, Not Replaces Them
SE002 nextomoro Humans&
SE003 Reworked Humans& Bets $480M That AI Can Be Human-Centric
SE004 arXiv Quiet-STaR: Language Models Can Teach Themselves to Think Before Speaking
SE005 Thinking Machines Lab Thinking Machines Lab
SE006 The SaaS News humans& Raises $480M Seed at $4.48B Valuation
SE007 Gartner Gartner Forecasts Worldwide AI Spending to Grow 47% in 2026
SE008 IEEE Spectrum 12 Graphs That Explain the State of AI in 2026
SU001 Crunchbase News Humans& Raises Huge $480M Seed Round At $4.48B Valuation For ‘Human-Centric AI Lab’ So far, not much is known about the company, founded in September 2025, which says its mission is to design an AI tool “around how people connect and work together, where collaboration and human insight remain central.”
SU002 VentureCapital.com Humans& Secures $480M in Seed Funding, Valued at $4.48B SV Angel and Georges Harik lead the seed financing round for Humans&, a company specializing in human-centric AI tools for communication and collaboration.
SU003 Gartner Gartner Forecasts Worldwide AI Spending to Grow 47% in 2026 “Up to this point, AI spending has primarily been driven by technology companies and hyperscalers,” said Lovelock. “Enterprises have yet to really flex their spending potential. That is coming and 2026 will be the inflection year. Currently, organizations show limited appetite for using AI to drive disruptive enterprise change. Instead, they favor tactical AI initiatives with incremental improvements in efficiency and productivity.
SU004 Microsoft WorkLab 2026 Work Trend Index report: Agents, human agency, and the opportunity for every organization The number of active agents in the Microsoft 365 ecosystem has grown 15x year over year, rising to 18x in large enterprises.
SU005 The SaaS News humans& Raises $480M Seed at $4.48B Valuation The company will use the funding to advance its AI platform for human collaboration, focusing on innovations in multi-agent reinforcement learning, memory, long-horizon planning, and user-centered AI product development.
SU006 Forbes The AI Talent Wars Have Hit Data Labeling A single relationship manager train-hopping could take their learned expertise with them and use it to win business contracts that can be worth up to hundreds of millions.
SU007 IEEE Spectrum 12 Graphs That Explain the State of AI in 2026 “We generally lack measures of how well a system (or agent) needs to function in a particular setting,” says Perrault. “Knowing that a benchmark for legal reasoning has 75 percent accuracy tells us little about how well it would fit in a law practice’s activities.”
SU008 Bizprofile Humans& Ai, Inc Redwood City, CA - filing information Everything above comes from state filings, so it can be out of date or incomplete. Claim the page and you can edit any field on it, then add the details filings never include.
SR001 TechCrunch Humans& thinks coordination is the next frontier for AI, and they're building a model to prove it Humans& will need endless large sums of cash to fund the expensive endeavor that is training and scaling a new model. That means it will be competing with the major established players for resources, including access to compute.
SR002 Reworked Humans& Bets $480M That AI Can Be Human-Centric "Humans&'s funding and valuation are highly unusual even by the standards of the current AI boom because of both the scale and the opacity of what it's achieved so early in its lifecycle," she said.
SR003 Bizprofile Humans& Ai, Inc Redwood City, CA - filing information Eric Zelikman from Redwood City CA, holding the position of Chief Executive Officer; Eric Zelikman from Redwood City CA, serving as the Chief Financial Officer; Eric Zelikman from Redwood City CA, serving as the Secretary.
SR004 Built In New AI Startup Humans& Raises $480M Seed at $4.8B Valuation The company will train its systems to be more interactive, enabling it to ask users questions and store their responses. The lofty cost of using existing AI techniques to train automated models in new ways is the driving factor behind the company’s sizeable capital raise.
SR005 International AI Safety Report International AI Safety Report 2026 This edition focuses on rapidly evolving general-purpose Al systems, including language, vision and agentic models. It also reviews associated challenges, including wider impacts on labour markets, human autonomy and concentration of power.
SR006 Gartner Gartner Forecasts Worldwide AI Spending to Grow 47% in 2026 Worldwide spending on AI is forecast to total $2.59 trillion in 2026, a 47% increase year-over-year, according to Gartner, Inc. a business and technology insights company. ... “Up to this point, AI spending has primarily been driven by technology companies and hyperscalers,” said Lovelock. “Enterprises have yet to really flex their spending potential.
SR007 NeoLabs Humans&: funding, founders and investors It has raised $480M in disclosed rounds, most recently a Seed of $480M at a $4.48B valuation in January 2026. It has not released anything outsiders can use yet.
SR008 The SaaS News humans& Raises $480M Seed at $4.48B Valuation The company will use the funding to advance its AI platform for human collaboration, focusing on innovations in multi-agent reinforcement learning, memory, long-horizon planning, and user-centered AI product development.
SR009 Daily with AI Humans& Raises a Staggering $480M Seed Round to Build “Human-Centric” AI Systems The startup plans to launch its first product within the year, further signaling that the AI funding landscape is broadening beyond language chat interfaces and into platforms to amplify human agency.
SV001 Recent Funding humans& raised $480M Seed (2026): investors, founders
SV002 CompaniesMarketCap CoreWeave (CRWV) - Market capitalization
SV003 Reuters AI Startup Humans& Raises $480 Million at $4.5 Billion Valuation in Seed Round
SV004 Feed the AI Biggest AI Funding Rounds of 2026 (So Far)
SV005 Gartner Gartner Forecasts Worldwide AI Spending to Grow 47% in 2026
SV006 IEEE Spectrum Stanford's AI Index for 2026 Shows the State of AI
SV007 The SaaS News humans& Raises $480M Seed at $4.48B Valuation
SV008 Bizprofile Humans& Ai, Inc Redwood City, CA - filing information