What the work actually involves

You write accounting problems that look like your real workload, then you check whether a model solves them the way a competent practitioner would. A typical unit of work is a scenario — a month-end close with an unreconciled intercompany balance, a lease classification under ASC 842, a partial-year depreciation schedule, an S-corp basis calculation, a sampling plan for a revenue test — paired with a defensible answer and the reasoning path behind it. Other tasks are comparative: two model responses to the same accounting prompt, and you rank them and explain in writing why one is stronger, flagging where a plausible-sounding answer cites the wrong standard, uses the wrong framework, or skips a judgment call an accountant would never skip.

The written feedback is the deliverable that matters most. "Incorrect" is nearly useless to a research team; "the model applied the lower of cost or market rule to inventory measured under FIFO rather than lower of cost and net realizable value, which is the correct treatment under ASC 330 for non-LIFO inventory" is what actually changes behavior. Expect to cite standards by reference, distinguish GAAP from IFRS treatment where it diverges, and be explicit when an answer depends on a facts-and-circumstances judgment rather than a rule.

What the screening looks for

  • A verifiable credential — CPA, CA, ACCA, CMA, EA, or a recognized equivalent — and three or more years of real practice in audit, tax, reporting, bookkeeping, controllership, forensic work, or accounting systems.
  • Depth in a named specialty rather than broad familiarity. Screeners follow up on the specific area you claim, and thin answers surface quickly.
  • Ability to write about accounting reasoning, not just perform it. Standards citation, error diagnosis, and clean prose under an AI-led interview.

Logistics

Fully remote and asynchronous; there are no client calls and no fixed shifts. Contributors commonly work 10–20 hours a week alongside a practice or an in-house role, drawing tasks from a queue. The path in is a resume or experience summary, a short form on practice area and certifications, and often a brief sample task. Strong contributors are moved into reviewer, lead, and domain-expert positions. The $80/hr figure is an observed rate for this funnel, not a guaranteed offer — actual pay varies by specialty, task type, and calibration results.