The number nobody puts in a case study
The Australian Bureau of Statistics asks businesses whether the thing they changed actually produced anything. In the 2024-25 Business Characteristics Survey, three quarters of innovating businesses reported a benefit. Of the remainder, 14% reported no benefits were achieved from innovating, and another 11% said it was too early to measure.
So something close to one business in seven did the work, paid for it, and got nothing back that it could name. That is not a story about bad technology. Most of those businesses bought something that worked.
The same survey asks what held businesses back from using technology at all. The most common answer was insufficient staff skills and capabilities, at 16%. That is the answer businesses give, and it is the wrong diagnosis for what happens at week six. By week six the skills are there. People sat the session, they did the task, they could do it again if you asked them in the room. What has gone is the reason to.
Why week six specifically
The first fortnight runs on novelty and on the fact that somebody is watching. There is a new thing, someone senior cares about it, and the person who ran the rollout is still around asking how it went.
Weeks three and four run on the first busy period. This is where the split happens. Under pressure people fall back to the method they can do without thinking, and for most of your staff that is still the old one, because they have done it four hundred times and the new one eleven.
Week six is when the person who ran the rollout has moved on to the next thing. The training is over, the project is marked complete, and nobody is asking. If the only thing holding the new way in place was attention, that is the week the attention stops.
It is a management attention problem wearing the costume of a skills problem. Businesses respond by booking more training, which addresses the costume.
What actually holds it, and none of it is training
Three things, and all three sit in a manager's week rather than in a staff member's skill set.
The first is that a named person still asks about it after the project is closed. Not a survey and not a dashboard. One line in a weekly meeting that already happens, owned by the manager of the team doing the work rather than by whoever ran the rollout.
The second is that the manager does it too. Where a team leader kept using the old spreadsheet privately because it was faster for them, their team read that correctly and went back. Staff calibrate off what their direct manager actually does, not off what the training said.
The third is that the new way is the only one still available. Where the previous template, form or spreadsheet is still sitting on the intranet, it gets used the first time someone is under pressure, and every time after that.
What to do about it
Put a date in the calendar six weeks after the last training session, before the training happens, and give it to the line manager rather than the project. On that day, watch three people do the task and count how many use the new method. That is the whole review.
If it is most of them, the change has held and you can stop looking. If it is not, you have found it at six weeks instead of at the renewal, and the fix at six weeks is usually one conversation with one manager rather than another round of training for twenty people.
None of this is about AI. It is ordinary change management, and it decides whether the licence you are paying for is doing any work by March.