Skip to main content
PraktickAI
CoursesFor TeamsServicesDocsBlog
CoursesFor TeamsServicesDocsBlog
HomeBlogAI gave you a vague plan? Turn your task brief into a usable prompt
October 9, 2026·8 min read

AI gave you a vague plan? Turn your task brief into a usable prompt

Prompt EngineeringProductivity
Jump to section
  • 1. Start with the messy request
  • 2. Copy the complete workshop prompt
  • 3. What an acceptable draft could look like
  • 4. Three checks before you use the answer
  • 5. Copy the worksheet for your own task
  • 6. Try a different task: staff induction
  • One possible completed prompt
  • Illustrative answer to that prompt
  • 7. Why use this structure?

“Help me plan a workshop” can leave you with a polished list and no way to tell whether it fits your situation. A usable brief tells AI what is known, what is still undecided and what the answer must contain. Below, turn a scrappy library request into a planning table you can actually check—then try the same method on a staff induction.

All situations here are fictional. The sample table is an authored illustration, not output from a recorded AI test. It shows what an acceptable answer might contain; a model may respond differently. You can complete the worksheet without using AI.

1. Start with the messy request

We want an evening workshop at the library for 12 beginners. One facilitator, 60 minutes. We already have a room and a projector. Please make a plan. The date and budget are not confirmed.

That request contains useful facts. It does not specify the workshop topic or learning outcome, and “we have a room” does not confirm a booking for a particular evening. A calendar date, spending limit or promise that participants will master a skill would add information the brief never supplied.

Part of the briefWhat belongs here
Given factsLibrary; evening session; 12 beginners; 60 minutes; one facilitator; existing room and projector.
Missing decisionsTopic and intended learning outcome; date and start time; budget. Room availability for the chosen evening is unverified.
ConstraintsPlan within 60 minutes and one facilitator. Do not assume purchases, additional staff or participant devices.
Desired outputA short facts/unknowns summary, a timed draft agenda and questions needed before confirmation.

“Missing decisions” can also contain things to verify. Keep those visible instead of quietly turning them into facts. If you want AI to suggest an option, explicitly ask it to label the option as a proposal.

2. Copy the complete workshop prompt

A prompt is simply the instruction you give AI. This one separates the input from the rules and specifies the answer. Copy the whole block; the labels are part of the instruction.

Task: Draft a plan for a fictional library workshop.
Write in plain English for the library coordinator.

GIVEN FACTS
- An evening workshop for 12 beginners.
- One facilitator; session length: 60 minutes.
- The library already has a room and a projector.

UNKNOWN / NOT CONFIRMED
- Workshop topic and intended learning outcome.
- Date and start time; room availability that evening.
- Budget and participant devices.

CONSTRAINTS
- Do not invent a date, budget, booking or learning outcome.
- Do not assume purchases, more staff or participant devices.
- Keep every unresolved item visible as “not confirmed”.
- Label suggestions as “proposal”, not confirmed facts.
- Draft only; do not book, spend or announce the event.

OUTPUT
1. Summarize the given facts and unresolved items separately.
2. Give a provisional agenda table with columns:
   Minutes | Activity | Resources | Status / dependency.
   Intervals must cover 0–60 minutes without gaps or overlaps.
   Keep topic-specific work conditional on topic confirmation.
3. List up to five questions to resolve before confirming the plan.
4. End with a check of facts, unknowns and the requested format.
   Report any mismatch rather than claiming the plan is confirmed.

The improvement is something you can inspect: you have defined the columns, the time limit and how to handle missing information. These instructions do not guarantee a correct answer. They give you a clearer basis for accepting or repairing one.

3. What an acceptable draft could look like

Given facts: an evening library workshop, 12 beginners, one facilitator, 60 minutes, existing room and projector. Not confirmed: topic, learning outcome, date, start time, room availability that evening, budget and participant devices. The time split below is a proposal; none of the activities has been booked.

MinutesActivityResourcesStatus / dependency
0–5Welcome and explain the session goalOne facilitator; roomProposal; goal wording depends on topic and outcome confirmation.
5–15Introduce the agreed topic with one exampleFacilitator; projectorProposal; topic and suitable example not confirmed.
15–35Guide one beginner practice activityFacilitator; roomProposal; choose the exercise after topic confirmation. No device or purchase assumed.
35–50Discuss the practice and take questionsFacilitator; roomProposal; content depends on the exercise.
50–60Recap and invite a next stepFacilitator; projector if usefulProposal; next step depends on the agreed learning outcome.
  • Questions before confirmation:
  • 1. What topic and one achievable beginner learning outcome do we want?
  • 2. Which date and evening start time are proposed?
  • 3. Is the room available then, with workable projection and suitable access?
  • 4. What budget, if any, has been approved?
  • 5. Will participants have devices, or should the practice work without them?

The durations add up to 5 + 10 + 20 + 15 + 10 = 60 minutes. That checks the arithmetic, not whether the session will work. The coordinator still needs to choose the topic, check practical arrangements and confirm the plan.

4. Three checks before you use the answer

  • Facts: Compare the answer with the original brief. Are 12 beginners, one facilitator and 60 minutes preserved? Reject any invented date, spending limit or booking.
  • Unknowns: Can you still find the unresolved topic, outcome, date, start time, room availability, budget and devices? A suggestion must remain a proposal.
  • Output: Does the answer include the requested summary, four-column agenda and up to five questions? Do the time intervals cover exactly 0–60 minutes?

