A 24/7 agentic system that continuously reviews member records against HEDIS quality measures, classifies why each care gap is open, and routes the right action to the right team months before the CMS reporting window closes. It runs as two agents over one shared evidence base:
A scheduled sweep across the entire member population that returns a prioritized queue of open care gaps, designed to run against EHR, lab, and claims feeds. Each gap is classified by why it is open: missing care, missing documentation, or a failed data submission. A deterministic classification and routing engine, written in code rather than prompted, attaches the accountable team to every determination and escalates urgency as the end of the measurement year approaches.
Deep-dive chart review on a single record, imported from an EHR or uploaded directly: a typed note, a lab PDF, a scanned image, or a 90-page hospital discharge packet. It extracts the text layer or runs vision OCR on messy scans, then works through a real multi-iteration tool-use loop, retrieving measure specs and ontology synonyms from a local vector database with deterministic matching and semantic-similarity fallback running in parallel.
A prototype covering two measures (Diabetes HbA1c Control and Controlling High Blood Pressure) through a config-driven measure registry. All patient data shown is synthetic.