Ask around and you'll find plenty of Australian businesses that have already done AI training. Ask what changed afterwards and the answer is usually a version of "everyone found it really interesting".
That's a failure, and it's a designed one. Four of the five reasons below are settled before the session begins, which is the good news: they're fixable by choosing differently, not by finding a better presenter.
1. The room left without permission
The most common one, and the most invisible.
A team learns some techniques. Nobody with authority has said which tool is sanctioned, what may not go into it, or what has to be checked before something goes out. So on Monday, every individual faces the same question privately — am I allowed to use this on real work? — and the cautious ones, usually the most senior and most useful, quietly answer no.
The fix costs forty minutes: the boundary block goes first, and the person who can approve it sits in the room while it happens. We put the reasoning for that sequence in its own article because it's the single highest-leverage design choice.
2. The example was somebody else's work
Sessions built on generic examples produce generic confidence. People nod along to a demonstration about a fictional business, then go back to their own documents and discover the fictional one didn't have their template, their compliance line, or their forty-page appendix.
A session should be built on one job from the business, chosen beforehand, that at least half the room does regularly. That's it. The difference in what people do afterwards is not subtle.
What makes a good job to build the session on
- Several people in the room do it, at least weekly
- It's mostly writing or restructuring, not judgement
- It can be demonstrated without client information
- It has a recognisable finished state
- It's currently done after hours
3. Nothing was saved
The clearest predictor. A session that teaches method produces a good afternoon. A session that ends with two or three saved setups — holding your context, your examples, your format — produces a change that's still there next month, because the effort of getting a good result has dropped to nearly zero for one named job.
Without that, every future use starts from a blank box, and a blank box at 4pm on a busy day loses to the old way every time. This is what the reusable setups module exists for, and it's why a session that runs out of time before reaching it has failed even if the first three hours went well.
4. It was aimed at the wrong altitude
Two versions of this.
Too low: a tour of features and prompts, aimed at people who were never going to become enthusiasts. They don't need to enjoy the tools; they need one job to be less awful.
Too high: strategy, disruption, what this means for the profession. Interesting, discussable, and it leaves nobody with anything to do differently on Monday. If a session can be summarised as "AI is important and we should think about it", the business has bought a conference talk.
The altitude that works is a specific job, done all the way to something you would actually send.
5. Nobody checked three weeks later
Most training gets evaluated by a feedback form at the end of the day, which measures whether the afternoon was pleasant.
The question that measures anything: which job is now being done differently, and by whom? One named job, done differently by more than one person, is a successful session. Enthusiasm with no named job is a pleasant afternoon.
Ask it at three weeks, not three months. By three months the answer is unrecoverable — people can't remember what they changed — and the window for a small correction has closed.
If you've already done a session and got nothing from it, that's worth ten minutes on a call. The diagnosis is usually one of the five above, and the fix is rarely another workshop.
Start an enquiryThe awkward implication
If four of these are decided before the session, then most of the value of training sits in the design: which job, who's in the room, what order the blocks run in, and whether anything gets saved.
Which also means you can do a fair amount of this yourself. The guides cover the boundary and the sequence at no cost, and a business with a clear-headed owner and one obvious job to fix can run a decent version internally. Where we tend to earn the fee is in regulated settings where the rule has to be right, and on the setups — because the gap between one that works in the room and one a team still uses in March is mostly experience of watching them break.