LangChain reported that Included Health built Dot, a federated multi-agent healthcare guide, on LangGraph and Deep Agents, with LangSmith used for clinical oversight. According to the post, multi-turn simulation evaluations moved four teams onto Deep Agents in under two weeks without significant regressions, and the deployment saw 75% more chat engagement and 99%+ high-risk detection.
The practical mechanism is using multi-turn simulations as a migration gate, so teams can check regressions on realistic conversations before switching frameworks. The source is a vendor post and does not define how high-risk detection or engagement were measured.