E411: How AI Is Creating a New Generation of Venture Firms

Episode Summary

Executive Summary: Andre explains how a machine-learning mindset led him to automate venture capital at Earlybird: building AI systems to reduce repetitive work, improve sourcing and screening, and institutionalize judgment. He argues VC is increasingly shaped by power-law outcomes, founder branding, capital-raising ability, and sector-specific founder fit, while AI mainly adds edge through proprietary data, better prep, and faster, more disciplined decisions.

Main Topics: Why Andre Joined Earlybird (Priority: 4/5): He chose Earlybird after researching Europe’s top VC firms and selecting the most technical, engineering-oriented one, aligning with his own engineering/ML background. VC as a Groupthink-Ridden Power-Law Industry (Priority: 5/5): He argues the industry became network-driven and consensus-seeking because false negatives are extremely costly in a power-law world where a few winners drive most returns. Portfolio Construction, Discipline, and Anti-Portfolio Learning (Priority: 5/5): Earlybird historically led rounds and aimed for strong ownership, but Andre emphasizes the cost of missing winners like Lovable and the need to adapt thresholds over time. Founder Traits by Stack: Capital vs Talent (Priority: 5/5): He distinguishes deep tech, infra/frontier AI, and application-layer companies, saying founder requirements vary: capital attraction matters more lower in the stack; talent attraction and distribution matter more higher up. AI-Native Venture Capital Platform (Priority: 5/5): Andre describes Earlybird’s shift from dashboards to AI chat interfaces, using AI for meeting prep, memo drafting, term-sheet benchmarking, network intelligence, and internal knowledge workflows. Metrics, Measurement, and Institutionalizing Judgment (Priority: 4/5): He stresses that measuring coverage, hit rate, committee decisions, and follow-on outcomes allows the firm to codify tastes, improve sourcing, and reduce dependence on individuals. Learning, Unlearning, and Change Management (Priority: 4/5): The biggest obstacle to AI adoption was cultural resistance, not technical feasibility; teams needed incentives, quick wins, and new workflows to unlearn old habits.

Key Arguments: Venture capital is not purely a people business; ML can automate repetitive work and improve both efficiency and decision quality. VC has become more groupthink-prone because investors fear missing rare outliers, making false negatives more expensive than false positives. A disciplined ownership strategy is valuable, but rigid thresholds can cause costly misses; flexibility must evolve with market conditions. Founder quality is not one-size-fits-all: different sectors require different combinations of research credibility, capital attraction, distribution skill, and policy access. Founder branding is now a scalable asset because it helps attract talent, capital, customers, and partners in crowded markets. The real alpha in AI-native VC comes from proprietary data: meeting transcripts, investment memos, committee notes, and decision patterns that competitors cannot access. The best use of AI is to remove low-value screening and preparation so humans can spend more time on a small number of high-quality decisions. Adoption failures in VC are often behavioral and organizational; metrics and incentives are needed to drive workflow change. Great founders and investors need a 'prepared mind'—the right information diet and mental space to recognize opportunity when it appears.

Data Points: Earlybird age: 30 years old (next year) - Andre says Earlybird is one of the oldest and most established European VC funds. Portfolio composition (historical): About 80% B2B companies - This was part of why Andre viewed Earlybird as the most technical firm. Initial ownership target: 15% initial shareholding on average - Earlybird’s target across the fund for ownership in investments. Strong shareholdings historically: Toward 20% or more - Andre cites historical Earlybird deals where initial ownership was very high. Lovable missed at pre-seed: Around 8% ownership opportunity - Andre says Earlybird passed because the shareholding was below threshold. Lovable implied outcome: North of 50x, most likely fund-return level - Andre estimates the opportunity could have returned at fund level. Company founder age pattern: Average founder age: 34 - He says unicorn founder age has been robust over 15 years. Earlybird engineering team peak: 8 full-time engineers - At peak, engineering was about 20% of the firm’s total team. Current engineering team: 3 senior engineers - Today they maintain and expand the data infrastructure. Inbound opportunities: 10,000 investment opportunities per year - Earlybird’s strong brand and network generated this annual inbound flow. Coverage before platform improvement: 70-72% hit rate - In 2019-2020, Earlybird saw roughly this share of competitors’ relevant opportunities in Europe. Current coverage: 95-96% hit rate - Andre says Eagle Eye now sees 19 out of 20 opportunities in Europe. Timing advantage: At least 6 weeks before funding round close - Their system surfaces opportunities well before rounds close. Portfolio size: About 35 companies - Andre says an ideal early-stage fund portfolio is around this size. Annual investment count: 10-12 investments per year - Derived from the firm’s portfolio construction target. Benchmark dataset: 1,000+ investment memos - Used to train memo-generation systems. Committee survey archive: 200+ investment committees - Used to capture decision data and correlate it with outcomes. Machine learning benchmark: As good as the best investor in a sample of 120 investors - A 2020 paper benchmarked machine learning against human investors at saying no. Fund count: 17 funds - Andre references Earlybird’s track record across multiple funds. UM assets: $2.5 billion under management - Used when discussing firm growth and compounding brand effects.

Pivotal Quotes: "I thought I disagree. And that triggered me to do some research." — Andre: Explaining how he approached the claim that venture capital cannot be automated. "The combination of the three really leads to investors looking into the mainstream and getting into this groupthink." — Andre: Describing network closeness, power-law economics, and FOMO in VC. "The biggest blocker for a majority of the investment firms." — Andre: Referring to cultural change as the hardest part of AI adoption in VC.

Implications: VC firms will increasingly compete on proprietary data, workflow automation, and judgment codification. Human taste still matters, but winners will use AI to narrow the funnel, prep better, and focus scarce time on the few decisions that drive returns.