Short Briefing · Evidence current through 2026-09-20
How do I define a useful result before prompting?
The bike shop opens at nine. The new supervisor receives a polished handover saying a collection job is ready and a delivery arrives at eight thirty. Neither claim appears in the notes. This is a fictional example, with mock text we wrote. Before asking for a better draft, define what a useful result must preserve: the right audience, the right deliverable, and checks that expose unsupported additions.
- For
- general education, not made specifically for children
- Use it to
- Write audience, deliverable and acceptance criteria first.
Watch the briefing
Briefing
The bike shop opens at nine. The new supervisor receives a polished handover saying a collection job is ready and a delivery arrives at eight thirty. Neither claim appears in the notes. This is a fictional example, with mock text we wrote. Before asking for a better draft, define what a useful result must preserve: the right audience, the right deliverable, and checks that expose unsupported additions.
Start with the audience. This handover is for the opening shift supervisor, who did not work the previous evening. That person needs to know what is ready, what remains pending, and what to do first. A public announcement would need different information. So would a mechanic's technical report. Naming the reader helps us choose what belongs in this document without pretending that one draft can serve every purpose.
Next, define the deliverable. We want one page, no more than one hundred eighty words, with three headings: ready, pending, and first actions. The word limit is a drafting constraint; actual page fit depends on layout and must be checked later. This is an opening handover, not an instruction to contact customers, alter job records, or approve equipment. A clear format makes the result easier to inspect.
Now set acceptance criteria from six invented notes. The shop opens at nine; the front desk is tidied. Collection job B seventeen awaits customer contact, with no promise that it is ready. Tube delivery timing is unconfirmed. The supervisor checks the till using the shop's existing procedure. No mechanic inspection or equipment-safety approval is recorded. A useful draft must retain those facts and uncertainties without adding an owner, time, or completed action.
Put those decisions into the request: draft a handover for the opening supervisor, under one hundred eighty words, using our three headings and the supplied notes. Preserve uncertainty and do not invent confirmations. Then review the illustrative draft against the criteria. The earlier ready-for-collection claim fails because the notes do not establish readiness. The eight-thirty delivery fails because the timing is unknown. Smooth wording cannot make either addition acceptable.
Our corrected mock puts the tidied desk under ready. Under pending, it keeps customer contact and delivery timing unresolved, and notes that no inspection or safety approval is recorded. First actions retain the supervisor's till check under the existing procedure and the nine o'clock opening. We still check the word count, every source fact, and the final layout. This is a planning handover, not a replacement for competent inspection or approved operating rules.
Write audience, deliverable and acceptance criteria first. That gives you a reason to accept, correct, or reject the draft before polishing its tone. These checks are our editorial adaptation of evaluation principles, not a guarantee that a tool will comply. Keep final judgment with the person responsible for the work. Next, we will break a task that is too large into one inspectable deliverable at a time.
One insight you can use
Write audience, deliverable and acceptance criteria first.
What remains uncertain
The approved narration states the applicable limits; teaching examples are not measured outcomes or individualized recommendations.
Disclosures
- AI-assisted production and synthetic narration. Original teaching examples and diagrams; linked third-party sources retain their respective rights.
Corrections
- No corrections have been recorded.
Original sources and limits
See what supports the briefing
-
https://developers.openai.com/api/docs/guides/evaluation-best-practices
developers.openai.com · 2026-09-17
Define success criteria, evaluate representative cases and incorporate human feedback; developer guidance adapted editorially to human checklists.
-
https://developers.openai.com/api/docs/guides/prompt-engineering
developers.openai.com · 2026-09-17
Explicit instructions/context and non-deterministic outputs; first-party developer guidance, not evidence for our fictional workflow's effectiveness.
-
https://www.nist.gov/itl/ai-risk-management-framework
www.nist.gov · 2026-09-17
Voluntary risk-management framework; original small-task adaptations are not NIST endorsement or certification.