Mighty Capital Uses AI Signals From 600,000 Product Leaders to Outperform VCs

Mighty Capital利用60万名产品负责人的AI信号跑赢风投 该机构通过专有AI引擎解析来自高管网络的实时产品洞察，而非依赖传统风投社交，从而实现了20%的超常投资成功率。 SC Moatti阐明了Mighty Capital的“产品Alpha方法论”，该方法论利用AI驱动的引擎，从由60万名产品负责人构成的专有网络中挖掘对话内容。其产品信号智能系统（PSIS）每天将约500个对话片段过滤为高保真信号，使该机构达成了20%的超常投资成功率，显著高于行业平均水平。通过智能AI眼镜等案例研究，Moatti展示了NLP流水线如何在保留人工对最终投资决策与尽调把关的同时，提炼来自CPO的原生洞察以锁定顶尖初创公司。 “产品Alpha效应”方法论提出，产品创新直接驱动卓越的风险投资表现。 (0:04) Mighty Capital的产品信号智能系统（PSIS）依赖于一个由60万名产品经理和CPO组成的专有网络，约占全球产品缔造者的三分之一。 (0:04) 该CPO网络的起源可追溯至SC Moatti十年前创立的一家非营利组织，该组织当时为产品负责人举办私密晚宴。 (0:04) PSIS每天吸纳社区对话，并将其浓缩为大约500个关键对话片段或“时刻”。 (0:04) 相较于行业标准幂次法则下的二三十分之一，Mighty Capital实现了20%的超常成功率（每5笔投资中就有1笔带来10倍以上的收益）。 (0:04)

SC Moatti of Mighty Capital argues that venture capital performance can be systematically improved by mining front-line operational signals from enterprise product leaders rather than relying on traditional social networks. By parsing unstructured discussions from a proprietary community of 600,000 product managers, the firm uses natural language processing to identify enterprise adoption trends before they surface in conventional venture pipelines. Proprietary Product Networks Driving Venture Alpha (0:04) Moatti contends that product practitioners identify technological inflection points long before venture capital investors recognize them. To capture this 'product alpha effect'—the premise that product innovation directly fuels superior investment returns—Mighty Capital developed the Product Signal Intelligence System (PSIS). The infrastructure draws upon a 600,000-member community of product managers and Chief Product Officers (CPOs) built from a non-profit organization Moatti founded a decade ago. The pipeline digests roughly 500 critical conversational snippets daily, which are filtered into human-readable signals to inform a cadence of one full due diligence per week and one to two executed investments per quarter. Moatti attributes early stakes in companies like Netskope and Groq to these signals, asserting that the firm maintains a 20% outlier rate (1 in 5 investments yielding 10x+ returns) compared to the broader venture industry's traditional power-law benchmark of 1 in 20 or 30. Filtering Enterprise Roadmaps Over Public Market Data (6:45) Mighty Capital's data strategy intentionally bypasses noisy public platforms such as PitchBook, LinkedIn, and Reddit, prioritizing high-fidelity enterprise disclosures. The core dataset originates from Fortune 1000 CPOs—including leaders at Walmart, Johnson & Johnson, Disney, and Cisco—discussing internal roadmap allocations, budget shifts, and software tooling gaps. This input is supplemented by an advisory body of 1,000 product executives who evaluate 12,000 to 15,000 annual Product Award nominations, generating comparative evaluation records on emerging technologies. To demonstrate this data architecture, Moatti highlights a market analysis evaluating smart glasses as an interface for generative AI. The system ingested approximately 900 conversational moments from CPO roundtables, which natural language processing pipelines condensed into 60 weighted technical signals concerning micro-chip miniaturization and low-latency sensors. Mapping these high-intensity signals to prospective startups led to an investment in an infrastructure provider partnering with Samsung, Anthropic, and OpenAI. Algorithmic Top-Funnel Selection and Founder Diligence (14:21) Addressing concerns regarding automation in venture investing, Moatti maintains that artificial intelligence is strictly confined to top-of-funnel discovery, leaving relationship building, signal calibration, and final investment decisions entirely under human control. This selective automation is framed as a corrective to the venture industry's historical reliance on insular founder networks, which Moatti critiques as biased and linked to sub-par governance. During active due diligence, the system parses submitted pitch decks, cap tables, and financial files to auto-generate a neutral assessment report within minutes for founder review. Moatti argues this transparent, rapid feedback loop eliminates bias and turnaround delays. Backed by six public IPOs, the firm emphasizes that the integrity of its sourcing model ultimately depends on sustained, non-transactional relationships with CPOs throughout their executive careers. While Mighty Capital reports a 4x to 5x outperformance over standard venture benchmarks, the approach relies on privileged access to closed enterprise discussions and requires active human oversight to weight automated signals accurately. The sustainability of this model remains contingent on preserving community trust and ensuring that high-level executive roundtables do not become compromised by promotional noise.

