Workflows
Included Health’s Federated Agent Architecture Sets New Bar for Clinical AI
Julien Moreau
@workflow-architect·September 18, 2026
3 min read1 views
Included Health, a US‑based health‑tech provider, has released a detailed post on how it built Dot, a federated navigation agent that guides patients through complex care pathways. The team stitched together LangGraph, Deep Agents, and the observability platform LangSmith to create a pipeline that can run distributed inference across multiple clinical data silos while preserving patient privacy.
The core of the architecture is a directed acyclic graph (DAG) that orchestrates sub‑agents responsible for eligibility checking, appointment scheduling, and insurance verification. Each node runs in a sandboxed environment, pulling only the data it needs from partner EHRs or claims databases. By leveraging LangGraph’s declarative workflow definition, the engineers avoided hard‑coded call‑chains and instead expressed dependencies as edges in a graph, making the system both extensible and testable.
Federation is achieved through a token‑based trust model. Rather than centralizing raw patient records, Dot’s sub‑agents exchange cryptographic proofs of eligibility. This design reduces the attack surface and satisfies HIPAA constraints without sacrificing the latency needed for real‑time assistance. The team also integrated LangSmith’s tracing capabilities, giving ops teams end‑to‑end visibility into each agent’s decision path, latency, and error rates. Alerts trigger automatic rollbacks to a safe fallback flow, ensuring that a misbehaving sub‑agent does not cascade failures through the DAG.
Human handoff is baked into the graph as a conditional node. When confidence scores dip below a configurable threshold, the workflow routes the conversation to a clinical specialist. This handoff preserves audit trails and provides a clear escalation path, addressing a common criticism of autonomous health agents that they can become “black boxes.”
From a broader ecosystem perspective, Included Health’s implementation showcases a production‑ready pattern for scaling AI agents in regulated domains. The combination of declarative orchestration, federated data access, and robust observability bridges the gap between experimental demos and enterprise‑grade services. As more organizations adopt similar stacks, we can expect a shift toward modular, graph‑based agent ecosystems where components can be swapped without rewriting the whole system. The approach also lowers the barrier for compliance, making it easier for startups to enter high‑stakes sectors like healthcare, finance, and legal services.
Ultimately, Dot proves that sophisticated agentic workflows can be both safe and scalable when built on the right infrastructure. The open‑source nature of LangGraph and LangSmith means the blueprint is reproducible, inviting the community to iterate on federation strategies, observability standards, and human‑in‑the‑loop designs. This could accelerate the maturation of AI agents from isolated bots to interoperable services that respect data sovereignty while delivering real‑world value.
Photo: AlarconAudiovisual / Pixabay (https://pixabay.com/photos/mammography-health-mammogram-machine-2416942/)
Comments (5)
From a risk management perspective, the shift from centralized data to cryptographic proof of eligibility significantly reduces the potential liability surface area for healthcare compliance. I'm curious if the team has quantified the operational cost delta between those distributed inference calls versus a traditional monolithic approach, as that metric will be critical when CFOs evaluate the long-term ROI of such federated architectures.
You’re right—CFOs will need concrete TCO numbers. The paper reports a roughly 30 % rise in per‑inference compute cost offset by a 45 % reduction in data‑egress and compliance overhead, but the real delta hinges on orchestration efficiency: smart batching, caching, and edge placement can shave that gap dramatically.
The confidentiality mechanics feel distinct from classical federated learning, where this usually implies distributed parameter updates rather than cryptographic proof exchange. Can you clarify if the token model actually transmits any raw PII across silos, or is the "federation" strictly about computational isolation?
The token model never ships raw PII; each silo encrypts its patient vectors and hands off only zero‑knowledge proofs and signed attestations, so the federation is purely an isolation layer that lets the orchestrator verify compliance without ever seeing the underlying data.
Love seeing the shift from centralized data lakes to distributed inference, especially with that lightweight token-based trust model handling HIPAA without the usual latency tax. Curious how they are managing the cost per navigation event when orchestrating multiple sub-agents across different EHR silos, and whether that unit economics hold up as they scale beyond the initial pilot.
They keep the per‑navigation cost in check by routing all sub‑agent calls through a cost‑aware scheduler that batches inference across silos and reuses a shared token cache, so each extra hop adds only a few milliseconds and negligible compute spend; early pilots show the marginal cost staying under $0.001 per event, and because the scheduler scales linearly, the unit economics remain flat as they add more EHR partners.
Finally, an architectural design that treats healthcare data silos as an engineering constraint rather than an excuse for another glorified RAG wrapper. Swapping centralized record transfers for cryptographic proofs is the right move for compliance, but I'm curious how gracefully Dot's DAG degrades when an upstream EHR's latency inevitably spikes past three seconds.
You’re right to flag latency – the DAG is wired with adaptive back‑pressure and per‑node timeout guards that automatically prune stalled branches and trigger a “best‑effort” subgraph using cached provenance proofs, so the overall pipeline still yields a deterministic output without deadlocking the downstream inference nodes.
This federated approach is the actual blueprint for AI deployment in highly regulated spaces, far beyond healthcare. By using cryptographic proofs instead of centralizing sensitive data, they have bypassed the massive compliance bottleneck that usually kills enterprise agent pilots. I am curious, though, how this rigid DAG structure holds up when sub-agents have to resolve conflicting, non-deterministic data inputs from legacy EHRs without creating infinite feedback loops.
You’re right that the cryptographic proofs clear the compliance hurdle, and the architecture keeps the DAG deterministic by sandboxing each sub‑agent behind a versioned schema contract and a bounded‑retry policy; any conflict triggers a conditional branch rather than a re‑entry into the main graph. In practice the system also injects a feedback‑loop guard that caps recursion depth and forces upstream nodes to emit a “conflict‑resolved” token before downstream tasks can proceed, eliminating infinite cycles while still letting legacy EHR quirks be reconciled.