Institutional-grade investing requires more than off-the-shelf AI; it demands domain-specific expertise to bridge the gap between raw data and actionable alpha. Joe O'Donnell, founder of Canary Data and former Tiger Global investor, argues that successful AI integration in public markets relies on a "layer cake" architecture: proprietary datasets, specialized intelligence layers, and rigorous fine-tuning. Because foundation models are inherently "lossy" and lack native judgment, they function best as information-gathering tools rather than autonomous decision-makers. True investment value stems from human-led judgment, which is why firms often struggle to replicate high-performance AI stacks internally. Instead of relying on generic productivity tools, investors must prioritize specialized agents that mimic the precision of expert analysts, ensuring that AI serves as a force multiplier for conviction rather than a source of dangerous, unearned confidence.
Sign in to continue reading, translating and more.
Open full episode in Podwise