Research / Theme
AI as Cognitive Infrastructure
Artificial intelligence is often discussed in terms of prediction accuracy or automation efficiency. While these capabilities matter, they capture only a narrow view of AI's potential role in institutional decision-making.
For complex investment environments, the more consequential role of AI lies in supporting cognition rather than replacing it. Institutions do not fail because calculations are unavailable — they fail because reasoning breaks down under uncertainty, pressure, and incomplete visibility. The bottleneck is judgment, not computation.
This research theme examines AI as cognitive infrastructure: systems designed to augment how professionals reason, prioritize, and anticipate — rather than to generate answers in isolation. Such systems help surface blind spots, coordinate multiple analytical perspectives, and maintain coherence as conditions evolve.
Treating AI as infrastructure shifts the focus from outputs to processes. The question becomes not what the model predicts, but whether the system helps decision-makers recognize what must be examined, challenged, or monitored next.
MARY is this research operationalized. The architecture separates deterministic computation — regime scoring, Black-Litterman math, DCF models, backtests — from LLM synthesis, which explains and integrates results rather than generating them. Confidence levels propagate through the system so language matches evidence. Every claim is traceable to a data source. The practitioner retains the decision. This is cognitive infrastructure: it extends analytical capacity without replacing accountability.