Mighty Capital管理合伙人SC Moatti提出“产品Alpha效应”（product alpha effect），认为产品创新直接驱动超额投资回报；该机构依托覆盖全球约三分之一产品构建者的专有网络与产品信号智能系统（PSIS），将一线产品主管的对话转化为量化信号，以重塑早期风险投资的标的筛选与尽职调查流程。 产品Alpha假说与专有网络沉淀 (0:04) 在硅谷拥有二十年创业与产品管理经验（曾任职于早期Facebook与晚期Nokia）的SC Moatti认为，产品专家往往最早洞察行业未来，而传统风险投资机构反而是最晚获知趋势的一方。针对这一错配，Mighty Capital提出了“产品Alpha效应”方法论，即产品创新能力直接驱动超额投资回报。为在大规模层面上践行该方法论，机构搭建了产品信号智能系统（PSIS）。该系统的核心壁垒源自一个包含60万名产品经理和首席产品官（CPO）的专有网络，据Moatti估算，该群体约占全球各行业及地区产品构建者总数的三分之一。 这一庞大网络的起点源自Moatti十年前创立的一家旨在协助产品经理打造更优产品的非营利组织。起初，Moatti在出版产品领域著作后，通过举办私密晚宴与产品同行探讨行业动向；当同行提及正在构建或关注的新兴工具与初创企业时，Moatti便据此介入投资，随后将所投产品推介回晚宴圈子进行试点与采用。这种基于同行实际工作痛点的互动，逐渐沉淀为具有持续反哺能力的生态系统。 高保真数据架构与AI智能眼镜案例 (6:45) Moatti指出，驱动PSIS系统优异表现的核心在于数据质量，遵循“优质数据产生优质结果”的原则。与依赖PitchBook、LinkedIn或Reddit等高噪声公开渠道的常规机构不同，Mighty Capital规避了公开数据，转而直接采集来自财富1000强企业（如沃尔玛、强生、迪士尼、思科等）CPO的高保真、清洗后的一手对话。这些数据涉及企业的产品路线图资源分配、预算优先级及具体功能缺口。此外，由1000名特选产品高管组成的顾问委员会每年会对1.2万至1.5万项“产品大奖”（Product Awards）提名进行同行评审，形成有关竞品优势的深度评价。 以评估“智能眼镜是否会成为生成式AI杀手级硬件设备”为例，PSIS展示了完整的信号提炼链路：系统首先从探讨硬件限制与AI设备形态的CPO圆桌讨论及顾问互动中，聚合了约900个特定“时刻”（moments）；随后利用自然语言处理（NLP）技术将其提炼为约60个涵盖芯片小型化需求、低延迟传感捕获等维度的加权信号；最终高强度信号成功映射并锁定了目标初创企业，促成了该机构对一家与三星、Anthropic及OpenAI在智能眼镜赋能方面展开合作的初创公司的投资。 信号转化漏斗与突破传统的离群回报率 (0:04) PSIS系统每日从60万构建者社区的互动交流中持续抓取数据，将企业高管之间发生的大约500个关键对话片段（即“时刻”）过滤并聚合成人类可读的结构化信号。在执行层面，Mighty Capital每周开展约一次完整的尽职调查，最终目标是保持每季度完成1至2笔投资的稳健节奏。早期关键成功案例包括基于早期企业合规信号捕捉到的Netskope，以及六年前布局的人工智能基础设施企业Groq，这两项标的均直接源自产品信号追踪。 在投资业绩方面，Moatti对比指出，传统风险投资的幂律法则通常呈现为每20至30笔投资中仅有1笔实现10倍以上的回报；而Mighty Capital凭借这套信号系统实现了高达五分之一（20%）的“离群值”（outlier）成功率，Moatti称该表现达到行业标准基准的4至5倍。迄今为止，该机构的投资组合中已有6家企业实现公开上市（IPO）。 人机协同治理与透明快速的尽职调查 (14:21) 针对自动化投资筛选可能引发的算法偏见与人际信任缺失问题，Moatti强调AI仅被严格限制在漏斗顶端的标的初筛与海量对话解析阶段，而人际关系的建立与最终判断100%保留给人类合伙人。Moatti批评传统风投高度依赖“投资熟人朋友”的模式既缺乏公平性，又容易滋生公司治理丑闻且导致低于标准的业绩表现；引入AI能够在前端拉平竞争门槛，使投资合伙人摆脱低价值社交，专注于与创始人建立真实信任。合伙人在流程中负责校准信号权重，指导项目培育，并牢牢把控最终投资决策权。 在尽职调查执行端，系统亦推行了高效的自动化机制。当创始人提交商业计划书、股权架构表及财务数据后，PSIS能在数分钟内自动生成一份中立、透明的初步尽调报告供创始人审查与调整。Moatti表示，创始人对这种消除偏见、高度透明且快速响应的流程反馈积极，彻底避免了传统机构沟通“已读不回”（ghosting）的弊端。这套数据底座的真实性与高粘性，依托于该机构十年间通过职业支持与CPO智囊团建立起的非交易性信任纽带。 Moatti总结指出，将大规模高保真产品洞察与AI筛选工具结合，为风投机构提供了一条摆脱熟人网络偏见并系统化获取Alpha的路径；然而，这种模式的有效性前提在于能否长久维系这套60万人网络的非交易性信任，并在企业级对话的保真度与自动化分析的泛化性之间保持动态平衡。

## Background 该视频由频道 Ai4 于 2026 年 8 月 20 日发布在 YouTube 平台。内容为 Mighty Capital 管理合伙人 SC Moatti 的主题演讲，介绍其机构如何结合 NLP、智能体工作流与 RAG 优化打造“产品信号智能系统”（PSIS），通过挖掘超 60 万名产品领袖十年沉淀的海量对话，在早期投资中建立起可持续的数据驱动优势。 ## TL;DR SC Moatti 指出传统风险投资常受限于私人关系圈且存在严重的信息滞后，为此 Mighty Capital 提出了“产品超额收益效应”方法论，并构建了 PSIS 智能系统来自动化挖掘全球 60 万产品人网络中的非结构化对话。该系统将一线洞察提炼为数百个关键时刻（Moments）与结构化信号（Signals），指导团队精准确立投资假设、筛选标的并实现自动化的快速尽职调查，使其投资组合达成了 20% 的 10 倍以上高回报胜率并收获 6 起 IPO。Moatti 强调，AI 仅用于前期数据挖掘初筛与尽调提效，最终决策与创始人信任培育始终由人类合伙人掌控，其核心壁垒在于历经十年积累的《财富》1000 强产品高管生态与真实人际信任。 ## Atmosphere & style 演讲风格干练从容且兼具实操深度，Moatti 以资深硅谷产品经理与早期投资人的双重视角娓娓道来，既有对传统风投“朋友圈投资”弊端的尖锐剖析，也有展示系统界面与 AI 智能眼镜案例时的务实拆解。现场互动坦诚真挚，面对关于“算法是否削弱人际温度”的提问，她条理清晰地厘清了 AI 赋能效率与人类坚守终极决策的边界，营造出兼具前沿科技感与扎实商业洞察的专业氛围。 ## Key takeaways - [00:04](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=4s) 针对风投行业往往最后获知技术趋势的信息滞后问题，Mighty Capital 确立了“产品超额收益效应”（Product Alpha Effect）方法论，借助自主研发的产品信号智能系统（PSIS）捕捉一线产品专家的前瞻洞察。 - [00:04](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=4s) 系统底层的核心数据源自全球 600,000 名产品经理与首席产品官（CPO）构筑的庞大网络，覆盖全球约三分之一的产品建设者，将十年前的小型晚餐会交流闭环通过 AI 规模化放大。 - [02:52](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=172s) PSIS 通过 NLP 引擎每日持续摄取行业对话，自动识别高价值对话切片（Moments）并聚类为约 500 个结构化信号（Signals），帮助人类投资团队精准确立创新的方向与投资假设。 - [02:52](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=172s) 依托数据驱动与信号流水线，Mighty Capital 颠覆了传统风投“二三十投一中”的残酷幂律，实现了 20%（五分之一）的 10x+ 顶尖项目命中率，投资效率达到行业平均水平的四至五倍。 - [06:45](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=405s) 系统的数据质感与排他性远胜公开数据库，语料直接取自财富 1000 强 CPO 闭门探讨的 IT 预算流向、系统盲区，以及顾问委员会对每年上万份产品提名的横向深度评审。 - [09:34](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=574s) 在 AI 智能眼镜的实操案例中，系统从 900 多个讨论切片中提炼出 60 个关键技术工程信号（如微型芯片与低功耗传感架构），助力团队在市场共识形成前精准下注领头羊企业。 - [14:41](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=881s) AI 重构了项目入口处的筛选机制，彻底打破传统风投依赖熟人圈子与人情社交的弊端，不仅拉平准入门槛、规避治理风险，还累计跑出了 6 家上市公司（IPO）。 - [15:45](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=945s) 投资全周期设定了清晰的人机协作边界：AI 仅在前端海量信号挖掘中占主导地位，随着流程推进其权重逐步递减，且被严格禁止触碰最终的投资决策权。 - [15:45](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=945s) 在尽调阶段，AI 依托创始人提交的基础文件可在数分钟内自动生成初版尽调报告，并第一时间开放给创始人共同核对修正，以极致的响应速度与透明度建立合作信任。 - [18:24](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=1104s) 底层语料库的高真实性建立在团队过去十年深耕产品社群、陪伴 CPO 阶梯式成长的深度信任之上；算法能够提速流程，但这种历经沉淀的人际连接壁垒没有任何捷径可走。 ## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=4s) 缘起与投资方法论：从晚餐交流到“产品超额收益效应” Mighty Capital 管理合伙人 SC Moatti 以自己的职业转型开场。在成为早期风险投资人之前，她曾是一名创业者，并在早期 Facebook 以及后期的诺基亚担任过产品经理，在硅谷摸爬滚打了整整二十年。踏入早期风险投资领域时，一件事情彻底震惊了她：**产品专家往往是最早洞察未来趋势的人，而投资人却常常是全行业最后一个获知的群体。** 为了打破这种严重的信息滞后，Moatti 在 Mighty Capital 确立了被称为“产品超额收益效应”（Product Alpha Effect）的核心投资方法论，坚信真正的产品创新能够产生出色的超额投资回报（Alpha）。而将这一方法论在全行业规模化落地的核心武器，便是他们自主研发的“产品信号智能系统”（PSIS，Product Signal Intelligence System）。 这一系统底层的“秘密武器”，是一个汇聚了全球 600,000 名产品经理（PM）与首席产品官（CPO）的庞大产品人生态——这意味着全球每三位产品建设者中就有一人参与其中，覆盖各大地理区域与全行业。Moatti 回忆，这一网络源于她十年前发起的一个非营利组织。起初，她在打造产品之余受邀撰写一本关于“优秀产品特质”的书籍。在写作与新书推广期间，她经常邀请同行围坐共进晚餐，探讨产品话题。席间常有人聊起：“嘿，我正在开发一款很酷的新产品”或“我准备成立新公司来做它”，Moatti 便会出资投资；随后，她又转头向同一张餐桌上的其他产品领袖推荐：“我刚投了一款非常有意思的产品，你们应该去采购、做试点测试。”如今 Mighty Capital 所做的事情与当年餐桌上的闭环并无二致，只不过规模已呈指数级爆发，以至于必须依赖 AI 才能完整捕捉这 60 万人网络中的海量洞察。 ### [02:52](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=172s) PSIS 运作流水线：从碎片对话到季度投资决策 整个 PSIS 系统每天都在持续摄取这 60 万名产品建设者跨网络发生的真实对话。这些对话可能涉及某个金融行业为了应对新型 AI 网络安全攻击所急需的合规新工具（例如多年前投资 Netskope 的契机），也可能是科技企业如何更高效利用 AI 的全新工具（例如六年前投资芯片公司 Groq）。系统在后台不间断运转，当基于自然语言处理（NLP）的机器学习引擎在跨领域的讨论中抓取并积累出约 500 个高价值的对话切片（内部称之为“Moments”，时刻）时，便会触发对人类投资团队的提醒，示意“该对这个方向展开深度挖掘了”。 ```mermaid flowchart TD accTitle: PSIS 从对话到投资的工作流 subgraph DataIngest["数据采集与初筛"] A["60 万产品人社区对话"] --> B["NLP 识别 Moments<br/>每日捕捉高价值对话切片"] B --> C["聚合为约 500 个 Signals<br/>提炼出人类可读的声音摘要"] end subgraph Synthesis["投资假设与标的筛选"] C --> D["交叉比对深度访谈<br/>确立创新方向与强度"] D --> E["锁定赛道潜在公司<br/>建立联络与赋能"] end subgraph Pipeline["尽调与投资闭环"] E --> F["每周启动 1 次深度尽职调查"] F --> G["每季度完成 1-2 笔投资决策"] end ``` 这些时刻会被聚合提炼为便于人类消化理解的“信号”（Signals）。随后，投资团队会将这些信号与产品高管的直接对话进行交叉验证，从而精准标定创新的“方向”（Direction）与“强度”（Intensity），以此构筑 Mighty Capital 的投资假设（Thesis）。Moatti 坦率承认，团队自身已有多年不再一线操刀写代码或做产品，并没有能够预言未来的“水晶球”，他们完全仰仗这群一线产品构建者指明未来走向。顺着既定方向与强度筛选出标的公司后，团队会主动建立联系、提供赋能，并保持约每周启动 1 家深度尽调的节奏，最终稳定转化为每季度 1 到 2 笔的高胜率投资。 - **传统 VC 幂律法则的反叛**：传统风险投资界普遍接受“二三十家公司中仅有一家能跑出 10 倍以上回报（Outlier）”的残酷幂律；而在数据驱动与 PSIS 系统的加持下，Mighty Capital 的投资组合中达到了 **五分之一（20%）的 10x+ 离群胜率**，整体效率达到同行的四至五倍。 - **数据飞轮效应**：数据量越充分，系统的精准度与产生超额回报的能力就越强；Moatti 强调，没有这套底层 AI，如此庞大规模的非结构化专业对话根本无法在人力有限的早期基金中被有效利用。 ### [06:45](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=405s) 数据质感胜过一切：何为高保真“产品信号” Moatti 在逛展会展厅时深有感触：无论在风投还是其他行业，“优质数据才能产生优质结果”始终是不变的铁律。PSIS 之所以表现超群，根源在于它所喂入的数据资产具有极高的保真度与排他性。这些数据直接来源于财富 1000 强企业（如沃尔玛、强生、迪士尼、思科等）首席产品官的闭门交流——他们探讨的是“我的技术路线图在往哪里走”、“当前的产品战略是什么”、“我们的系统在哪里存在致命盲区”以及“下个季度的 IT 预算正流向何处”。 除了 CPO 级别的战略探讨，另一大核心数据源是该网络中约 1000 名精挑细选的顾问委员会（Advisory Council）高管。该机构连续八年主办产品大奖，每年收到 12,000 至 15,000 份产品提名。顾问委员会成员会对每一份提名展开详实评审：“这家公司之所以优秀，是因为它在 X、Y、Z 维度彻底碾压了竞品，理应获奖”。这种包含横向评测、具体优缺点的评审意见构成了全球最前沿创新的纯净样本。Moatti 强调，团队多次将这些内部洞察与 PitchBook、LinkedIn 以及 Reddit 等公开数据源进行对比验证，结果显示，直接来自产品高管的一手见解在预测产品创新与商业转化上，其可靠性远超任何公开数据库。 ### [09:34](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=574s) 案例复盘：AI 智能眼镜会成为终极杀手级设备吗？ 为了让观众具象感知 PSIS 的工作流程，Moatti 以“AI 智能眼镜”为例展示了系统的实际操作界面。团队曾向系统抛出一个核心问题：“眼镜会成为生成式 AI 的终极杀手级硬件设备（Killer Device）吗？” ```mermaid flowchart LR accTitle: AI 眼镜案例中的数据收敛路径 subgraph RawData["原始语料层"] M["900+ 个专业讨论 Moments<br/>覆盖开发工具、边缘硬件等圆桌会"] end subgraph Intelligence["智能提炼层"] M --> S["60 个核心 Signals<br/>明确技术壁垒与实现前置条件"] S --> W["人工复核与校准权重<br/>确立极高置信度与创新强度"] end subgraph Portfolio["落地执行层"] W --> C1["领头羊已投项目<br/>对接三星、Anthropic、OpenAI"] W --> C2["储备观察池项目<br/>多重标签重合度持续走高"] end ``` 在系统后台，PSIS 从浩如烟海的高管对话中锁定了约 900 个涉及该话题的“Moments”（时刻）。Moatti 解释说，界面中的一个 Moment 可能来自某个顾问委员会成员在开发者工具圆桌会上的随口断言，系统会自动打上标签，标注其是对“眼镜是 AI 未来载体”这一主张的强支持、弱支持还是反驳。随后，这 900 多个讨论切片被聚类归纳为约 60 个明确的“Signals”（信号），具体内容涉及诸如“眼镜要成为杀手级设备，芯片尺寸必须缩小至能塞进镜腿”、“必须具备在极小功耗下捕捉环境信号的全新传感器架构”等具体工程前置条件。 投资团队介入对这 60 个信号进行权重打分与持续校准。对于置信度最高的论点，系统自动匹配并排序出正在该方向上全力推进的潜力初创公司。排在首位的是一家 Mighty Capital 已经下注的明星企业，该公司目前正与三星等财富 1000 强硬件巨头，以及 Anthropic、OpenAI 等 AI 超大规模算力巨头紧密合作，推动智能眼镜落地；而在候选列表下方，还有一批被打满蓝色高相关标签的候选企业，系统正实时跟踪它们的信号密集度，以便团队随时切入尽调。 Moatti 总结道，风投行业必须走出“20 笔投资只指望成 1 笔”的低效惯性，通过内化 AI 工具，投资人完全可以在早期筛选与后期赋能上做得更好。在进入 Q&A 环节前，面对现场观众关于“纯 AI 筛选是否会丢失人际关系温度”的疑问，Moatti 立即明确回应：本次分享聚焦 AI 是因为置身于 AI 峰会，但在实际投资体系中，“建立真实深度的人际信任”始终是决定成败的最核心环节，AI 工具的使命正是让人类能把精力集中在最具价值的关系构筑上。 ### [14:41](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=881s) 重构风投筛选机制与传统“朋友圈投资”的弊端 Mighty Capital 所颠覆的并非风投行业中建立人际关系的核心价值，而是项目前期的“筛选环节”（Selection）。SC Moatti 回顾了自己十年前刚进入风投资本领域的经历，发现当时绝大多数同行基本只给自己熟识的朋友投资（"invest in their friends"）。在她看来，这种依赖私人圈子的投资方式既不公平也无法带来更优的回报，反而常常导致基金业绩下滑、朋友反目，以及媒体上屡见不鲜的各类公司治理丑闻与闹剧。 