Lesson 12 of 19
Practise a difficult scenario
Discuss a fictional problem and practise an appropriate response. Record what happened and what further checking is needed.
This lesson concerns artificial intelligence (AI) governance: how your organisation assigns responsibility, checks AI use and records decisions and follow-up. AI Governance keeps these records; it does not operate the external AI tool for you.
At the end of this lesson
- Organise a safe manual discussion.
- Record observations and response responsibilities.
- Turn the discussion into owned follow-up without claiming a passed tool test.
Before you begin
This lesson is for facilitators, reviewers and affected staff. You can read and complete the paper activity without an account.
Use the fictional facts below for paper practice. No AI tool or real incident is needed. Saving a workplace exercise needs current permission to edit the relevant activity. If access is blocked, record your proposed response in your notes and ask the responsible administrator to check access.
Key ideas
Tabletop exercise means discussing a fictional situation. Observation, correction, underlying control improvement and tool performance are different.
Follow these steps
1. Choose the exercise
Open Test your scenario from Tools and activities, or use the paper exercise below. A tabletop exercise is a discussion of a fictional situation.
Check the result: No external AI tool or real incident is needed.
2. Set the boundary
Identify the facilitator, participants, no-blame purpose and task. Use the linked facilitator guidance. Keep real client or staff information out of the learning activity.
Check the result: Participants know that this is a manual practice exercise.
3. Brief the participants
Read the source and draft facts below. Ask participants to work from those facts rather than inventing a tool result.
Check the result: Everyone has the same information for the discussion.
4. Discuss the response
Ask what should be withheld, who checks the source and who receives the concern. Consider how the response changes if the error has already been shared.
Check the result: The response identifies responsibility and an immediate action.
5. Record observations
Write what people actually said or demonstrated, what was corrected and what remains unresolved. Distinguish an observation from a proposed future action.
Check the result: The record shows the limits of the exercise instead of claiming a successful live test.
6. Assign follow-up
Record a priority task, responsible person and completion check. In workplace practice, use the supported activity review/save process and reopen the record.
Check the result: Follow-up can be tracked. One corrected example does not establish the tool’s accuracy or safety.
Practise
Paper activity — fictional information
'Fictional source: "The request is due on 20 October 2026." Fictional draft: "The request is due on 21 October 2026."' Identify the discrepancy, state the immediate response and name a follow-up checking task.
Write your response in your own notes before opening the answer. You do not need to sign in, change records or run an assessment.
Check your understanding
Does correcting this one draft prove that the AI tool is safe to use?
Check your answer
No. Withhold/correct the inaccurate draft using the source and refer it to the responsible reviewer. Record the checking gap and follow-up. One manual discussion or correction does not establish tool performance. Reread steps 5–6 if the response declares a successful operational test.
Check your work
- The exercise is clearly fictional and manual.
- The observation and response are traceable.
- Follow-up has an owner and completion check.
Continue learning
Optional worked examples
Sources
These National AI Centre resources inform the lesson. This course does not imply Australian Government endorsement.