A useful AI workshop gives people a bounded place to try the technology on work they recognise. The preparation matters as much as the session: the business needs to choose suitable tasks, approve the tools and information involved, and decide what people must still check for themselves.
What to take into your next decision
- Begin with a recurring task and a clear quality standard, not a tour of whichever tool is newest.
- Agree on approved platforms, information and access before anyone brings live work into the room.
- Teach participants to inspect sources, assumptions and omissions rather than accepting fluent output.
- Give each person a practical next step and create a follow-up point after the workshop.
Start with a task people already understand
Tool demonstrations can create excitement without helping anyone decide what to do on Monday. A stronger starting point is a recurring task that participants already know how to complete and evaluate.
Suitable workshop tasks have a visible input, a useful draft or analysis as an output, and a person who can recognise whether the result is good enough. Preparing an email from approved notes, restructuring a meeting summary or comparing a document with a checklist can be easier to examine than an open-ended request to ‘use AI in your role’.
Avoid tasks where an incorrect answer could immediately affect a person's rights, health, safety, finances or employment. Those areas may still benefit from carefully governed systems, but they are poor material for casual experimentation.
- What starts the task today?
- What information does the person need?
- What does a useful result look like?
- Which decisions require experience, authority or professional judgement?
Set the boundary before the session
Participants should not have to work out the organisation's data rules while they are learning. Confirm which platforms are approved, which account types should be used, which features an administrator has enabled and what information must stay out of the system.
Use sample, synthetic or appropriately de-identified material when live information is unnecessary. If real business information is required, the organisation should understand the provider's terms, retention settings, access model and privacy implications first.
Write the boundary in plain language and make it easy to find. ‘Use judgement’ is not enough when different people may reasonably make different assumptions about confidential or personal information.
Use a short intake to make practice relevant
Before the workshop, ask each participant about their role, one recurring task, the tools involved and the part of the work they find awkward. Ask what a good result needs to include and what commonly goes wrong.
The answers help a facilitator prepare examples without collecting unnecessary documents. They also reveal when participants need different exercises or when a proposed task should be excluded because the risk or access requirements are not ready.
The intake is also a chance to name expectations. The aim is not to make every task automatic. It is to learn where AI can assist, where it adds friction and where it should not be used.
Structure the workshop as a working loop
Give enough orientation for people to understand what the tool can and cannot see, then move into a real exercise. Participants should provide context, inspect the first result, identify what is missing and revise their instructions.
A good exercise makes the quality criteria visible. Instead of celebrating a polished paragraph, compare it with the source material, the intended audience and the business's actual standard. Ask what the output assumed and what a person would need to verify before using it.
Leave room for failure. A weak result is useful when the group can diagnose whether the problem came from missing context, a poor source, an unsuitable task or a limitation of the tool.
- Orient: explain the approved environment and the task boundary.
- Attempt: create a first result from suitable material.
- Inspect: check accuracy, relevance, tone and missing context.
- Refine: improve the instructions or the source information.
- Decide: record what a person must still review or approve.
Teach checking as part of using the tool
Verification should not be a warning delivered at the end. Build it into every exercise. Participants can trace statements back to supplied material, test whether a summary omitted an exception and compare a draft with an approved example.
It helps to separate different questions: Is the statement supported? Is important context missing? Is the tone appropriate? Does the person have authority to use this result? Does the result need specialist review?
The facilitator should be willing to say that a task is not suitable. Knowing when to stop, escalate or use the existing process is a practical AI skill.
Plan what happens after the room
A participant should leave with a small next action rather than a promise to ‘use AI more’. That action might be refining an instruction, testing the same task on approved sample material or asking an administrator about a missing control.
Nominate where useful instructions and lessons will live, who can approve broader use and how people can report a poor or concerning result. A planned follow-up lets the team compare what was genuinely useful with what sounded promising in the session.
That feedback is more valuable than an attendance count. It can inform further training, a workflow assessment or a decision to leave the process as it is.
The TechGuider introduction session
Our introduction session runs for at least 120 minutes, with up to eight participants and a follow-up after 14 days. Preparation happens beforehand and is additional to session time. We confirm the trainer, approved tools, suitable material and delivery arrangements before booking.
Sources and limitations
The practical method above is TechGuider’s suggested approach. These references explain the underlying guidance; they do not endorse TechGuider or prove an outcome for your business.
- NIST: Generative AI Profile
A 2024 companion covering generative AI risks, including confidently stated false information.
- Guidance for staff training on AI
Written for Australian Government staff and used here as a useful reference, not a rule for private businesses. It covers AI basics, prompting, output checking, privacy, security, bias and accountability.
- Guidance on privacy and the use of commercially available AI products
Australian privacy guidance for assessing products, handling personal information and training staff who use AI systems.
Suitability depends on your task, information and review process. A source-check date records when the references were checked; it is not professional certification.
How to request a correction ↗