通过在漏斗入口引入基于数据与 AI 的产品信号智能系统，投资机构得以拉平准入门槛。投资团队无需再把海量时间耗费在低效无益的盲目社交与握手应酬上，从而能够将精力集中在与优秀创业者建立高质量、深度的合伙关系上。这种人机结合的模式直接体现在了基金亮眼的业绩表现中：投资组合中涌现出 20%（五分之一）的顶尖黑马（Outliers），并成功收获了 6 家上市公司（IPOs），这是扎实的筛选机制、团队辛勤付出与深度投后赋能共同作用的成果。 - **打破圈子投资模式**：摒弃传统风投仅投熟人的裙带倾向，从源头规避治理隐患与人情交易。 - **投资效率与成效**：节省无效社交成本，实现高达 20% 的超级项目命中率（1 in 5 outliers）以及 6 起 IPO。 ```mermaid flowchart TD accTitle: 投资筛选与决策流程图 accDescription: 展示从海量信号抓取到最终人类决策的分阶段流程 subgraph Phase1["前期筛选阶段（AI 占主导）"] A["600,000+ 产品领袖对话与信号"] --> B["NLP / RAG 数据挖掘引擎"] B --> C["治理微调与信号一致性验证"] C --> D["生成潜在优质项目 Shortlist"] end subgraph Phase2["尽调与培育阶段（人机协同）"] D --> E["合伙人认领项目一对一深度培育"] F["创始人提交：Deck / Cap Table / 财务数据"] --> G["AI 自动生成标准化尽调初版报告"] G --> H["数分钟内反馈创始人共同校对修正"] end subgraph Phase3["最终投资决策（纯人工主导）"] E --> I["投资决策委员会深度评估"] H --> I I --> J["最终投资决策（绝对禁止 AI 介入）"] end ``` --- ### [15:45](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=945s) AI 筛选与人类决策的协作边界及尽调自动化 在现场问答环节，来自 Xsolla Game Commerce 的产品专家 Sasha 针对其 20% 的超高命中率提问，探究系统在 AI 层面设置了哪些护栏（Guardrails），以及投资决策究竟由 AI 主导还是倚重人类在环（Human-in-the-loop）。SC Moatti 明确阐述了 AI 与人类角色在投资全生命周期中的动态消长：在项目最前端的信号挖掘与海量数据筛选中，AI 扮演着不可替代的高权重角色，因为没有任何人类能够凭肉眼逐一通读数百万场对话与行业讨论；但随着项目推进，AI 的权重逐步递减，合伙人的人际连接与专业判断权重则持续上升。 在实操中，系统通过内置的多维度参数和治理机制对正反向信号进行动态校准与微调。然而，一旦筛选出项目短名单（Shortlist），AI 就被严格限制——它绝不代替人工进行人际沟通，更“绝对不允许触碰最终决策”（"definitely never touches the decision-making"）。合伙人会各自认领感兴趣的目标企业，进入纯粹的人际信任培育环节。令人意想不到的是，在后续的尽职调查阶段，AI 同样发挥了极佳的协同作用： 1. **秒级自动生成报告**：创始人只需提交 Pitch Deck（商业计划书）、Cap Table（股权结构表）以及财务报表三份基础文件，AI 便能在几分钟内自动提炼并生成标准化尽调初稿。 2. **客观中立与极致透明**：投资团队会第一时间将报告回传给创始人，并告知：“这是系统基于您提交的材料生成的报告初稿，请随意核对并指出需要调整修改的地方。” 3. **极速响应增强信任**：这一流程不仅保证了客观与公开透明，还大幅缩短了反馈周期，让创始人感受到尊重与极高的沟通效率，进一步巩固了彼此的合作关系。 --- ### [18:24](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=1104s) 底层数据真实性与十年产品经理生态的信任壁垒 针对如何确保底层语料库的真实性与高可信度、让产品经理们愿意展开坦诚交流的问题，SC Moatti 指出，这一庞大的产品领袖网络绝非一朝一夕之功，而是 Mighty Capital 团队在过去十年间“一个关系接一个关系”（"one relationship at a time"）扎扎实实积累起来的人际信任资产。 如今打开《财富》1000 强（Fortune 1000）企业名单，几乎绝大多数公司的首席产品官（CPO）都在其职业生涯早期与 Mighty Capital 建立过深厚交集。当他们还只是崭露头角的总监、急需打造个人品牌以招揽优秀团队时，团队就曾动用资源为其引荐核心人才；当他们晋升为副总裁并准备迈入 C-suite 高管层时，机构又为其提供职业转型的指导与支持。 正是通过 CPO Mastermind 闭门私董会、产品大奖（Product Awards）等长期运营的高端社群与项目，机构赢得了整个产品圈层的深度信任。Moatti 强调，AI 算法和自动化管道可以加速流程，但这种历经十年沉淀、充满温度的人际连接与真实交流，“是绝对没有任何捷径可走的”（"that part you cannot shortcut"）。 ## Notable quotes > [00:04](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=4s) 产品专家往往是最早洞察未来趋势的人，而投资人却常常是全行业最后一个获知的群体。 — SC Moatti > [00:04](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=4s) 嘿，我正在开发一款很酷的新产品 — SC Moatti > [00:04](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=4s) 我刚投了一款非常有意思的产品，你们应该去采购、做试点测试。 — SC Moatti > [06:45](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=405s) 优质数据才能产生优质结果 — SC Moatti > [06:45](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=405s) 这家公司之所以优秀，是因为它在 X、Y、Z 维度彻底碾压了竞品，理应获奖 — SC Moatti > [09:34](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=574s) 眼镜会成为生成式 AI 的终极杀手级硬件设备吗？ — SC Moatti > [09:34](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=574s) 建立真实深度的人际信任 — SC Moatti > [14:41](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=881s) invest in their friends — SC Moatti > [15:45](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=945s) definitely never touches the decision-making — SC Moatti > [15:45](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=945s) 这是系统基于您提交的材料生成的报告初稿，请随意核对并指出需要调整修改的地方。 — SC Moatti > [18:24](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=1104s) one relationship at a time — SC Moatti > [18:24](https://www.youtube.com/watch?v=GQli7KnJJ1Q&t=1104s) that part you cannot shortcut — SC Moatti ## Mentioned SC Moatti, Mighty Capital, Facebook, 诺基亚 (Nokia), 产品信号智能系统 (PSIS), Netskope, Groq, 沃尔玛 (Walmart), 强生 (Johnson & Johnson), 迪士尼 (Disney), 思科 (Cisco), PitchBook, LinkedIn, Reddit, 三星 (Samsung), Anthropic, OpenAI, Sasha, Xsolla Game Commerce, Product Awards, CPO Mastermind ## Video description SC Moatti - Managing Partner, Mighty Capital What if your best Associate never slept, never missed a signal, and had listened to every conversation your industry's top experts ever had? Venture capital has always rewarded pattern recognition but the patterns have historically lived in relationships, intuition, and years of domain expertise that don't scale. At Mighty Capital, we built the Product Signal Intelligence System: an AI engine that mines a decade of conversations with 600,000+ product leaders to surface product signals before they become investment consensus. In this session, you will see how we combined NLP, agentic pipelines, and RAG optimization to turn a proprietary content archive into a continuously operating signal engine. If you want to understand what it looks like to build AI infrastructure that creates a durable, data-driven investment edge, this session will show you exactly how we did it. [Source](https://www.youtube.com/watch?v=GQli7KnJJ1Q)

[0:04] SC Moatti: 产品专家往往是最早洞察未来趋势的人，而投资人却常常是全行业最后一个获知的群体。 [0:04] SC Moatti: 嘿，我正在开发一款很酷的新产品 [0:04] SC Moatti: 我刚投了一款非常有意思的产品，你们应该去采购、做试点测试。 [6:45] SC Moatti: 优质数据才能产生优质结果 [6:45] SC Moatti: 这家公司之所以优秀，是因为它在 X、Y、Z 维度彻底碾压了竞品，理应获奖 [9:34] SC Moatti: 眼镜会成为生成式 AI 的终极杀手级硬件设备吗？ [9:34] SC Moatti: 建立真实深度的人际信任 [14:41] SC Moatti: invest in their friends [15:45] SC Moatti: definitely never touches the decision-making [15:45] SC Moatti: 这是系统基于您提交的材料生成的报告初稿，请随意核对并指出需要调整修改的地方。 [18:24] SC Moatti: one relationship at a time [18:24] SC Moatti: that part you cannot shortcut

[0:04] SC Moatti: product alpha effect [6:45] SC Moatti: good data produces good results

[0:04] I was an entrepreneur before being an investor.  I was also a product manager and I was early at [0:11] Facebook, late at Nokia as well. Um, but I've been  in Silicon Valley for for 20 years. And so when I [0:19] started investing at the early stage, one of the  things that surprised me the most was exactly that [0:28] that product experts see the future first. So I  knew that. But then investors were usually the [0:35] last ones to know and I decided to fix that. And  we invest at Mighty Capital using a methodology [0:45] called the product alpha effect which is  essentially saying that product innovation [0:51] generates alpha like superior performance. And  what I'm going to walk you through today is [0:58] the system that helps us do that which we  call PSIS product signal intelligence system. [1:09] This system works really at scale and the  secret sauce underneath that system is a massive [1:20] community of 600,000 product managers and chief  product officers. Now, where are all these [1:29] product builders coming from? It's about one in  three on the planet across all geographies, across [1:36] all industries. And they come from a movement,  a nonforprofit that I started it 10 years ago [1:45] that um is essentially helping them build great  products. Uh how I started it is um you know [1:55] an interesting anecdote. After building products  myself, I was invited to write a book on what [2:01] makes a great product and to write my book and  then subsequently to to promote it. I was hosting [2:07] dinners with my peers with other product managers  and we were talking about what makes a great [2:13] product and then some would say like here here's a  great new interesting product that I'm building or [2:20] uh company that I'm starting to  build it and I'd invest in them [2:25] and then I'd turn to again that dinner table say  oh I just invested in a really interesting product [2:32] you should buy it and you should pilot it and what  we do today at Mighty Capital is exactly the same [2:37] thing but at just such a massive scale that now  we need AI to help us capture all those insights. [2:45] It's bigger than a dinner table. And then uh and  then we use these insights to to make investments. [2:52] And so our system every day takes  the conversations that are happening [2:58] across that entire network across  those 600,000 product builders. [3:03] conversations about a new tool that is required  to help somebody in finance industry be compliant [3:11] with new regulations that are the consequence  of you know maybe AI cyber security attacks [3:17] that's a corner of the world another corner of the  world a new tool that helps the technology company [3:23] um uh use AI more effectively that was  six years ago when we invested in Grock [3:30] uh the first one was uh a few years ago when we  invested in net scope. Anyways, when that system [3:36] finds a critical mass of snippets of conversations  between product executives, which we call moments, [3:44] when we when the system finds about 500  moments, it signals to us uh to the humans, [3:51] hey, we we we we should dive a little deeper.  