What the work looks like

You will be handed AI-generated commercial artifacts — a discovery call transcript, a quote against a customer RFQ, a product substitution recommendation, an email handling a short-ship or a damaged pallet — and asked to judge whether a working sales rep would actually send it. Much of the volume is head-to-head: two model responses to the same buyer scenario, and you pick the stronger one and write why. The 'why' is the deliverable. Reviewers who write 'Response B is better, it's more professional' are not producing usable training signal; reviewers who write 'B correctly flags that the requested quantity falls below the case pack and offers the next break, A quotes an eaches price that doesn't exist' are.

Expect scenarios that hinge on things AI models get quietly wrong: margin math on tiered pricing, freight terms and who owns the damage, MOQs and lead times, distributor versus direct channel conflict, an overpromised delivery date, discounting authority a rep doesn't have. You are also flagging soft failures — the model that caves on price at the first objection, the model that invents a product spec, the model that answers a purchasing manager the way you'd answer a consumer.

What the screen actually checks

Mercor's screening is AI-led and follow-up heavy. It will ask what you sold, to whom, at what order size, and through what channel, then push on the specifics: how you handled a price objection from a distributor, what your discount ceiling was and who approved anything beyond it, how you priced a mixed truckload. Vague seniority claims do not survive the second question. The bar is 4+ years in a wholesale or manufacturing B2B sales seat — account rep, inside sales, outside sales, sales consultant — selling non-technical goods with real product knowledge behind you. Written English matters more than usual here, because your rationales are the product.

Logistics

  • Fully remote and asynchronous; no set shifts, but sustained throughput matters
  • 20+ hours per week is a stated requirement, not a suggestion
  • Immediate start, approximately 4–6 weeks duration
  • Payment is task-based to begin with, released on first task approval; completing that first task inside the required window qualifies you for hourly pay across the rest of the project
  • Observed band is $70–110/hr; rates vary by project and are not guaranteed