The AI Problem You're Solving Isn't the One You Have
There is a particular kind of frustration that shows up amongst executive leaders right now. The organisation has bought the licences, run the training sessions, and rolled out the new AI tools across every team that asked for them. Six months later, adoption is patchy, half the team is quietly avoiding the new system, and the leader is left wondering what went wrong with the plan.
Nothing went wrong with the plan. The plan solved a technical problem when the organisation was actually facing an adaptive one.
Ronald Heifetz drew this distinction decades before generative AI existed, and it has never been more useful. A technical problem has a known solution, even if that solution is complicated to execute. You can hire an expert, buy the right software, or write the right process, and the problem resolves. An adaptive challenge is different. It requires people to change their values, their habits, or their sense of what their role actually is. No amount of expertise from the outside can do that work for them, because the work is internal.
Rolling out an AI tool is technical. Helping a senior manager accept that the parts of their job they were proudest of are now partly automated is adaptive. Teaching someone to use a new dashboard is technical. Helping a team stop hiding their AI use out of fear that it makes them look replaceable is adaptive. The second kind of problem is the one actually stalling most AI adoption, and it is the one most rollout plans never touch.
This matters because the instinct under pressure is almost always to treat the adaptive challenge as though it were technical. More training. Better documentation. A clearer mandate from the top. These things are not wasted, but they are aimed at the wrong target, and teams can feel the mismatch even when they cannot name it. What they are actually waiting for is permission to be uncertain, and a leader willing to sit in that uncertainty alongside them rather than paper over it with a rollout schedule.
The leaders who move through this well tend to do a few things differently. They name the discomfort in the room rather than working around it. They ask what people are afraid AI might take from them, rather than assuming the fear is only about competence. And they treat their own uncertainty about the technology as something to be honest about, not something to manage away with confidence they do not actually feel.
None of this replaces the technical work. The tools still need to be rolled out, the training still needs to happen, and the systems still need to work properly. But the technical work will keep stalling until someone in the room is willing to do the adaptive work alongside it. That is leadership work, not IT work, and it cannot be delegated to a vendor.
If you are watching an AI rollout stall in your own organisation, the first useful question is not "what training are we missing." It is "what are we actually asking people to give up, and have we said that out loud yet."
