ShapedSI Journal · Learning
AI training for Canadian teams: build habits around real work
A useful AI workshop leaves people with a repeatable way to do part of their job. Start with the tasks your team already performs and the tools it is allowed to use. Training should cover how to prepare an input, review an output, protect information and recognise when to stop and ask a person.
In this guide
01
Choose a few tasks by role
Ask participants which task they would like to make easier. Group the answers into a small set of exercises, such as preparing an internal summary, drafting a customer reply or turning notes into a checklist. Select examples that staff understand well enough to judge. Avoid spending the session on an unfamiliar demonstration that nobody can use afterward.
Use fictional, public or appropriately anonymised information for the exercises. Confirm which tools and account types are approved before the workshop. People should leave knowing where they can practise and what they may put into the tool, rather than receiving a prompt that assumes access to a different system.
02
Teach a complete working method
Give the task, the relevant context, the desired output and the limits. Then review what the tool produced against the original information. Ask participants to notice missing details, unsupported claims and a tone that does not fit the situation. A polished answer is not evidence that every statement is correct.
Have the team revise the request and compare the result. This helps people learn which instructions matter for their own work. Save the useful pattern with a short explanation of when to use it and how to check the answer. A template is easier to maintain when it describes a task instead of relying on a collection of clever phrases.
03
Agree on the review standard
Decide which outputs can remain internal drafts and which need a named reviewer before they are shared. For a customer response, that may mean confirming the service details, the recipient and any commitments. For a summary, it may mean checking the key points against the original document.
Include a stop condition in the exercise. If the tool cannot find supporting information, if the input contains material outside the agreed boundary or if the output could affect a consequential decision, staff should know who to ask. Practise that handoff, not just the successful example.
04
Make learning continue after the session
Choose an internal contact who can collect questions and keep the approved templates organised. Arrange a follow-up that reviews examples from actual use with sensitive information removed. The questions people bring back will show which parts of the method need more practice.
Judge the training by the quality of the work and the confidence to review it, rather than by how many prompts participants saved. Review whether the exercise helped with preparation, correction and the handoff to another person. Keep the training materials current when your tools, policies or services change.
Bring this to the conversation
A working brief.
- A short list of role-specific tasks
- Approved tools and practice data
- A repeatable input-and-review method
- Clear rules for sharing outputs
- A contact for follow-up questions
A few useful questions.
Do we need advanced technical knowledge?
Not for many everyday exercises. Start with a task participants understand and build from the review habits they already use.
Should everyone learn the same prompts?
A few shared principles help, but role-specific examples make the practice useful. Keep the task, data boundary and review standard visible.
Original sources
Sources and further reading.
Microsoft’s Copilot adoption resources include user enablement and champion support. The broader lesson for a workshop is to plan ongoing practice and support alongside the initial session.
Shape the next step