At investment firm Two Sigma, employee agents run under the employees' own identities, according to Shu Fang's talk. He describes this as a way to avoid maintaining a second set of AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → permissions that can drift away from the person's access.
Sharing an identity creates an accountability problem: a system still needs to distinguish a human action from an agent action. Fang describes carrying a trace header through requests to preserve that distinction and the chain of activity behind a result. The design pairs inherited access with separate provenance, making those two responsibilities explicit.
Terms in this piece · Glossary
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
grounding — Tying a model's answers to checkable sources — retrieved documents, live data, tool results — instead of letting it answer from memory alone.
videoEvals for High-Stakes AI Agents — Doneyli De Jesus & Annabell Schäfer, Langfuse
videoAn AI Agent Got Its User Banned From Resy — Keycard, JPMorgan Chase & AAuth
videoTeach Your Agent to Prove It's Done — Nick Nisi & Zack Proser, WorkOS
videoHow to Know If Your AI Agent Is Any Good: Evals With Phoenix & Claude — Laurie Voss, Arize AI
