A shortlist of names is easy to collect. Judging them is the work. Buy the session only if it changes a task the team already owns.
In 2026 plenty of pages will offer to train your staff. The useful part is what you can ask before anyone sends a proposal. The points below are that list.
What to ask
Start with the shape of the work. A demonstration shows a polished example on a sample the trainer chose. A hands-on session puts the team's own task on the table and lets people try, miss and correct it. If the proposal cannot name that task, you are booking a tour.
Ask who the room is for. One generic exercise for a mixed audience slides back to whatever the trainer can show everyone. The finance person, the lawyer, the service lead and the operator should each leave with a use that belongs to their job. A provider that runs the same exercise for every function is teaching a product, not a team.
Ask whether the trainer is tied to one product. If the only tool in the room is the one they sell, or the one they are paid to resell, the exercise will bend toward that product. The better question is which tools the organisation is already allowed to use, and whether the session can run on those.
Ask about data before anyone pastes a live file. What may be typed into a tool. What must stay out. Who is accountable when a person is unsure. Australia's Voluntary AI Safety Standard is the public document for organisations using AI. A provider that cannot say what may be pasted has not done this work.
Ask what happens after people go back to their jobs. Someone should own one use, and there should be a check that is not a fresh steering pack. When a team needs that follow-through built as a program, the AI adoption sprint is what Collective Campus runs for it. A proposal that ends when people stand up has sold the easy part.
Red flags
Treat these as reasons to keep looking. A case study with no task you recognise. A promise that everyone will feel confident. A tool list in place of a decision about the team's work. Slides that never open the tool people will have when they return.
Another flag is a room that is only allowed to admire the output. If people cannot say the tool was wrong, and cannot show the check they used, the session trained an audience. How to judge that check is in how to judge hands-on AI tool training.
In-house work and a public course
A public course puts people from different organisations in one room. The task has to be generic, because the group does not share a workflow. That can be a fair way to see a method. It is a weak way to change how this team works when they are back at their desks.
An in-house session can use the work already on the desk, the rules the organisation already has, and the roles that will touch the tool. It can also be plain about what not to use. Online platforms sit in a third group. They help a person explore. They are thin when the point is a decision the team will share.
The same questions apply for a team in Brisbane, Sydney or Melbourne, including people who are not in one room. Ask how the session changes for that city, or for people joining from elsewhere in Australia. A provider with one generic national page, and no answer for the city the team sits in, is guessing at the fit.
Questions to put to any provider
Put these to any provider, in writing. What task will we bring, and who owns it. Which roles are in the session, and what each person should be able to do afterwards. Which tools will we open, and can we use ours. What may be pasted, and what is refused. Who follows up, and what they are checking. If the proposal cannot answer those, the brochure is doing the selling.
Where Collective Campus fits
Collective Campus has done capability building for teams since 2015. AI is one part of that work, alongside testing ideas, customer experience and how people work together. AI is not the whole of the practice, and the session does not need the team to adopt a single vendor.
The AI training workshop is for a team that can bring a live task and will say what they already believe the tool will do with it. The session is hands-on. It is for people who are not engineers. People should leave able to run the check without the trainer beside them.
What to bring
Bring one live task, the roles that touch it, and the rule for what may be pasted. If every provider on the shortlist can only show a sample they carried in, you do not have a shortlist yet.
