What the work involves

You receive artifacts an AI model produced in response to a finance prompt: an operating model in Excel, a monthly variance commentary memo, a budget-vs-actual deck, a headcount plan, a scenario sensitivity table. Your job is to grade each one against a rubric and write feedback specific enough that an engineer who has never built a forecast can see exactly where the output failed. That means catching broken formula logic and circular references, but also the subtler failures reviewers miss: a bridge that doesn't tie, driver assumptions that contradict the stated scenario, revenue recognition treated casually, a cost line double-counted across departments, or commentary that describes a variance without explaining its cause.

Much of the work is spreadsheet and slide review, so tooling fluency matters as much as finance knowledge. Expect to open workbooks and trace precedents, judge whether a chart type suits the data, and flag decks that would not survive a CFO's read — inconsistent units, unlabeled axes, footnotes that contradict the body. Tasks arrive in batches; volume fluctuates.

What the platform screens for

Mercor's screening is AI-led and follow-up heavy. It probes for verifiable specifics — what you actually owned, at what scale, on what cadence — rather than titles. Expect to be pushed on mechanics: how you built a rolling forecast, how you handled a mid-year reforecast, what your close-to-commentary timeline looked like. The second thing measured is evaluation judgment: given a flawed output, can you rank errors by severity and justify the ranking, rather than listing every nit at equal weight? Vague answers and inflated scope tend to end screens early.

Logistics

  • Fully remote, asynchronous, contractor engagement billed hourly.
  • Observed band is $80–120/hr; actual rate depends on assessed depth and project, and is not guaranteed.
  • No fixed hours, but most reviewers commit 10–20 hrs/week to stay in active rotation.
  • Work is English-language throughout; written feedback quality is scored.