Turn the topic into a deliverable
Compare two fictional requests: > “Help with our evidence.”
“Review these ten public pages for whether their health claims match the cited sources, then provide a table of unsupported claims and suggested revisions.”
The second request identifies the material, review question, and output. It still needs boundaries, such as whether searching for replacement evidence is included.
Four decisions to record
| Decision | Example question |
|---|---|
| Depth | Is this a targeted source check or a systematic review? |
| Inputs | Who provides the documents and resolves missing information? |
| Review | Who can accept the work and how many revision rounds are included? |
| Use | Is the output for internal planning or public-facing claims? |
CDC's evaluation framework emphasizes aligning questions and design with intended use. That principle is useful here: a rapid internal scan should not later be described as a comprehensive evidence review.
A boundary is not a refusal to help
New questions often emerge during good work. A change can be recorded with its effect on time, cost, or the deliverable. This gives both parties a chance to choose rather than allowing an implied obligation to grow unnoticed.
For example, discovering that several sources are unavailable might lead to a narrower review or a separate retrieval phase. It should not silently turn uncertain claims into approved claims.
Define completion in observable terms
“High-quality advice” is hard to verify. “A claim-by-claim table with source links, limitations, and a review meeting” is inspectable. Contract terms, legal obligations, and professional scope require their own appropriate review; this article offers a project-planning framework rather than legal advice. A clear scope makes the scientific work easier to evaluate on its actual terms.
