What the work actually involves

You build the artifacts a consulting engagement would produce, then reverse-engineer the prompt that should elicit them. A typical unit of work is one deliverable — a two-page client memo on a make-versus-buy decision, a market sizing spreadsheet with visible assumptions, a five-slide recommendation deck — paired with a task description precise enough that a model has a fair shot, and a rubric or reference answer specific enough that another reviewer could grade consistently. The second half of the role is evaluation: reading model output against your reference, deciding whether an answer is wrong or merely structured differently, and writing feedback that names the analytical failure rather than the stylistic one.

The hard part is not writing consulting content. It is being explicit about things consultants normally leave tacit — why the market was segmented that way, which assumption the recommendation actually hinges on, what would make the answer defensible in front of a client versus merely tidy. Models are often fluent and confidently unsupported; catching that requires someone who has been asked "where does that number come from?" by a partner.

What the platform screens for

  • Real engagement experience, described concretely — client type, your role in the workstream, what you personally produced.
  • Excel and PowerPoint fluency at working depth: build-up models, sensitivity, clean exhibit logic — not template decoration.
  • Judgment about grading: can you separate a genuine analytical error from a valid alternative approach, and justify the call in writing.
  • Native-level English writing, since deliverables and feedback both become training signal.

Logistics

Fully remote and asynchronous, with no fixed hours. Contributors typically pick up batches of tasks and complete them on their own schedule; volume fluctuates with project demand, so this suits people with 5–15 discretionary hours a week rather than those needing steady full-time load. Pay is per accepted work at rates observed in the $100–150/hr range — not guaranteed, and dependent on task type and quality review. Expect a calibration or trial task before sustained volume.