What the work actually looks like

You spend your hours reading and scoring AI output that imitates a technical seller. A typical batch might include a cold email to a plant engineer about a flow meter, a discovery-call transcript for a mass spectrometer purchase, a quote package with mismatched part numbers, or an objection-handling turn where the model concedes on price before it has established value. You judge each on two axes at once: is the technical claim about the product true and specification-accurate, and would this actually move a real buying committee. Much of the work is head-to-head comparison — two responses, same prompt, pick the better one and write the reason in a few tight sentences that another reviewer could act on.

  • Pairwise ranking with written rationale, usually the bulk of the queue
  • Single-response scoring against a rubric (accuracy, sales effectiveness, customer relevance)
  • Occasional rewrites showing what a competent rep would have sent instead
  • Flagging fabricated specs, invented certifications, and hallucinated lead times

What the platform screens for

Mercor's screen is AI-led and follow-up heavy. It wants a verifiable sales history — what you sold, to whom, at what ACV, how long the cycle ran, who signed — not a title. Expect it to press on product specifics: if you say you sold analytical instrumentation, be ready to name the competing platforms, the specification a lab manager cares about, and the objection you heard most. It also tests evaluation judgment, which is a different muscle from selling: can you separate "this email is stylistically smooth" from "this email states a false compatibility claim," and can you explain your preference without hiding behind taste. Written English is assessed continuously through your rationales, not as a separate step.

Logistics

Fully remote and asynchronous — you pull tasks from a queue on your own schedule, though throughput expectations are real and 20+ hours a week is a floor, not an aspiration. Duration is roughly 4–6 weeks with an immediate start. Payment is task-based at first: you're paid on approval of your first task, and completing it inside the stated window is what qualifies you for hourly pay across the rest of the project. The $70–110/hr band reflects observed rates on comparable Mercor domain projects and is not a guarantee for any individual contributor.