What the work involves
You'll work through a queue of presentation tasks that mirror what a pharma brand team actually produces. A prompt might ask for a launch readiness deck for a specialty product six months from approval, a quarterly market performance review with NBRx/TRx trend framing, a payer value story slide sequence, or a lifecycle strategy narrative for a brand facing LOE. Sometimes you author the reference deck yourself; more often you evaluate a model's attempt — scoring whether the storyline holds, whether the data slide would survive a leadership review, whether the claim on slide 7 is one a medical/legal/regulatory reviewer would ever pass — and then write the correction with reasoning a model can learn from.
The craft matters as much as the content. Slide architecture, action titles that carry the argument, chart selection, density, the difference between a deck built to be presented and one built to be read — these are the distinctions the lab is trying to teach. Generic "good presentation advice" is not useful output. What's useful is you saying that a competitive landscape slide built as a four-quadrant grid buries the access story, and showing the version that doesn't.
What the platform screens for
Ethos runs an AI voice screen before any task access. Expect it to press on specifics: which brands, which phase of lifecycle, whether you owned the brand plan or contributed to it, who your audience was when you presented. It will also test judgment — given a flawed AI-generated slide, can you name the failure precisely rather than calling it "unclear"? Candidates who speak in agency-deck abstractions tend to stall under the second and third follow-up. Candidates who can describe a specific pre-launch deck they built, the objection they were anticipating, and the slide that answered it tend to move forward.
Logistics
- Fully remote and fully asynchronous — no standing meetings, no fixed hours
- 5–20 hours per week, self-selected; higher volume is generally available
- Observed rate is $80/hour; pay bands on these programs vary by task type and are never guaranteed
- Work is done in a browser-based task environment with PowerPoint-equivalent deliverables; expect to have a working PowerPoint install
- Engagements run in waves — expect periods of heavy queue and quieter stretches