Vague Friday Requests Aren’t Briefs — Clarify First

The problem

“Can you look into this and send me something by Friday?” sounds like an assignment. It isn’t a brief.

It leaves the real questions unanswered: What decision should this support? Who is the audience? Is “something” a slide, a one-pager, a recommendation, or talking points? What does “done” mean by Friday?

The tempting move is to paste the request into your AI chat tool and ask for the deliverable. That’s backwards. If the job is unclear, the model will still produce something that looks finished — confident tone, neat structure, plausible detail. You then spend Friday polishing invented requirements, or discover in the meeting that you built the wrong thing.

Sometimes the useful first application of AI is not producing an answer. It is helping you ask a better question.

Try it safely

Use a synthetic or authorised example. Here’s one that mirrors real workplace vagueness without needing your company’s data:

“Can you look into whether we should expand the partner program and put something together for next week’s management meeting?”

Notice what’s missing: what “expand” means, what evidence exists, recommendation vs decision frame, deck vs one-pager, and which decisions the meeting should leave with.

A strong first move is not to invent a strategy pack. It is to propose a short management discussion brief — a one–two pager or discussion-point outline — covering outcomes to achieve, decisions required, next steps, and open questions / what else to discuss — before anyone writes recommendations.

Method

Paste this into your usual AI chat (ChatGPT, Claude, or whatever your workplace allows). Swap in your own (safe) request at the end.

Help me turn the request below into a draft work brief.
Use only what the request states as confirmed information.
Do not silently fill in missing details.

Produce:
1. What is explicitly known
2. Important unknowns or assumptions
3. Up to five clarification questions (most important first)
4. A proposed deliverable — clearly labelled as a proposal
5. A draft definition of done — subject to confirmation

Do not carry out the underlying assignment yet.
Do not invent analysis, options, or recommendations.

Request:
“Can you look into whether we should expand the partner program and put something together for next week’s management meeting?”

What a good output looks like (and what to correct)

A useful draft brief will look roughly like this — your model’s wording will differ:

1. Known: Someone wants input on whether to expand the partner program, for next week’s management meeting.
2. Unknowns / assumptions: What “expand” covers; existing evidence; meeting format; recommendation vs framed decision; ownership.
3. Clarification questions: e.g. What decision should management leave with? Discussion frame or recommend-and-approve? One–two pager or short deck of discussion points? Which constraints are already fixed?
4. Proposed deliverable (proposal): A one–two pager / discussion-point outline for management covering: outcomes to achieve; decisions required; next steps (as proposals only); open questions / discussion agenda. Not a full strategy deck and not invented analysis or recommendations.
5. Draft definition of done: Requester-confirmed discussion brief ready for the management meeting by [date] — subject to confirmation.

Corrections you’ll often need:

• If AI invents market data, partner names, or “recommended expansion to APAC” — strike it. That’s an assumption wearing a lab coat.
• If AI jumps into writing the strategy or scoring options — stop and re-run with “Do not carry out the underlying assignment yet.”
• If the proposed deliverable sounds like a final recommendation pack — rewrite it as a discussion brief proposal: outcomes, decisions needed, next steps, agenda items.
• If “definition of done” sounds like a guarantee — rewrite it as subject to confirmation.

Check before you use it

• Separated requested facts from AI-assumed facts
• Clarification questions would actually change the work if answered differently
• Proposed deliverable is labelled as a proposal (discussion brief / outline), not a commitment or finished strategy
• You will confirm the brief with a human before producing the real deliverable

Essential distinction: A proposed one–two pager of discussion points is not an agreed requirement. A suspected expansion option is not a finding.

Make it yours / when not to use this

Adapt: Swap the sample for a real request from your week — after stripping anything sensitive (see Before you paste). Change “management meeting” to your channel (steering pack, leadership sync, Slack update, email).

Don’t use this method when: The requester already gave a clear brief and you’re stalling; the work is high-stakes and you need a human scoping conversation first (AI can still prep questions); or you’re about to paste confidential context into an unapproved tool.

Your challenge (about 5 minutes)

Run the prompt on one fictional or safe vague request aimed at a leadership or management discussion. Then answer:

1. One assumption the AI introduced.
2. One clarification question that would materially change the work — especially what decisions the meeting must leave with.

Did this produce something you could use, and what did you have to correct?

Next: Sounds Good ≠ Correct — One Verification Habit. New here? Start Here · AI Newbie Brief

One Useful Task — from AI Newbie. Understand AI. Build real things.