This guide gives you a reusable starting pattern. It is designed to help you see the work more clearly; it is not a substitute for judgment, source checking, or responsibility for the result.
Prepare the inputs before you ask for output.
The model only sees what you give it. Spend a few minutes naming the reader, desired result, and uncertain information. This makes a first draft easier to assess and reduces the need for decorative rewriting later.
Give the task a useful brief.
You are helping me prepare a one-page research brief. Goal: [state the decision or question] Audience: [who will read it] Source notes: [paste labeled notes and links] First, group the notes by claim. Preserve the source label beside each claim. Then produce: (1) the question, (2) three evidence-backed findings, (3) open questions, and (4) a short recommendation only if the notes support one. Flag claims that lack a source, are dated, or need verification. Do not invent facts or citations.
Replace every bracketed field with your real context. Read the output before reuse.
Four steps that keep the result usable.
- 1
Name the decision before collecting anything.
- 2
Label each pasted note with its source and date.
- 3
Ask AI to organize claims and expose gaps.
- 4
Open original sources for decision-critical claims.
Why a brief beats a summary
A summary compresses material. A useful brief helps someone make a next decision. It needs a concrete question, the evidence that matters, what is still unknown, and the confidence behind any recommendation.
Give the model source labels it can preserve
Instead of pasting a wall of text, add short labels such as [Interview A, 12 Aug]. The output can keep claims tied to the source that supports them, making a later human check much faster.
Use the output as a review surface
The useful moment is not when the first draft appears. It is when the draft lets you notice what is missing: an old number, a weak comparison, or a conclusion that reaches farther than the evidence.
A worked example: sizing a service decision
Suppose the question is whether a support team should move to a new ticketing tool. Labels such as [Vendor demo, 09 Aug], [IT security review, 11 Aug], and [Support lead interview, 12 Aug] keep every claim traceable. The draft brief can then group cost, migration, and security claims separately, so the person deciding sees which parts rest on a live demo and which rest on a price sheet from May.
What to do when the brief looks confident
Confidence in a brief is not the same as certainty in the sources. If the draft states a number without a label, ask for the label before treating it as fact. If two notes conflict, keep both visible instead of letting the model pick one. A brief that still has visible open questions is working correctly; a brief that reads smoothly may have hidden the seams.
Run a human check.
- Can each factual claim be traced to a note or source link?
- Does the brief separate observations from a recommendation?
- Are dates and limitations visible where they matter?
- Would a reader know what to verify next?
AI is strongest here when it makes missing information, structure, and options easier to see. The moment an output becomes a claim, commitment, or decision, bring a person back into the loop.