Use AI to Draft One Clarifying Question Before You Act

A polished draft can hide a decision nobody has made. When a missing answer changes what you will tell other people, use AI to help frame a question for the right person. Let that person make the decision.

For: Anyone drafting a public note, update, or small plan from a verified, non-sensitive source.

Use it to: Identify one consequential unknown, ask its named decision owner a neutral question, and keep the answer open until a human confirms it.

The fictional note

This puzzle-swap example, the mistaken flyer line, and the proposed question are constructed teaching examples. They are not a real event, a message sent to anyone, or an observed response from a separate AI run.

1. The neighborhood puzzle swap is planned for Saturday, November 7, 2026, 10 a.m. to noon. The library lobby reservation is confirmed.
2. Flyer copy is due Thursday, November 5, at noon.
3. Sign-up wording for the flyer is undecided: whether participants must sign up has not been settled. Maya is the person who will decide; no answer has been recorded.
4. Leo offered to bring boxes. The number of boxes is unknown.

The dates are part of the fictional note. A separate calendar check confirmed that November 5, 2026 is Thursday and November 7 is Saturday.

Suppose a draft flyer says, “No sign-up needed. Just drop in.” It sounds friendly, but line 3 does not support it. A later question such as “Maya, since people can drop in, can we print that?” repeats the unsupported assumption. Asking Leo about the box count may be useful for setup, but it will not settle the sign-up wording on the flyer.

Find the question that changes the next action

Before asking AI to write anything, identify four things in the source:

1. The open decision: Is sign-up required? The note does not say.
2. The person who can answer: Maya is named as the decision owner. AI and the flyer writer are not.
3. The action affected: The public flyer needs accurate sign-up wording.
4. The timing: Flyer copy is due Thursday, November 5, at noon.

That gives you a neutral question to review, not send automatically:

Maya, what should the November 7 puzzle-swap flyer say about sign-ups, given that the copy is due Thursday, November 5 at noon?

The question does not supply Maya's answer. It names the artifact and deadline without claiming that sign-up is required or optional. Line 3 supports the recipient and uncertainty; line 2 supports the deadline; line 1 supports the event date. If Maya has not answered, the sign-up line is still open.

A reusable AI prompt

Use only a harmless note or material you are authorized to share with the AI service:

Using only the numbered note below, identify the one unresolved fact or decision that would change the public-facing next step. Draft exactly one neutral question to the person explicitly named as its decision owner. Include the relevant artifact and deadline only when the note supplies them. Then list the exact note lines supporting the recipient, unknown, artifact, and timing. Do not answer the question, invent an owner, choose a policy, or send a message. If the owner is missing, write “owner unknown” instead of guessing. Label anything the note does not establish “cannot verify.”

NUMBERED NOTE: [paste a non-sensitive note you are authorized to use]

The prompt asks for a small output and a source-line check. It is a drafting aid, not a delegation of Maya's choice. OpenAI's prompt guidance recommends a clear task, necessary context, a desired form, and iteration. Its accuracy guidance warns that a confident answer can still be wrong, so important details need checking.

Check the question before a person sees it

Run four checks against the original note, not against the AI's explanation alone:

If Maya answers, record what she actually decided and when. Check the final flyer against that decision before sharing it. If no answer arrives, do not turn silence into permission for a drop-in claim. Leave the sign-up wording open and ask the responsible person whether the flyer should wait or omit that claim. The note does not decide that fallback for you.

The box count is a separate unknown. Ask Leo about it when setup planning needs the number. It should not distract from the one decision that changes the public sign-up line.

Limits and one action to try

A well-phrased question can still point to an old or incomplete note. Verify that a real note is current and approved for this use. Do not paste private messages, workplace records, or personal details into an AI service unless you are authorized to use it for those records. AI can help draft a question; only the responsible human can make the decision and confirm the current fact.

NIST's AI Risk Management Framework 1.0 discusses objective evidence and human oversight in the broader context of trustworthy AI. These sources support the checking approach. They do not evaluate this fictional puzzle swap, prove this particular prompt works, or measure any benefit.

Try it: Take one harmless, verified note with an unresolved decision. Ask AI for one neutral question and the source lines behind it. Check the person, unknown, artifact, and timing yourself before you decide whether to ask the human owner.

Sources and disclosure

• OpenAI, “How do I create a good prompt for an AI model?”, checked October 29, 2026. The page displayed “Updated: 9 days ago”; no absolute update date was shown.
• OpenAI, “Does ChatGPT tell the truth?”, checked October 29, 2026. The page displayed “Updated: 6 days ago”; no absolute update date was shown.
• NIST, AI Risk Management Framework 1.0, section 3, “AI Risks and Trustworthiness”, 2023 excerpt checked October 29, 2026. NIST states a revision is in progress.

Disclosure: AI assisted research and drafting. Short Briefing constructed and checked the fictional note, faulty wording, prompt, and example question. No separate model response, real puzzle-swap event, sent message, measured gain, or subscriber outcome is claimed. Any companion video uses synthetic narration and original project graphics.