Former Spotify VP warns VC: Stop using AI on bad data
Garbage in, garbage out: VCs use LLMs to speed reports, not fix raw data.
Henrik Landgren, former VP of Analytics at Spotify, brings a data-centric perspective to venture capital in a Crunchbase op-ed. He contrasts Spotify's granular user-click tracking with the VC industry's reliance on founder-packaged data, calling it a fundamental information asymmetry problem. Landgren argues that most VC AI adoption is misguided: using LLMs to produce faster reports or summarize pitches does not fix the core issue—bad input data. He emphasizes the old maxim "garbage in, garbage out," noting that AI stacks are often deployed for instant gratification rather than genuine workstream transformation.
Landgren proposes a radical shift: investors should plug directly into a company's raw financial systems—payment records, accounting, marketing performance—instead of waiting for curated pitch decks. This independent sourcing, akin to a home survey not done by the seller, would let analysts start diligence at 70% completion, reserving human judgment for team dynamics and founder charisma. He warns that capital-efficient, high-retention startups are overlooked because polished data hides their true risk profile. Better data access means faster conviction on attractive deals, giving early term sheets a competitive edge. Landgren concludes that fixing data pipelines before layering on AI is the only path to smarter, more confident VC investing.
- Landgren argues VCs adopt AI superficially—using LLMs to speed summaries instead of fixing underlying data quality.
- He advocates for direct access to company financials (payment records, accounting systems) to reduce information asymmetry.
- Independent data sourcing could let analysts start diligence at 70% completion, saving time for human judgment on team dynamics.
Why It Matters
VCs must prioritize raw data access over AI speedups to avoid funding bad bets and missing hidden gems.