AI adoption is not a tooling problem. It is a behaviour-change problem.

A lot of companies are still treating AI adoption like a technology rollout. They give people access to tools, run some training, publish guidance, and expect behaviour to follow. However what happens in practice might not meet expectation.

What actually changes behaviour is much messier and much more human. People adopt AI when they see colleagues they trust using it in real work. They adopt it when peers ask, “How are you using it?” in normal team conversations. They adopt it when someone shares an example, not a polished success story. In other words, AI adoption looks less like a training programme and more like a culture change.

The blocker is often not access to models or lack of formal learning. It is whether people feel safe to experiment, admit confusion, compare notes, and learn in the flow of real work.

There is also a deeper issue underneath this, especially in engineering. Many developers do not only write code because that's what the job demands, they do it because they enjoy it. So if the future gets described too narrowly as “the AI writes, you review,” that is not an exciting evolution of the craft. It's a downgrade.

That is why organisations need to solve two problems at once. First, help people use AI better in their jobs right now through practical, peer-led learning. Second, give them a more motivating picture of where their value moves next which might be problem framing, steering intent, context engineering, shaping direction, owning trade-offs, and apply judgment in different ways at each stage of AI-assisted workflows.

The organisations that do this well will not just train people on AI. They will make AI use visible, normal and discussable.