Build a short practice exercise that shows what someone understands, how they use AI and where they still need support.
A team has attended an AI workshop. Everyone has tried a few prompts and produced a polished summary. What can they now do that they could not do before?
That is a useful question to answer before adding another tool demonstration to the calendar.
On November 5, OpenAI described how Australian universities are combining AI access with practical learning and assessment. Its account includes students practising realistic scenarios and institutions setting expectations for when AI belongs in an assignment. It describes programmes and plans, rather than establishing that access alone produces better workplace performance.
One example offers a useful starting point for business training. UNSW's October 13 explanation of its assessment approach distinguishes work where students may use AI, work where they must demonstrate capability without it, and work where using AI is itself part of the learning goal. The university describes this as a programme design decision, with different assessments serving different purposes.
A workplace exercise can borrow that distinction while staying focused on the job people need to learn.
Start with one observable skill
Replace a broad objective such as “learn AI for sales” with something a colleague can review: prepare a short account brief that separates confirmed facts, open questions and proposed next steps.
Use a fictional account and a small, clearly labelled practice pack. Include a product description, a meeting note and a simple table of invented figures. Give the pack a version and provide the same materials to everyone. Keep real customer information and confidential documents out of the exercise.
Write down what a useful brief must contain before the session begins. For this example, the reviewer might look for accurate figures, clear attribution to the practice documents, explicit uncertainty and a next step that follows from the evidence.
Choose the role of AI deliberately
Give each part of the exercise a reason.
Start with a short task completed without AI, if the aim is to see whether the person can interpret the underlying material. Ask them to identify the account's stated goal and calculate a simple total from the practice table. This establishes what needs teaching before a generated answer enters the discussion.
Then make AI part of the task. Ask the participant to produce a first draft using the approved tool, check it against the practice pack and revise it. Have them identify one useful contribution from the tool and one part they changed or rejected.
For a final task, let them choose whether to use AI. Ask why they chose that approach. Someone may decide that a short calculation is quicker to do directly, while using AI to organise a longer note. The explanation gives the trainer something more useful to discuss than a count of prompts.
These are suggested workshop exercises, not a claim that UNSW uses this sales-training format.
Review the explanation alongside the result
A well-written brief can conceal a weak understanding of the material. Ask the participant to walk through a few concrete decisions:
- Which source supports this statement?
- What did you change after checking the draft?
- Which question remains unanswered?
- What additional information would change your proposed next step?
Include a deliberately missing detail in the practice pack, and tell participants that some questions may be unanswerable from the supplied information. A useful response should preserve that gap rather than fill it with an invented fact.
Keep the discussion supportive. If someone cannot explain a calculation or locate a supporting passage, use that finding to decide what to practise next. The exercise should help the trainer give specific feedback.
Check what carries over
A few days later, offer a different fictional account with the same learning objective. Look for whether the person can apply the method to unfamiliar material, not whether they remember the first prompt.
Keep a short record of the skill practised, the feedback and the next exercise. If you also measure time, define what the clock includes, including checking and correction. That connects to being precise about an AI pilot's claimed time savings.
A useful training outcome is a person who can explain the work, spot a gap and improve an answer. Tool access and a good demonstration can support that outcome. The practice task is where the team can see what still needs teaching.