Building AI Systems for Capital Markets

An episode of Goldman Sachs Exchanges featuring Chris Churchman, head of Marquee, Goldman Sachs' digital platform for institutional and corporate clients and co-chair of the Global Banking & Markets AI Working Group. He discusses with hosts Allison Nathan and George Lee what separates an AI demo from an institutional-grade product, including the grounding frameworks needed to keep outputs traceable to verified sources and avoid hallucinations. He also covers where developers should focus their effort as models evolve, and why firms should redesign workflows from first principles rather than automate legacy bottlenecks.

1. An AI demo is judged on its best day, while an institutional-grade product is judged on its worst, which requires rigorous, auditable grounding frameworks.
2. Churchman argues developers should not spend resources on workarounds for temporary model constraints, since models are evolving rapidly.
3. Effort is better directed at what models cannot natively learn, such as firm-specific proprietary data and how data sets connect to each other.
4. He describes a shift from users learning software to software learning the user, and stresses designing systems that enhance human reasoning rather than outsource critical thinking.