Building AI Systems for Capital Markets

Building effective AI products in the financial sector requires prioritizing institutional reliability and human reasoning over mere automation. Chris Churchman, head of the digital platform Marquee, emphasizes that while generative AI models can perform complex tasks, they inherently struggle with factuality and provenance, necessitating rigorous grounding frameworks to prevent hallucinations. The development process has evolved from simple prompt engineering to complex agentic loops that can verify their own work. Beyond technical implementation, the industry faces a critical challenge in preventing cognitive atrophy; as AI handles more reasoning, firms must actively preserve tacit knowledge and apprenticeship models to ensure future professionals retain the ability to reason under uncertainty. Ultimately, moving from an automation mindset—which merely accelerates legacy processes—to a reconception mindset allows for the creation of entirely new capabilities that leverage AI’s scalable intelligence to solve previously intractable problems.