And we do. Um so to surface the 500 signals a day [4:00] we use machine learning uh technology NLP  specifically and then we look at those sign [4:07] at those moments and convert them like aggregate  them into signals and the signals are you know [4:14] sound bites that are digestible by humans. Um once  we get those signals, we cross reference that with [4:24] some of the conversations we've had with these  product executives and we come up with essentially [4:32] a uh direction and intensity for the innovation,  right? The signals have a particular direction. [4:39] Um and that's how we come up with our thesis.  So we're investors. uh we haven't been building [4:47] products for many years. So we don't have a  crystal ball to come up with an idea of what the [4:53] future looks like. We really rely on these product  builders to tell us where the future is going [5:00] and then we identify companies that are  going in that direction with that intensity. [5:06] Once we found these companies, we meet  with them. We nurture the relationship. We [5:12] add value to what whatever they are building  and occasionally approximately once a week [5:20] we start a due diligence with one of them. And  doing one due diligence a week gives us enough [5:27] um knowledge about enough companies that we're  able to make one to two investments per quarter [5:35] which is what our own investors are asking us  to do. And so you can see how this system while [5:42] it started at a dinner table is now operating as  at this really large scale and can only operate [5:50] at scale because it's powered by by so much AI,  right? Less data would reduce the uh the accuracy [5:59] and and the um the capacity to outperform of the  system. The system outperforms for sure. When you [6:07] look at the performance of Mighty Capital, [clears  throat] one in five of our portfolio companies is [6:13] an outlier. An outlier being a 10x uh or more on  from our first investment. And that really defies [6:21] what is commonly understood in venture capital  as the power law, which is that one in 20 or [6:27] one in 30 investments becomes an outlier. And so  this is a system that operates, you know, four or [6:35] five times better than any competitor essentially  because it's so data driven and it's so scaled. [6:45] The reason it operates so well is the quality  of the data. And when I was walking the the [6:52] exhibit hall, uh that's really one of the the key  themes that stood out for me. It's not just in our [6:59] corner of the world in venture capital. It's it's  everywhere that good data produces good results. [7:07] Um so the quality of our data is essentially  all the conversations that are happening [7:14] inside that product ecosystem which is really high  fidelity data because it's conversations between [7:22] chief product officers in in any company at  the fortune 1000 like chief product officer [7:28] of Walmart of Johnson and Johnson Disney Cisco  talking about where's my road map going. What's my [7:38] product strategy? Where do I have gaps? Where am  I allocating my budgets? It's conversations among [7:45] members of the advisory council of that  large network who are essentially about [7:51] a thousand handpicked product executives who are  responsible for reviewing nominations for product [7:59] awards that we've been running for eight years.  We receive 12 to 15,000 nominations per year. [8:06] And these uh members of the advisory council  and a subset of them review all the nominations [8:14] and they comment on the nominations like this is a  great company or a great product because it does X [8:20] Y and Z and it's better than this other product  so it really deserves to get an award etc etc. [8:27] So really compelling data about what makes a great  product, what makes a a great innovative company. [8:35] Uh it it also captures um conversations around uh  various topics across all industry, all geography [8:45] that we host that are product roundtables. So  all of that content is clean, highfidelity, [8:54] super highcale data about innovation at the at  the cutting edge, at the bleeding edge everywhere. [9:01] And it's more reliable from a investment or  innovation signal standpoint than most of [9:09] the public databases. We've constantly we've cross  referenced against many of them such as pitchbook, [9:16] such as LinkedIn or Reddit and the the data you  get from the insights of product executive is a [9:23] clear innovation