What the work actually involves

You write audit problems that a strong senior or manager would find non-trivial, then judge how a model handles them. A task might be a set of facts about a client's revenue recognition cut-off, a walkthrough of a three-way match control, or a sampling plan for a substantive test of details — and your job is to produce the scenario, the defensible answer, and the reasoning trail that explains why competing answers are wrong. The second half of the work is evaluation: reading model output, deciding whether it applied the right standard rather than merely naming it, and writing feedback specific enough that an engineer who has never audited anything can see where the reasoning broke.

Expect topics across audit planning and materiality, risk assessment and significant risks, tests of controls versus substantive procedures, evidence sufficiency and appropriateness, sampling, going concern, related-party and fraud considerations, and documentation standards. Some batches are US-centric (AICPA/PCAOB); others may call for ISA framing. Being explicit about which framework you are applying matters more than which one you know best.

What the platform screens for

  • Standards fluency under follow-up. A screener will push past a first answer — you should be able to say why a procedure addresses a specific assertion, not just that it is a common procedure.
  • Verifiable audit history. Firm type, client sizes, industries, the roles you held on engagements, whether you signed or reviewed workpapers.
  • Written precision. The output is text. Ambiguous fact patterns and hedged rationales get rejected downstream.
  • Calibration. Willingness to grade a fluent, confident, wrong answer as wrong, and to explain the failure concretely.

Logistics

Fully remote and asynchronous — no standing meetings, work drawn from a task queue. Contributors commonly report 10–20 hours a week, with volume varying by project. Pay is hourly and contractor-based; the $75–150 band is what has been observed on this category rather than a guaranteed rate, and placement within it tracks credentials (CPA/CIA), depth of experience, and whether you are authoring or reviewing.