What the work actually is
You are not running a billing department here. You are the person who reads what a model produced — a populated CMS-1500 or 837P segment, a suggested modifier, a claim-edit resolution, an appeal rationale — and decides whether it would survive a clearinghouse scrub and a payer adjudication. Most tasks arrive as a case with source documentation and one or more AI outputs. You score them, mark the specific defect (wrong place-of-service, missing NDC, modifier 25 without a separately identifiable service, taxonomy mismatch on the 837), and write a short correction that explains the payer rule behind it. Some batches ask you to author gold-standard responses or adversarial cases the model tends to fail.
What the screen looks for
- Verifiable operational history. Volume you managed, payer mix, professional vs. facility, what your first-pass acceptance rate was and how you moved it.
- Transaction-level depth. Loop and segment familiarity on the 837, how rejections differ from denials, what a clearinghouse edit catches versus what the payer catches downstream.
- Payer specificity. Generic answers about "following CMS guidelines" screen out fast. Naming a MAC LCD, a commercial payer's timely filing quirk, or a Medicaid state plan difference does not.
- Evaluation judgment. Can you distinguish a genuinely wrong output from one that is merely different from how you would do it, and can you explain the difference in writing that a non-billing annotator could follow?
Logistics
Fully remote, asynchronous, contract. Work is delivered in batches with deadlines rather than shifts; most contributors commit 10–20 hours per week, though some projects scale up for short periods. Pay is reported around $80/hr at this level, varying by project and assessed depth — treat it as observed rather than promised. Expect an onboarding calibration round where your annotations are compared against a rubric before full volume opens up.