Dental insurance still runs on people working phones, payer portals, PDFs, and spreadsheets: verifying benefits, chasing claim status, processing coordination of benefits, validating provider data, credentialing dentists. Overjet is replacing that work with agents (voice, browser, and document agents built on a shared platform) that are already running in production for enterprise payers and DSOs. Every agent that works earns a customer’s trust and shrinks their human queue.
Every agent that doesn’t lands in a human review queue the next morning, with a customer asking why. The Lead PM decides which jobs an agent should be given, how we know it did them correctly, and when a person steps in.
You’ll own an agent product area end to end, reporting to our Insurance product leader and working daily with Engineering, ML, Customer Success, and Sales. What makes this hard is that there is no established playbook for it. Agent behavior is probabilistic, so the spec is an eval set rather than a PRD. Success is measured in tasks resolved without a human, not features adopted. The customers are regulated enterprises with real SLAs, and the workflows are messy, undocumented, and different at every payer.
If turning an operating procedure nobody has written down into an agent that holds up on call number 10,000 sounds like the fun part of the job, keep reading.