Included Health partners with employers to provide healthcare navigation services.
Their Dot platform uses an AI guide built on a federated multi-agent architecture.
Historical routing systems forced users through rigid, predefined decision trees. This approach failed when complex needs required natural conversation flow.
The new system uses LangGraph for its main conversational router structure. Sub-workflows handle specific journeys like scheduling or finding specialists.
Deep Agents ensure consistent voice and tone across all service teams. A global platform prompt manages tone without independent team management.
Filesystem features let agents pass conversation summaries between workflows. Members never repeat themselves when moving between different services.
Skills define service boundaries, eligibility rules, and edge case handling. The model uses progressive disclosure to present the right options first.
A virtual filesystem stores skill descriptions for quick model selection. Full skill files load only when the model decides they are relevant.
