Choose a familiar work situation
Base the exercise on a task participants recognise, such as summarising notes or preparing an internal update. Use a clearly fictional organisation and invented details that cannot be mistaken for a real case. A fictional service team might need a summary of a short planning meeting. Keep the example focused enough to complete in the available practice time. Avoid introducing an unfamiliar industry problem that makes participants spend the session learning the background instead of practising the task and reviewing the output.
Provide enough context to work fairly
Include the audience, purpose, source information and required output format. If the task depends on a business rule, supply the fictional rule in the exercise. Do not expect participants to guess hidden criteria that the facilitator reveals only afterwards. Mark any deliberately missing information so the review can discuss whether the response asked for clarification or invented an answer. The exercise should distinguish a poor instruction from a poor result. That gives participants a practical way to improve their own requests without assuming the tool always understands unstated context.
Write the review points before testing
Define the facts the output must preserve and the decisions it must not invent. For fictional meeting notes, the review might check whether an owner was actually assigned and whether a proposed date remained a proposal. Avoid one model answer that treats every harmless wording difference as wrong. Use criteria that focus on meaning, completeness and suitability for the audience. Keep the source available during review. A polished summary should not pass if it quietly changes an unresolved question into an agreed action.
Include a manageable difficult case
Add one or two realistic complications, such as conflicting notes, an ambiguous abbreviation or an absent detail. The purpose is to practise recognising limits, not to embarrass participants with a puzzle. Ask what the person should verify before using the output. Keep higher-consequence advice and live business actions outside the exercise. If the source cannot support an answer, a useful outcome may be a clear question for the owner. Do not design every example so that producing a complete-looking response is always rewarded.
Keep practice material separate from live work
Label files and messages as training examples and use approved storage and tools. Avoid real contact addresses or identifiers that could accidentally reach someone if copied into another system. The exercise should stop before sending, publishing or changing live records. If participants want to adapt their own material, route it through the agreed preparation process first. A facilitator should not assume that removing one name makes a document suitable to share. Use fully fictional material when the approval or preparation requirements are uncertain.
Review the exercise with participants
Ask participants to explain which parts of the output they would keep, correct or verify. Compare different attempts against the same criteria and discuss the source evidence behind disagreements. Record questions the example could not resolve and improve it before reuse. Do not present success on a prepared exercise as proof of performance across real work. The training result is that participants practised a task and its review process. Any later workplace trial needs its own approved scope and checks.