Purpose and fit

Use this to turn a specific work example into shared learning. Invite the people who create, review and receive the output. Keep the discussion grounded in the task and the evidence available.

Before you start

Use an authorised existing work example, its brief and available reviewer notes. Supply the acceptance criteria and any known corrections; say which records are missing. If there is no completed example yet, request an agenda and blank learning log only, rather than creating an invented success case.

Use only information you are authorised to share with the chosen AI tool. Adjust the prompt to the fields and context you actually have; a specially created dataset is not required.

Required inputs

  • One authorised, sanitised example with the task brief, inputs and output
  • The acceptance criteria and a record of human corrections
  • Tool/version or configuration details needed to understand the example
  • Representatives from the affected roles and a facilitator

Practical workflow

  1. Prepare a short case packet. Remove unnecessary identities and confidential material. Record what is unavailable or reconstructed.
  2. Explain the intended result and the conditions under which the task was performed. Let reviewers inspect the example before hearing a success claim.
  3. Compare the output with the acceptance criteria. Identify useful parts, defects, missing checks and downstream effort.
  4. Separate observations about this case from hypotheses that need another test. Ask what might change in a different team, task or data setting.
  5. Agree a small number of actions with owners and follow-up evidence. Reuse a practice only within the context the reviewers can support.

Expected output

Expect a case-specific agenda and draft learning log, or an unfilled template if evidence is absent. The output is not an agreed meeting record or proof that the practice works generally.

An agenda and a learning log. A suggested 45-minute agenda, adjustable to the task:

  • 5 minutes: task, context and boundaries
  • 10 minutes: inspect the example and review evidence
  • 15 minutes: discuss useful practices, failures and downstream work
  • 10 minutes: choose changes or further tests
  • 5 minutes: confirm owners and review points

Learning-log fields:

Observation | supporting example | uncertainty | implication for this workflow | proposed change/test | owner | evidence to review

Human checks

The facilitator gives reviewers space to disagree and avoids turning the session into a performance ranking. Someone receiving the work checks downstream usefulness. The process owner decides whether a proposed change is within current authorisation. No one treats a generated summary as the definitive record without checking it.

Limitations

One example cannot establish a general success rate or justify broad rollout. Selected demonstrations can hide difficult cases and failed attempts. If a live incident is identified, use the existing incident route immediately rather than waiting for the meeting.

Starter prompt

Prepare a practice-review agenda and learning-log draft for this sanitised case: [case packet]. Intended output and acceptance criteria: [criteria]. Participants’ roles: [roles].

Summarise only what the packet supports. Separate observed facts, reviewer interpretations and questions for further testing. Identify review/rework effort, possible downstream effects and missing evidence. Suggest discussion questions that help different functions test the conclusions. Leave decisions, action owners and due dates blank unless supplied. Do not invent consensus or describe this case as a proven success.

Keep facts, interpretations and further-test questions separate. Link each factual observation to the supplied case evidence; mark unsupported conclusions unresolved. Leave consensus, owners and dates unrecorded unless provided. Match the agenda length to the available meeting time.

Before drafting, ask a focused question if a missing detail materially changes the task or acceptance criteria. Otherwise produce a clearly limited draft and list the gaps. Label any non-critical assumptions; never invent facts, figures, sources, approvals or completed actions.

Data and AI output disclaimer

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.