The work
Ethos is staffing this for a foundational AI lab that wants its model to produce consulting artifacts a partner would actually put in front of a client. You'll spend your time on three kinds of tasks: authoring reference deliverables from a brief (a market entry deck, a sensitivity-tested case model, an operating model redesign summary), critiquing model-generated versions of the same, and rewriting weak outputs into something defensible. Expect real files — .xlsx with working formula chains, .pptx with a horizontal storyline — not chat transcripts.
The critique work is where most of the value sits. You'll be asked to say precisely what is wrong with an AI-drafted deck: that the executive summary asserts a conclusion the appendix doesn't support, that the market sizing double-counts a segment, that the driver tree in the model is hardcoded where it should be an input. Written rationales are read by researchers and other reviewers, so vague verdicts like "not client-ready" are near-useless.
What the screen looks for
- Ownership, not proximity — whether you built and defended the deliverable or supported someone who did
- Specific engagement types you've run end to end: diagnostics, market entry, operating model, post-merger, cost transformation
- Modelling depth, including how you structure sensitivities and what you refuse to hardcode
- Whether your criticism is diagnostic (what's wrong, why, what would fix it) or merely evaluative
An AI voice screen is part of the process and will follow up on the specifics you give, so bring engagements you can discuss in detail without breaching confidentiality.
Logistics
Fully remote and fully asynchronous, 5–20 hours per week with room for more if you want it. Tasks are claimed from a queue rather than assigned on a calendar, so evening and weekend work is normal. $100/hour is the rate observed for this listing; pay on these programs varies by task type and reviewer tier and is not guaranteed.