Deterministic today
Code mapping, plan-section lookup, rule checks, and the explanation are deterministic in the current demo. Conversational generation is future work.
Healthcare · Member advocacy
Member360 explains what happened after a claim is adjudicated, who needs to act, which plan rule applies, what the member may owe, and what to do next. If it cannot identify a code or rule, it stops instead of guessing.
The problem
An explanation of benefits may contain a denial code without answering the questions a member actually has: Who needs to fix this? Which rule applies? Is there a deadline? What should I say when I call?
That gap creates avoidable calls, delayed appeals, and anxiety. The product is designed around the member’s next decision, not around another claims dashboard.
The product flow
Read a structured, synthetic claim and EOB fixture and identify the outcome that needs an explanation.
Use an exact denial-code lookup. An unknown code stops the flow rather than producing a plausible answer.
Use a deterministic lookup for the named SBC section and keep the explanation tied to that source.
Name the action owner, financial context, call script, and appeal path in plain language.
Trust by design
Code mapping, plan-section lookup, rule checks, and the explanation are deterministic in the current demo. Conversational generation is future work.
The public prototype is deliberately separated from real member records and live payer systems.
Missing evidence is an outcome the product can show. It is not a gap for a model to fill.
What is real today
The public build is an offline working demo with a React interface, a FastAPI/Pydantic service, synthetic fixtures, exact JSON code mapping, and a deterministic SBC lookup. Live payer integration, EDI 835 ingestion, vector retrieval, and production LLM behavior are future work and are not present in the current v2.
Check the public source ↗