If AI writes “The room is booked for Tuesday”, a focused repair is: “Tuesday and the booking are unsupported. Replace them with ‘not confirmed’, retain the given facts, and regenerate the affected rows.” Then check the repaired answer again. If the brief itself is wrong or incomplete, update it before asking for another draft.

5. Copy the worksheet for your own task

Replace every bracketed field with your own non-sensitive information. A field you cannot fill stays “not confirmed”. Choose a task where a draft is useful, such as preparing a meeting agenda, rather than asking AI to make commitments for you.

TASK AND READER
[What draft do I need? Who will use it?]

GIVEN FACTS
[Only information supplied or verified for this task.]

UNKNOWN / NOT CONFIRMED
[Missing decisions and facts still to verify.]

CONSTRAINTS
[Time, resources, scope and actions not authorized.]
Do not invent missing facts. Mark suggestions as proposals.
Keep unresolved items visible as “not confirmed”.
Draft only; do not make commitments or take external action.

OUTPUT
[Required sections, columns, length and any timing rule.]
Separate given facts, unknowns and proposed actions.
End with a check of facts, unknowns and requested format.

MY ACCEPTANCE CHECK
Facts match the brief: [ ]
Unknowns are still visible: [ ]
Output matches my requested structure: [ ]

6. Try a different task: staff induction

A fictional office needs a 30-minute induction for six new colleagues. One shared public checklist is available. No date or session leader is confirmed. Prepare a draft agenda; do not schedule or send invitations.

Before reading the answer below, sort this request into the four worksheet sections and write your desired output. Keep the date and leader unresolved. A useful answer can suggest an agenda without claiming somebody has agreed to lead it.

One possible completed prompt

TASK: Draft an induction agenda for the office coordinator.
FACTS: Six new colleagues; 30 minutes; a shared public checklist.
UNKNOWN: Date, start time, session leader, meeting format/location.
CONSTRAINTS: Do not schedule, send invitations or invent a leader.
Label the agenda as a proposal; keep unknowns not confirmed.
OUTPUT: Facts/unknowns summary; agenda table with
Minutes | Activity | Resource | Status / dependency;
up to three questions needed before confirmation.
Cover 0–30 minutes with no gaps or overlaps.

Illustrative answer to that prompt

This sample answer was written by us, not recorded from an AI run. Given facts: six new colleagues, 30 minutes, one shared public checklist. Not confirmed: date, start time, session leader and meeting format/location. The agenda below is a proposal; nothing is scheduled and no invitations have been sent.

MinutesActivityResourceStatus / dependency
0–5Welcome and explain the purpose of the inductionSession leader, once confirmedProposal; requires a confirmed leader and meeting arrangements.
5–20Walk through the shared public checklistAvailable public checklist; confirmed leaderProposal; the leader still needs to check the checklist’s relevance to these colleagues.
20–30Questions and recap of unresolved itemsPublic checklist; confirmed leaderProposal; requires a confirmed leader and meeting arrangements.
  • Questions before confirmation:
  • 1. Who will lead the session?
  • 2. What date and start time are proposed?
  • 3. Will it be in person or online, and what location or meeting link will be used?
  • Check: 5 + 15 + 10 = 30 minutes; six colleagues and the checklist remain given facts.
  • Check: date, start time, leader and meeting arrangements stay unresolved; no invitation was sent.
  • Check: the answer distinguishes proposals from facts and uses the columns you requested. A room reservation or a named leader would fail the fact check.

7. Why use this structure?

Anthropic’s prompting guidance recommends explicit instructions, output formats and constraints, and discusses examples. Its overview starts with defining success and checking it. OpenAI’s prompt engineering guide explains using headings and lists to separate instructions and context, and examples to show expected responses. Sources checked on 9 October 2026. The worksheet is our practical adaptation, not a provider guarantee.

Your next step: copy the worksheet, fill it with one real non-sensitive task, and use the three checks on the resulting draft. Keep one unresolved decision visible instead of letting a convincing sentence decide it for you.

  • 1. Start with the messy request
  • 2. Copy the complete workshop prompt
  • 3. What an acceptable draft could look like
  • 4. Three checks before you use the answer
  • 5. Copy the worksheet for your own task
  • 6. Try a different task: staff induction
  • One possible completed prompt
  • Illustrative answer to that prompt
  • 7. Why use this structure?
LinkedInX / Twitter
Karel Čech

Karel Čech

Developer and AI consultant. I help technical teams adopt AI in their daily workflow — from workshops to long-term strategies.

LinkedIn →

Stay ahead with AI insights

Practical tips on AI for dev teams. No spam, unsubscribe anytime.

By subscribing you agree to our privacy policy.

Related posts

AI training for your team: start with five answers that hold up

Claude Frontier Academy starts with a named project. Define five FAQ answers, their evidence, an accountable owner and an acceptance test using this worksheet.

Your slides are ready. Can you answer these three objections?

Rehearse a six-week documentation pilot: replace “AI will improve productivity” with a decision, a quality check and a stop rule. Copy the worksheet for your next proposal.

One brief, three AI ad concepts: a worked example you can reuse

Turn a specific brief into three distinct ads. Follow the sample brief, complete copy and visual ideas, one unsupported claim and its correction, plus a prompt you can copy.

PraktickAI

AI training & consulting for technical teams

CoursesFor TeamsServicesDocsBlog
karel@praktickai.appLinkedIn

© 2026 PraktickAI — SkyVisual s.r.o. All rights reserved. Privacy Policy · Terms & Conditions · Complaints Procedure