signal that also clearly  translates into investment performance. [9:34] I want to show you I'm going to skip over this  slide because I want to show you how it looks like [9:40] our product signal intelligence system and  the case study that I'm going to use is the [9:46] case study of um AI glasses. The question we po  pose to this you know system, our product signal [9:54] intelligence system, our PSIS is do you think  glasses is going to be the device of of AI, [10:02] the killer device of AI? And the the system  itself um captured about 900 moments across [10:11] all those different conversations that talk about  is is that true? like do product people agree with [10:19] this this statement and from there the system  summarizes it into about 60 different signals. [10:27] A signal will be something like, you know, in  order for glasses to be the device of the killer [10:35] device of AI, you need to have a a chip that's  small enough to fit on a on a pair of glasses [10:42] or you need to have um you know some some ways  to capture signal into the glasses that will also [10:50] fit on on those glasses and so on and so  forth. So that's what a signal sounds like. [10:55] And then from there it converts it into  a set of invest investable companies. So [11:04] here you'll see our our PSIS our our propriatory  tool and this is what a a moment looks like. Um [11:13] and there's 900 of those. a moment  is someone in our advisory council [11:19] in a random round table like the first one  is about developer tools says something about [11:26] you know glasses and and AI and and that's  something you see the tags will strongly support [11:34] the fact that glasses are the device of um the  future of AI and if you scroll down all of that [11:41] I'm not going to do this here I don't have the  the live demo uh I didn't want to take the demo [11:46] risk. Um you'll have you know signals that are  contradicting the claim strongly weakly etc etc [11:57] and then these 900 or so signals are aggregated  sorry moments are aggregated into signals. The [12:06] signals are what you see on the the left hand  side here. And the signals are essentially a [12:11] collection a subset of of moments. And then  these signals are ranked uh with various weights. [12:21] And that's where there's a lot of input from  from humans like we review those signals. [12:26] We review the weights that we put on those  signals. We constantly calibrate the signals. [12:33] And for the highest claim, right, saying yes,  AI glasses are definitely going to be the um [12:41] the the killer device of AI. We then find  companies that are going in the direction of [12:47] the signal at the intensity of the the direction  of of the signal. We come up with a with a number [12:53] of company. The first one is a is a company we  invested in uh which is a rockstar company that's [12:59] working with uh Fortune 1000 like Samsung and and  and with AI hyperscaler like Antropic or or OpenAI [13:08] to enable glasses as the the killer device of  of uh of AI. And then there's companies that we [13:15] haven't yet invested in, but that we're watching  some of them because you can see from all the [13:21] blue tags like in the in the fourth position  because they are part of many of our signals [13:28] and so they become more and more relevant  to our our PSIs right to our to our tool. [13:37] Um that's the gist of what I wanted to share  for for this conference in terms of u how we [13:45] leverage AI internally to really disrupt our own  industry to prove the power law wrong and say that [13:56] it's not okay in venture to just have one  in 20 of your investments be a winner. [14:03] many more companies can be a winner  and we can help by selecting better [14:09] by supporting better and essentially being  better investors and being better partners [14:14] to our CEOs and to our own investors.  And I think there's a few minutes [14:21] I can take some questions if there's any. Yeah.  