Purpose and fit

Use this when a team starts with a tool, a demonstration or a broad ambition such as “use AI in operations”. Produce a brief that explains why the work needs to change. A leader should be able to understand the problem without knowing which technology might address it.

The useful outcome is an agreed question worth investigating. It is too early to promise savings or select a supplier. Keep the scope to one recurring difficulty affecting a named group.

Before you start

Start with any existing authorised request, process note or recent example that shows the difficulty. One example is enough to draft a provisional brief; add more only if available. Remove identifying details and supply the affected roles, intended improvement and known constraints. Adapt the prompt to your actual notes rather than creating a dataset.

Expect a provisional one-page problem brief, not proof of prevalence, savings or AI suitability. Each observation should point to a supplied example; causes remain hypotheses and the next investigation addresses a named uncertainty.

Required inputs

  • One or more recent, sanitised examples of the difficulty, if available, including the result received by the next person; otherwise start with a clearly provisional description.
  • The people who request, perform and depend on the work, plus an owner able to resolve scope disagreements.
  • Available observations of waiting, rework, missing information or unmet needs; label estimates and gaps.
  • Constraints on quality, access, effort and timing, and any earlier attempts to improve the work.

Practical workflow

  1. Describe the current situation in plain language. State who is trying to achieve what, where the difficulty occurs and what consequence follows. Remove product names from this first description.
  2. Walk through the examples with both the person doing the work and its recipient. Separate an observed symptom from a proposed explanation; repeated corrections may reflect unclear requests rather than poor drafting.
  3. Define the smallest useful outcome. Specify what the recipient should receive or experience differently and the unacceptable trade-offs. Leave numerical targets unresolved if there is no defensible basis.
  4. List plausible responses: remove a requirement, clarify information, change responsibility, use an existing feature, introduce AI assistance or leave the process unchanged. Give each a reason to consider it.
  5. Identify the uncertainty that prevents a sensible next decision. Choose a short investigation, such as observing several requests or comparing source completeness, and name who will collect the evidence.
  6. Have the problem owner and affected people revise the brief together. Record disagreements and a review date rather than turning uncertain assumptions into an agreed fact.

Expected output

A one-page opportunity brief using these copyable fields:

Problem owner: [role/name] Affected people: [performer and recipient] Current situation: [task and friction] Observed examples: [sanitised references] Consequence: [what becomes harder] Desired outcome: [observable improvement] Unacceptable trade-offs: [boundaries] Possible causes: [hypotheses] Options: [including non-AI and no change] Critical unknown: [question] Next investigation: [owner, evidence and review date]

Hypothetical example: an internal support team rewrites incomplete requests before assigning them. The brief asks how to improve request completeness. An intake change and AI-assisted clarification remain separate options until the team understands where information is lost.

Human checks

The recipient confirms that the proposed outcome would help. Frontline colleagues check whether the examples are typical, and the problem owner checks that the scope is actionable. Challenge explanations that blame individuals without examining unclear instructions, conflicting incentives or missing information.

Limitations

A clear problem statement does not prove that AI is suitable. A few examples can reveal questions but cannot establish their frequency. Revisit the brief when observation contradicts the original explanation or when the intended beneficiary changes.

Starter prompt

Help frame a problem before selecting a solution. Use only these sanitised notes: [examples, affected roles and constraints]. The accountable problem owner is [owner]. Describe the current situation, the consequence and a bounded desired outcome without assuming AI is needed.

Separate observations, possible causes and missing evidence. Compare removal, process clarification, existing tools, AI assistance and no change. Produce a brief with these fields: problem owner; affected people; current situation; observed examples; consequence; desired outcome; unacceptable trade-offs; possible causes; options; critical unknown; next investigation, including its owner, evidence and review date. Suggest one small investigation that would resolve the most important uncertainty. Use supplied examples; any optional illustration must be labelled hypothetical and kept separate from evidence. Do not invent baseline figures, select a supplier or authorise operational changes.

Adapt the requested fields to the authorised inputs actually supplied. If examples are incomplete, retain a narrower problem statement and ask for the one missing fact most likely to change it. Preserve conflicting accounts by role or source instead of inventing consensus. State the expected output and acceptance criteria before drafting. Ask for essential missing information; where a useful partial draft is possible, label its limits and assumptions. Do not invent facts, evidence, results or approvals.

Data, model variability and your responsibility

Use relevant data that exists or that you provide, with accuracy and completeness sufficient for the task. Different AI models and versions may produce different results, and responses may vary between runs. You are responsible for checking and validating the output against your own requirements, source information and acceptance criteria before relying on it.