What the work involves

A frontier AI lab is building systems that draft appeals, categorize denials, and surface root causes across the revenue cycle. Your job is to tell them where those systems are wrong — and why, in terms a model can learn from. Day to day that means reading AI-generated appeal letters and judging whether the clinical argument, the cited policy, and the supporting documentation would actually hold up with the named payer; checking whether a denial was correctly mapped to its CARC/RARC combination and denial category; and stress-testing prevention recommendations against the operational reality of front-end registration, authorization, and coding workflows.

Expect structured annotation more than prose writing. Tasks typically ask for a rating, a specific error label, and a short written rationale pointing to the exact sentence or code that fails. Some batches ask you to author a gold-standard appeal for a given denial scenario so the model has a reference. Others ask you to compare two model outputs and defend a preference. Vague feedback like "weak appeal" gets rejected; "cites LCD language that was retired in 2023 and omits the timely-filing date required for this Medicaid plan" is what gets used.

What the platform screens for

  • Verifiable operational depth. Mercor's AI interviewer will probe specifics: denial categories you personally worked, overturn rates you were accountable for, how you built an appeals calendar under varying payer deadlines.
  • Code-level fluency. Expect follow-ups on distinguishing CARC from RARC, common medical-necessity denial patterns, and where technical denials differ from clinical ones.
  • Evaluation judgment. Can you separate an appeal that is well-written from one that is likely to be overturned? These are different, and the screen tests whether you know it.
  • Written English under scrutiny. Your rationales are the product. Sloppy or generic writing disqualifies.

Logistics

Fully remote and asynchronous. Contributors commonly commit 10–20 hours per week, though some projects scale up during active labeling pushes. Work is contract-based via Mercor; pay is hourly at rates observed in the $70–93 range for this role, with the exact figure set at offer and not guaranteed. Certifications (CPC, CCS, CRCR, CHFP) and hands-on time with denial analytics platforms strengthen an application but don't substitute for management-level operating experience.