Henrik Landgren argues that venture capital firms are applying AI incorrectly by focusing on speed rather than data quality, according to Crunchbase News.
Current VC Data Practices
Landgren states that investing is 50% science and 50% art and relies on founders’ charisma and “it” factor. Once a promising founder is identified, investors receive a vast collection of data that is cherry-picked and packaged by the founder. Pitch decks and company websites can be created in a single afternoon while AI models aid data slicing, increasing information asymmetry.
Landgren’s Background and Critique of AI Adoption
During his time as VP of analytics at Spotify, Landgren used software to store every user click, moving beyond Excel spreadsheets. He found the investment industry far behind in data practices. Most teams adopt AI in a misguided way by using LLMs to summarize pitches or create reports, which fails to improve overall efficiency because an LLM is only as good as its input data.
Proposed Data Infrastructure Approach
The solution begins with direct connections to payment records, marketing performance, accounting systems, and board reports rather than founder-packaged materials. Investors would plug into a company’s financials to see hidden faults before founders decide to disclose them. This allows analysts to start diligence at 70% completion instead of from a standing start, reserving human judgment for team assessment and the “it” factor.
Competitive and Future Implications
Better data access makes investors more competitive in attractive deals by enabling faster conviction without weeks spent locating and cleaning data, according to Crunchbase News. In five years, companies worth funding will include AI-powered hardware, infrastructure, and new deep tech categories that require reassessment of performance evaluation since traditional income models will no longer apply. The industry must stop asking how AI can speed up existing processes and instead determine what a better process looks like, according to Crunchbase News.
