E418: AI, Venture Capital & the Future of Investing

Episode Summary

Executive Summary: The conversation argues that AI is reshaping venture capital itself, not just portfolio companies. It highlights how VCs use LLMs for deep research, diligence, memo red-teaming, and portfolio analysis, while firms remain mostly grassroots in adoption. The discussion also covers MCP-enabled workflows, buy-vs-build decisions, the importance of culture and values in AI-native firms, and where venture alpha will come from as routine document parsing becomes automated.

Main Topics: AI as a transformation of venture capital workflows (Priority: 5/5): AI is changing how VCs research founders, source deals, conduct diligence, manage portfolios, and handle back-office work. Adoption is experimental and often driven by individual investors rather than firm-wide mandates. Research, sourcing, and diligence with LLMs and data connectors (Priority: 5/5): VCs use internal notes, internet data, and specialized datasets to build pre-meeting dossiers, accelerate learning in new technical domains, and tailor explanations to their own technical fluency. MCP helps unify these workflows across tools like Notion, Salesforce, and Affinity. Grassroots adoption versus institutionalization (Priority: 4/5): Most AI usage in VC is bottom-up, with investors choosing tools like Claude, ChatGPT, Perplexity, Gemini, and enterprise stacks based on personal preference and compliance constraints. Some workflows, such as memo red-teaming, eventually become firm policy. MCP and the future of software interfaces (Priority: 4/5): Model Context Protocol is presented as a layer above APIs that lets users prompt an AI agent to access software, making products more 'headless' and easier to use through natural language while preserving structured backend data access. Build versus buy in venture software (Priority: 5/5): Firms should build internal, single-player tools when workflows are bespoke, but buy external software for multiplayer use cases involving portfolio companies, auditors, compliance, and verifiable audit trails. Network effects and data moats also influence the decision. What AI commoditizes and where alpha remains (Priority: 5/5): Routine document parsing and data extraction are becoming automated, while human judgment, founder support, relationship-building, and evaluation remain core sources of value. The speaker argues alpha shifts toward taste, trust, and difficult-to-evaluate work. Hiring, culture, and becoming AI-native (Priority: 5/5): To build an AI-native firm, culture and mission alignment matter more than raw intelligence alone. The speaker emphasizes slow hiring, strong values screening, working in the open, and developing managers from within.

Key Arguments: AI is transforming VC workflows end-to-end, from sourcing and diligence to portfolio support and back office operations. VCs are using LLMs most effectively for deep research before meetings, especially to generate dossiers on founders, companies, LPs, and sectors. MCP makes it easier for AI tools to use company software by abstracting APIs into prompt-based interactions, enabling more seamless agent workflows. Most AI adoption in VC is still grassroots and path-dependent on individual investor preferences, not top-down firm mandates. Some AI workflows, like red-teaming investment memos, are valuable enough to become standardized firm processes. Firms should build internal tools only when the use case is highly specific and internal; external, auditable, multi-stakeholder workflows are better bought from vendors. The next frontier in venture alpha will not be parsing documents by hand but applying human judgment, founder support, and strategic decision-making. As AI increases productivity, firms may externalize more mundane work and internalize more strategic or core-competency work. AI-native firms need a culture that embraces transparency, experimentation, and strong alignment with mission and values. The best hiring decisions in startups depend heavily on values and mission fit because small teams amplify misalignment quickly.

Data Points: Standard Metrics customer base: 12,000+ companies - The speaker referenced the size of the platform's dataset used for venture analytics. Company size: 68 people - Used to describe the speaker's company scale while discussing culture and hiring. Revenue run rate in AI example: 60 billion revenue run rate - Cited while discussing the scale of AI adoption and industry revenue numbers. Fund reserves allocation: 30-50% reserves - Used to explain why follow-on investment decisions matter so much for fund performance. SpaceX returns example: $127 billion returned - Mentioned in the context of concentrated investing and backing winners repeatedly. Top decile growth: Fastest growing AI companies growing faster than any companies ever in human existence - Described qualitatively to emphasize extreme growth at the top end, without a precise numeric figure. Revenue per FTE crossover: AI companies crossed non-AI companies for $100M+ revenue firms - Referenced a report showing AI companies overtook non-AI peers on revenue per employee for the first time. Onboarding process: Better part of a week - New hires are flown to San Francisco or New York for in-person onboarding with their manager. Engineering managers: 4 of 4 started as ICs - All current engineering managers at the company were promoted from individual contributor roles.

Pivotal Quotes: "AI is going to transform venture capital itself." — Speaker: Central thesis of the discussion about AI changing VC workflows and operating models. "The most important thing is probably the people that really desire to build that kind of firm." — Speaker: On what matters most when creating an AI-native venture firm: culture and people before tools. "Hire slow and screen relentlessly for mission and values alignment." — Speaker: Advice the speaker would give their younger self about startup hiring and team cohesion.

Implications: VC firms that adopt AI thoughtfully can research faster, improve diligence, and spend more time on high-value judgment and founder support. The winners will combine strong data infrastructure, human trust, and culture-driven adoption rather than relying on tools alone.