What the work actually involves

You are handed AI-generated outputs on finance topics — a tax treatment question, a DCF walk-through, a retirement distribution scenario, a valuation memo — and asked to decide whether the reasoning holds. That means two distinct jobs. First, quantitative verification: independently recomputing a stated figure rather than eyeballing whether it looks plausible. Second, rule verification: tracing a claim back to a primary source — the Internal Revenue Code, a Treasury regulation, an SEC rule, IFRS or ASC guidance — rather than trusting that the model's confident citation is real. A substantial fraction of the errors you will catch are hallucinated citations, correct rules applied to the wrong entity type or filing status, and arithmetic that compounds a small early mistake into a wildly wrong final number.

Feedback is structured. You are not writing a paragraph of impressions; you are flagging the specific step where reasoning broke, stating the correct treatment, and citing what makes it correct. Some tasks ask you to rank two model outputs; others ask you to rewrite one into the version an expert would have produced. Researchers may follow up on your annotations, so the reasoning trail matters as much as the verdict.

What the platform screens for

  • A verifiable credential in good standing — CPA, EA, CFP, or CFA. Mercor's screen asks about credential status and jurisdiction, and this is checked.
  • Substantive practice, not adjacent exposure. Two-plus years actually doing tax, planning, FP&A, banking, PE, or advisory work.
  • Recomputation instinct. Screens frequently present a figure and ask how you would check it. Answers that describe pulling the underlying inputs and rebuilding the calculation land better than answers that describe sanity-checking magnitude.
  • Source discipline under unfamiliar conditions. Expect a question about a rule outside your daily practice; the point is whether you know where the authority lives and say so honestly, not whether you already know the answer.

Logistics

Fully remote and asynchronous — there are no standing meetings and no set hours. Work arrives in batches and you accept what you have capacity for. Observed pay for this project runs $100–200/hour for U.S.-based professionals, set by credential and assessed depth rather than negotiated; pay bands are as reported and not guaranteed. Payment is weekly via Stripe Connect under an independent contractor agreement, which means you handle your own tax withholding. Most contributors treat this as a nights-and-weekends supplement to a primary practice, and volume fluctuates with project phase.