Henrik Landgren, founder of Gilion, is pushing venture capital firms to rethink how they deploy AI. The problem isn't AI itself. It's garbage data feeding AI systems.

Landgren argues that VCs are applying machine learning to poor-quality information sources, undermining the entire analytical value proposition. The fix requires building proper data infrastructure from the ground up, connecting directly to authoritative sources like financial systems, payment platforms, and accounting software. This eliminates the noise and intermediaries that corrupt datasets.

Better data infrastructure delivers concrete wins for investors. Due diligence becomes faster and more reliable. Pattern recognition improves, surfacing overlooked startups that traditional screening misses. Investment decisions themselves become both quicker and more accurate, giving firms competitive edges in hot deals.

The VC world has embraced AI tools with enthusiasm, but many implementations skip the unsexy foundational work of data hygiene and infrastructure. Firms pull data from multiple fragmented sources, apply AI models, and wonder why results disappoint. Landgren's thesis inverts the priority: data quality precedes useful AI application.

Gilion operates in this gap. Rather than wrapping AI around existing messy datasets, the company focuses on the plumbing. Direct connections to financial and payment systems create clean, real-time data feeds that AI can actually work with.

This matters because venture capital decisions rest on information asymmetries. Better data accessibility and quality compress those asymmetries, democratizing access to high-quality information. Early-stage investors could level the playing field against mega-funds with larger research teams.

The argument also confronts a broader startup culture myth. Tech companies often assume AI solves everything if you feed it enough data. Landgren counters that data source quality matters infinitely more than data volume. One clean financial statement beats a thousand scraped web pages.

VCs sitting on underperfor