What the work actually involves
You are not using the AI to do office work — you are deciding whether its output would survive contact with a real workplace. Projects fall into three broad shapes. The first is authoring tasks: taking a document, financial model, or deck from your own professional life, stripping it to a realistic starting state plus a constraint, and writing the reference answer a model attempt gets scored against. The second is grading: opening model-generated files and judging whether the formula is correct and robust to inserted rows, whether the formatting holds after someone else edits it, whether the deck's argument actually lands. The third is failure-hunting — flagging outputs that look fine on screen but break under file conversion, a different Office version, shared editing, or a locale that uses commas as decimal separators.
What the platform screens for
Mercor's 20-minute AI interview probes depth in one tool rather than breadth across all of them. Expect follow-ups on specifics: which functions you reach for and why, how you'd audit someone else's model, what you know about the seams between Excel and Sheets or between Word and PDF. Written reasoning matters as much as tool skill — most graded work is a judgment plus a paragraph explaining it, and vague justifications are the most common reason submitted work gets rejected downstream. Comfort with ambiguity is explicitly named: projects arrive with imperfect instructions and the expectation is that you flag them, not quietly guess.
Logistics and honest expectations
- Fully remote and asynchronous; no fixed hours, and project work is typically taken in blocks rather than daily shifts.
- Rates on listings in this field have been observed at $60–70/hr, set per project by scope and depth. This listing itself names no rate — the project invitation does.
- This is a standing pool, not an opening. You will not receive an accept or reject decision. Matching can happen within a week or take several months, depending on what clients are running.
- You apply once, complete the domain expert interview, and remain eligible across every network you qualify for.