The work
You receive a task prompt and the artifact an AI model produced against it: a draft standard operating procedure, a current-state/future-state process map, a control plan, a Kaizen or DMAIC readout deck, a takt-time or capacity spreadsheet. Your job is to judge whether the artifact would survive contact with a real operation. That means checking whether the SOP is actually executable by the person named in it, whether steps are sequenced correctly, whether measurement definitions and units hold up, whether formulas in the spreadsheet compute what the label claims, and whether the deck's recommendations follow from the data shown. You write structured feedback tied to rubric dimensions — not a general impression, but specific defects with locations and reasoning.
A large share of findings are unglamorous: an SOP with no revision control or approval line, a swimlane diagram where handoffs vanish, a control chart with the wrong subgroup logic, a slide whose title asserts a conclusion the chart does not support. Aesthetic and presentation quality counts too — formatting, alignment, readability, consistent terminology — because these artifacts are graded as deliverables, not drafts.
What the screen looks for
- Five or more years of hands-on process improvement or SOP authorship work, described concretely: which processes, which environment (manufacturing, clinical, shared services, logistics), what changed and how you knew.
- Real fluency with Google Slides and PowerPoint, plus Sheets/Excel — you will be asked about it because so much of the output is slide-based.
- Whether you can hold a position under follow-up: state a judgment, then defend or revise it when the interviewer pushes.
- Calibration instincts — knowing when an output is wrong versus merely different from how you would have done it.
An advanced degree helps but does not substitute for operational experience. Regulatory context (ISO 9001, GMP, FDA QSR) or Lean Six Sigma certification is a plus where it reflects actual project work.
Logistics
Remote, asynchronous, hourly, contractor. You claim tasks from a queue and work in blocks that suit you; most contributors do 10–20 hours a week, and volume fluctuates with project demand. Pay is per hour logged on reviewed work. The $80–120/hr range reflects rates observed on Mercor for this specialization and is not a guarantee — the offered rate depends on the project and your assessed depth.