Federated AI Agents in Healthcare: What Included Health's Dot Means for Malaysia

AI Summary

Included Health, a US healthcare company, has built "Dot" — a federated AI agent for healthcare navigation — using LangChain's Deep Agents library, the LangGraph orchestration framework, and LangSmith for observability. "Federated" means one front-facing agent coordinates a team of specialist sub-agents instead of a single chatbot answering everything. The two features that make this story worth a Malaysian executive's attention are human handoff (the agent escalates to a person when it hits its limits) and clinical oversight (qualified clinicians supervise the system's behaviour). The pattern is directly applicable to Malaysian hospitals, insurers, and government service portals — provided PDPA obligations around sensitive health data are handled properly.

Key Takeaways

• Federation beats the monolithic chatbot: a coordinator agent delegating to specialist agents is easier to test, debug, and upgrade than one giant prompt trying to know everything.
• Human handoff was designed in from the start, not bolted on — the single most important design decision for any AI touching customers in a regulated industry.
• Clinical oversight means humans with professional accountability stay in the loop, which is what regulators, insurers, and patients actually require.
• Tracing and evaluation tooling (LangSmith, in this case) is the quiet infrastructure that makes agentic AI auditable — without it, you cannot prove your agent behaved correctly.
• The architecture is industry-agnostic. A Malaysian takaful operator, hospital group, or GLC service desk could run the same coordinator-plus-specialists-plus-escalation pattern within a quarter.

What Happened

LangChain documented on its engineering blog how Included Health built Dot, an AI agent for healthcare navigation. Included Health operates in the care navigation business — helping members find the right care, understand their coverage, and coordinate the administrative work around treatment. Dot is the AI system fronting that function.

The build rests on three pieces of technology. LangGraph is an orchestration framework for building stateful, multi-step agent workflows — instead of a single call to a large language model (LLM), you get a graph of steps that can branch, loop, and hand work between