New: Boardroom MCP Engine!

Ready to put this into action?

Get the complete AI Integration Playbook β€” Practical AI implementation guide β€” prompt engineering, workflow automation, and ROI frameworks.

Article 014 Β· Part 2

Make AI Ask the Missing Questions

Find the unknowns that could change the plan before polishing the plan.

By Randy Salars Β· Published

On this page
  1. Ask what could make the plan fail
  2. Request a small number of useful questions
  3. Give answers in a form the plan can use
  4. Separate assumptions from decisions
  5. Catch contradictions before they become promises
  6. Stop asking when there is enough to draft
  7. Students: clarify the assignment before solving the wrong problem
  8. Your exercise

Find the unknowns that could change the plan before polishing the plan.

The birthday party plan is beautiful. It has a cooking station, warm food, a cake presentation, and a minute-by-minute kitchen schedule.

There is one problem: the rented room has no kitchen.

The assistant did not know that. The person asking for help did not think to mention it. Together, they built a detailed answer on an unstated assumption.

An early question could have saved the work. But asking questions is not automatically useful either. A twenty-question interview about napkin colors is a poor start when the real uncertainty is whether the room can hold the guests.

The skill is to identify the few unknowns that would materially change what happens next.

Ask what could make the plan fail

Consider this fictional request:

Help me plan a two-hour birthday gathering for 24 people in a rented community room. I want something simple, with food and one group activity.

Before choosing decorations, the assistant needs to know whether the room permits the intended use, what facilities are available, and how much can be spent. Attendee needs may also change the options.

Not every unknown carries equal weight. Compare these:

Missing informationWhy it matters now
Room facilities and restrictionsCould rule out cooking, equipment, or activities
Remaining budgetDetermines which options are feasible
Attendee access and participation needsMay change layout, timing, and activity choices
Preferred napkin colorUsually can wait until a workable plan exists

The test is practical: β€œIf the answer changed, would I choose a different plan?”

If yes, the question probably deserves attention. If no, a modest, labeled assumption may be enough.

Request a small number of useful questions

A better opening request would be:

Before drafting, ask up to three questions about missing information that would most affect feasibility, cost, or participation. Explain briefly why each matters. Use labeled provisional assumptions for minor preferences.

For this scenario, useful questions are:

  1. What facilities and restrictions does the room have, including access, seating, and food preparation?
  2. What budget remains after the room booking, and what must it cover?
  3. What attendee needs should shape food and activities, without sharing private medical details?

These questions group related information. They should still be manageable to answer. If one question contains twelve hidden questions, shorten it.

β€œUp to three” is a starting limit, not a reason to ignore a fourth critical issue. The assistant can flag an additional blocker and explain its effect. You can decide whether to investigate it before proceeding.

Give answers in a form the plan can use

Suppose the organizer provides the following fictional answers:

  • The room booking covers the gathering plus 30 minutes of setup and 30 minutes of cleanup.
  • There is a level entrance, 24 seats, and six tables. The organizer has confirmed these details with the venue.
  • The room has no kitchen or refrigeration. Open flames are prohibited.
  • The remaining budget is $180 for refreshments, paper goods, and activity materials.
  • The group includes adults and children. The organizer wants an activity people can join while seated and has not yet collected food preferences.

The assistant can now propose a useful draft: a seated game, a simple room layout, and refreshment options to check once food needs and suitable handling arrangements are known.

It should not quietly add a warming tray, assume an accessible restroom from the level entrance, or invent dietary information. Each of those would extend the evidence beyond what the organizer supplied.

Separate assumptions from decisions

An assumption is a temporary working choice. A decision is a choice someone has actually made.

For this party, β€œUse a simple paper sign” might be a harmless provisional assumption. β€œAll guests can climb stairs” would be an unacceptable substitute for checking the route they need to use.

Ask the assistant to keep a short list:

Confirmed: 24 guests, two-hour gathering, $180 remaining, no kitchen or refrigeration, no open flames.

Provisional: One seated group game and simple reusable decorations.

Unresolved: Food preferences and suitable refreshment arrangements; any venue access details needed by attendees beyond those already confirmed.

This list keeps the draft honest. It also tells you what you can change without reopening every settled part of the project.

The broader AI Fluency framework treats deciding what to delegate and describing the task as separate competencies. Asking targeted questions helps you make both decisions deliberately. Claude Academy: The 4D Framework.

Catch contradictions before they become promises

Suppose someone adds, β€œKeep the total below $180, including a $150 entertainer and $60 of refreshments.”

The stated items total $210, leaving a $30 gap before paper goods or activity materials. The assistant should surface that conflict. It should not make the budget fit by silently reducing a confirmed price.

A useful response would give choices: remove the entertainer, find a genuinely confirmed lower price, reduce another expense, or increase the budget. The person responsible chooses.

Ask for the same treatment when timing conflicts. A two-hour event cannot contain three separate one-hour sessions unless they overlap or the scope changes. A tidy schedule can still be impossible.

Stop asking when there is enough to draft

Clarification should lead to useful work. It can become a delay tactic when every small preference is treated as a prerequisite.

After answering the important questions, say:

Use these answers to draft a workable plan now. Put remaining minor decisions in a short final list. Flag only unresolved issues that prevent a part of the plan from being feasible.

You do not need to know the final sign wording before arranging the tables. You do need the table count.

For a reversible choice, a draft can be the best way to discover your preference. Seeing two possible schedules may help you choose faster than discussing scheduling in the abstract.

Students: clarify the assignment before solving the wrong problem

A student might ask for help preparing a presentation without mentioning its length, required sources, or assessment criteria.

Try this instead:

Help me plan my presentation using the assignment sheet below. First identify requirements already answered by the sheet. Ask only about missing details that would change the plan. Do not invent my teacher's expectations.

The assistant should read what is available before asking you to repeat it. If the rubric says five minutes, β€œHow long is it?” is unnecessary. If permitted AI use is unclear, the teacher's policy needs clarification; the assistant cannot resolve it by guessing.

Once the requirements are known, the student can plan their own argument and sources. Keep a record of decisions so a later outline does not drift away from the assignment.

Your exercise

Start with the incomplete party request. Write three essential questions and explain what each answer could change. Then use the supplied fictional answers to create a one-page requirements brief.

Include confirmed facts, provisional choices, unresolved issues, and the next useful task. Do not turn unknown food needs into invented restrictions or assurances.

Completion check: Every question earns its place, answers affect the plan, and unresolved constraints remain visible without preventing unrelated work.

Get the AI Dispatch

Weekly insights on ai & technology β€” delivered to your inbox. No spam, unsubscribe any time.

Want to choose specific topics? Customize your interests