A self-funded plan can generate tens of thousands of paid claims a year. Reading each one line by line, against the EOB and the plan document, is exactly the kind of high-volume pattern work that slows a manual audit down. Duplicate charges, upcoded visits, and unbundled panels all leave a trail, but finding that trail across thousands of claims takes time.
Where we brought in AI
We use Claude, Anthropic's AI, to help our team move through that first pass faster. It reads the itemized bill alongside the plan's records and surfaces the claims that look most likely to carry an error, so our auditors can spend their time where it counts instead of scanning claims that were paid correctly.
A first read, not a final answer
Nothing AI flags gets treated as a finding on its own. Every flagged claim is reviewed and confirmed by a certified auditor before it becomes a dispute. The judgment call, and the responsibility for it, stays with a person.
Your data, handled the same way
Claims data that runs through this process is covered by the same safeguards as everything else we handle: a signed BAA, encryption in transit and at rest, and access limited to the audit itself. You can read the full program on our security page.