Descriptions:
Tashara Fernando, Head of Data and AI at Man Group — the systematic investment manager overseeing more than $200 billion in assets for pension funds, sovereign wealth funds, and major institutions — gives one of the most candid public accounts of end-to-end AI deployment in regulated financial services. Trading signals developed entirely by AI (idea generation, data acquisition, back-testing, strategy writeup, and productionization) are currently running live on real capital at the firm, with humans reviewing but not originating the output.
The technical foundation is Anthropic’s Skills system, which allowed Man Group to encode systematic trading workflows — data cleaning, price stitching, outlier detection, back-test execution — as reusable capabilities that Claude agents can invoke reliably across teams. But Fernando’s most valuable contribution is a frank account of the governance failure mode that nearly derailed the program: power users were encoding personal workflows rather than organizational standards, leading to a cost-center routing bug that surfaced during a show-and-tell and was symptomatic of a deeper accountability gap.
The solution was a centralized skills marketplace where process owners — not power users — are accountable authors. With that governance layer in place, agents could be trusted to execute complex multi-step trading workflows using shared, validated skills. The talk offers a clear, repeatable playbook for any enterprise moving from AI experimentation to production-scale deployment in a high-stakes domain.
📺 Source: Claude · Published May 21, 2026
🏷️ Format: Workflow Case Study
