Most AI training produces enthusiasm. Very little of it produces lasting behaviour change. Here is why — and what we designed differently.
There is a pattern that recurs in corporate learning and development that the industry has known about for decades and has largely failed to solve. A programme runs. Participants are engaged. Feedback scores are high. Three months later, very little has changed about how people work.
AI training has the same problem — and in some ways a worse version of it, because the tools themselves change so quickly that even people who do retain what they learned find that the specific steps they practiced are different when they try to apply them in their own environment.
We thought about this problem carefully before we built a single activity. Here is what we concluded.
The most common format for AI training is a presentation. Someone demonstrates the tools. Shows what they can do. Participants watch, take notes, feel impressed. Then they go back to their desks and open their usual applications.
The problem is not that the demonstration was bad. The problem is that watching someone else use a tool is categorically different from using it yourself. The muscle memory is not there. The confidence is not there. The specific knowledge of how to prompt for your own work — not a generic example someone else chose — is not there.
Watching someone drive a car does not teach you to drive. It shows you that driving is possible and gives you some vocabulary for it. The learning only starts when you are in the driver’s seat.
A single day of AI training can shift attitudes. It can demonstrate possibility. It can motivate. What it almost never does is change behaviour, because behaviour change requires repetition, and one day of any new skill does not produce enough repetition for a habit to form.
The research on skill acquisition is clear: new behaviours require repeated practice across multiple sessions with sufficient spacing to allow consolidation. A single day of AI training followed by a return to normal work patterns is a demonstration, not a development programme.
This is why AI Advantage runs across two full days, and why Education Excellence for schools runs across multiple sessions with deliberate spacing between them. Not because we want to sell more time — because less time simply does not work.
This one is underappreciated. Many AI training programmes use professional or enterprise tools — Microsoft Copilot, Salesforce Einstein, Adobe Firefly — because they are integrated into the software the organisation already pays for. The training makes sense in context.
But for the large majority of South African professionals, particularly in SMEs, NGOs, education and the independent sector, those tools are not available after the training room. The moment someone goes back to their desk and the subscription is not there, the practice stops. The habit never forms.
We made a deliberate decision to build every GEMIS Technical Academy programme exclusively around free tools. ChatGPT. Claude. Gemini. NotebookLM. Canva. Gamma. Perplexity. Every tool we teach with is available free, immediately, to every participant, the moment they leave the room. There is no friction between the training and the application.
AI training built around fictional companies and made-up scenarios requires participants to do an extra piece of cognitive work that is rarely talked about: translation. They have to take what they learned in a context that is not theirs and figure out how it applies to their actual work.
That translation step is where most of the value leaks out. Some participants do it naturally. Many do not, not because they are not intelligent but because they are busy and translation is an extra step that gets deferred and then dropped.
Every activity in our curriculum is built around a real scenario from the role profile being trained. A school principal works on a school communication. A compliance manager works on a POPIA policy. An entrepreneur works on a pitch deck for their actual product. The translation step disappears because the scenario is already theirs.
The participants who retain the most are the ones who used the tools on their actual work during the training — not on demonstration scenarios, not on hypotheticals. On the real thing. This is why we ask participants to bring live work to every session. Not as an afterthought. As a design principle.
There is one more factor that most training providers have not addressed: AI tools change. What you learned six months ago may be superseded. The interface has changed. A new capability has been added. A better tool has emerged. Training that was accurate when it was delivered can become misleading within months.
Our Living Curriculum Engine — the system that powers all GEMIS Technical Academy content — automatically refreshes every activity on a quarterly cycle. Not because we want to sell updates. Because training that is out of date is actively harmful. It teaches people to use tools in ways that no longer reflect how those tools actually work.
If you have been through AI training that did not stick, we would genuinely like to understand why. Not because we want to critique whoever delivered it, but because the answer usually tells us something useful about what the next training needs to do differently.
Every GEMIS Technical Academy programme is built around hands-on practice, real work scenarios, free tools and the Living Curriculum Engine that keeps content current. See the difference for yourself.