The work

You receive an artifact produced by a model — a supplier evaluation matrix, a category sourcing strategy deck, a vendor risk assessment memo, a should-cost model in a spreadsheet, or a draft SOW — plus a rubric and the prompt the model was working from. Your job is to decide whether a working procurement professional would accept it. That means checking whether the total cost of ownership logic holds, whether the weighted scoring actually sums the way it claims, whether contract terms referenced are plausible and internally consistent, whether the deck's narrative would survive a category review with a CFO in the room, and whether formatting and chart choices are defensible rather than decorative.

Feedback is the deliverable. Vague scores are not useful to the people training these systems — you need to name the specific error, explain why it matters commercially, and say what a correct version would have looked like. Expect to write several structured comments per task, and expect your reasoning to be spot-checked against other evaluators working the same rubric.

What the screen looks for

  • Verifiable depth. Five-plus years in sourcing, category management, vendor governance, or contract administration — and the ability to talk concretely about spend under management, categories owned, and negotiation outcomes.
  • Rubric discipline. Whether you can hold a consistent standard rather than drifting toward your own house style or personal preferences.
  • Error taxonomy. Whether you distinguish a factual error from a stylistic weakness from a missing-but-not-requested element — over-penalizing is as much a calibration failure as under-penalizing.
  • Artifact fluency. Real facility with Slides and PowerPoint, Excel and Sheets. A meaningful share of tasks are deck- and spreadsheet-based, and aesthetic and layout judgment is scored.

Logistics

Fully remote and asynchronous — tasks are pulled from a queue, no standing meetings. Most contributors work in blocks of a few hours and treat this as supplemental to existing consulting or in-house work. Volume fluctuates with project cycles, so treat availability as variable rather than a guaranteed weekly number. Pay is hourly and stated as observed on the platform, not guaranteed; onboarding typically includes a paid or unpaid calibration set before full task access.