What the work involves

You'll spend most of your time on two kinds of tasks. The first is authoring: given a brief — a CEO's earnings narrative, a product keynote at a developer conference, a thought leadership talk for an industry summit — you produce the model-facing gold standard, both the spoken script and the deck that carries it. The second is evaluation: you're shown two or more AI-generated attempts at the same brief and asked to rank them, then justify the ranking in writing that a researcher who has never written a keynote can act on. Expect specific recurring workflows: speech-to-slide alignment (does the deck support the spoken line or compete with it), CEO narrative development, conference and keynote decks, earnings call supporting materials.

The critique is the product. "Slide 6 is weak" is worthless; "slide 6 puts four claims on one chart while the script only lands one, so the audience reads ahead and stops listening — cut to the single revenue line and move the other three to appendix" is the deliverable. Most reviewers who wash out do so because they can produce good decks but can't explain, in transferable terms, why one draft beats another.

What the platform screens for

Ethos runs an AI voice screen before any human contact. It probes whether your four-plus years were genuinely spent writing for senior leaders or adjacent to people who did — expect follow-ups on specific venues, specific formats, how you handled a principal who rewrote your draft, and what you do when the strategy team's slides contradict the speech. Portfolio claims get pressure-tested: naming the event and your actual role matters more than adjectives. PowerPoint depth is checked concretely too — master slides, chart construction, rebuilding a deck under time pressure.

Logistics

  • Fully remote, asynchronous; no fixed hours and no standing meetings.
  • 5–20 hours per week, with more available to reviewers whose output holds up.
  • $70/hour is the observed rate on this listing, not a guarantee; rates vary by task type and calibration performance.
  • Work is claimed from a queue, so consistent weekly availability matters more than large irregular bursts.