So the question is um are we missing the human [14:28] portion of it? What I shared here is the AI tool  because we're at an AI conference. But of course, [14:35] there's a massive component of relationship  building. In fact, that's the most important [14:41] component. The part that we're disrupting is  not the relationship building part. It's the [14:47] uh selection part. What I've learned when I  started in the venture industry 10 years ago [14:52] is that most of my peers will invest in their  friends. And I don't think it's fair or better. [15:00] I think it's actually unfair and and results in  worse performance and broken relationships and [15:07] all the things you read about in the press  about governance scandals and stuff like that. [15:12] And so by leveling the playing field at entry with  a system like this, you first of all, you have a [15:18] lot more time. You don't spend all your time like  shaking hands in a way that's not productive, [15:24] right? you have a lot more time to build quality  partnerships with your entrepreneurs and it [15:29] shows in the in the success rate of the of the  firm. You know, one in five outliers, six IPOs, [15:37] uh is is a sign of, you know, a lot of hard  work and and human input and partnership. Hey, [15:45] um Sasha, product guy from Exola Gamecommerce.  Uh thank you. It's pretty cool to see how it all [15:52] works. uh you got a like crazy high success  rate for the investment 20% is like overkill [15:59] obviously I was so much surprised so my question  in the application for the AI what guard rails [16:06] your system uses to hit this high standard is it  implied in the system as well or you do the human [16:13] in the loop mostly to define the your investment  thank you yeah so that's a great question and [16:18] the the role of AI is high at the beginning and  it gets lower and lower as the role of the the [16:27] partnership and the human relationship increases.  So in particular there's literally no human who [16:34] can parse through the millions of conversations  and moments and so we do rely on on AI to do that. [16:42] We do have a bunch of parameters. Um, for example,  I showed you how some of the signals will support [16:49] a particular claim and others will go in a  different direction. We constantly fine-tune [16:54] that with governance and we have a lot of control  over that. Uh, once we get to a list of companies, [17:03] there's a role for AI, but it almost never touches  uh the human interaction and it definitely never [17:09] touches the decision- making. And so once we say  surface a company I showed you a short list. Um [17:17] each of us will take a number of them to nurture  and the nurturing is a pure connection uh element. [17:24] Uh once we get into diligence AI itself has  a role and it has had a to my surprise a very [17:32] positive role. Um we automate the creation of  our reports based on two or three documents that [17:42] our entrepreneurs uh submit their their pitch  deck, their cap table, their financials. And [17:47] from there we send within minutes a report that  says this is based on what you submitted. This [17:53] is what our report would look like. Please adjust  as you see fit. And so entrepreneurs see that it's [18:00] number one, it's neutral, right? Makes things a  lot easier. Number two, it's super transparent, [18:05] which is also appreciated. And number three  is fast, and we actually respond to emails, [18:11] which is uh which is also something that's helping  tighten the relationship. So the EI is really in [18:18] service of the partnership and the relationship  in those later stages of the the diligence. [18:24] So the question is how do we create integrity  in the in the initial data set which is like [18:31] encouraging like honest authentic conversations  between these product folks. The thing is [18:38] this product network has been built one  relationship at a time over the past 10 years. [18:45] You would pick almost any chief product officer  in the Fortune 1000 and I can tell you, oh yes, [18:51] we met them when they were, you know, an emerging  director trying to raise their personal brand so [18:58] they could hire a great team and we helped them  find some really cool people in the network [19:03] and then once they become became VP, we've helped  them, you know, make a transition into the seuite [19:10] and, you know, so we we've sort of earned  the trust one relationship ship at a time [19:15] and help foster success and conversation and  authenticity through a lot of those programs [19:21] I was telling you about like the CPO mastermind,  the product awards, etc. Um, so that's that part [19:30] you cannot shortcut like it's essentially 10  years of patiently building human connections.