# Turn a research brief into a small next step

Research becomes useful when it changes what you do next. Identify the uncertainty that could reverse your decision, then choose a small action that produces evidence about it. A long report is not proof that the plan will work.

## Separate facts from the decision hypothesis
For a fictional scheduling product, research may establish that competitors offer timezone labels. It does not establish that your intended customers will switch tools because of them. The decision hypothesis might be: “Teams arranging cross-region calls struggle to confirm the selected timezone.” That needs observation or conversation with those teams.

## Define the test before collecting feedback
| Element | Fictional pilot |
| --- | --- |
| Question | Can participants identify the timezone before booking? |
| Task | Book a supplied appointment in the prototype |
| Observation | Where they look and whether they verify the zone |
| Decision | Revise the label if it is repeatedly missed |
| Limit | Small pilot; no conversion estimate |

Do not change the success criterion after seeing the result just to call the test successful. If you change the question, record that as a new test.

## Ask AI for a bounded next step
```text
Read this brief and list assumptions that could change the decision.
Choose one assumption to test cheaply and reversibly.
Propose a task, observations, decision rule, and limits.
Do not invent participants, results, demand, or statistical confidence.
Brief: [paste]
Resources and constraints: [facts]
```

## Close the loop
After the action, record what happened in plain language. Keep observations separate from your explanation of them. “Two participants missed the label” is an observation; “the label needs more contrast” is a proposed explanation that can be tested.

Update the brief with the new evidence and the next decision. If the test did not answer the question, say why and adjust it. The purpose is to reduce important uncertainty, not to accumulate a sequence of reports that all recommend continuing. A useful outcome can be changing the plan, narrowing the audience, or deciding not to build a feature yet.

---
SkillStall · 2026-10-02