Your AI pilot is stalling on trust, not technology

Most AI business cases have a line for licences, a line for integration, and a line for training. Almost none have a line for trust. That omission is where a lot of pilots quietly die.

The logic is short. AI feeds on your organisation's data: its workflows, its institutional memory, its assumptions. That data exists, but most of it is not coming to you. It sits in silos, held tight by teams, parked under layers of compliance and hesitation. People do not hand over data they have not been asked to expose, because exposing the work (or the work not done) is an act of vulnerability. They only do that with a requestor they trust. No trust, no data. No data, no AI impact.

So the bottleneck in your AI programme is rarely the model. It is the same layer every transformation stalls at: people, and more precisely the behaviours of the leaders asking for the data.

Patrick Lencioni's Five Dysfunctions of a Team still maps this cleanly, even when the dysfunction now plays out in data pipelines rather than meeting rooms. Absence of trust shows up as data hoarding. Fear of conflict shows up as AI steering groups that are too polite, or quietly avoided. Lack of commitment shows up as a roadmap nobody really believes in. Avoidance of accountability gets fuzzier when outputs are probabilistic, so it has to be replaced by pride in the work. And inattention to results turns the whole effort into theatre, all vanity metrics and no business value.

Read that way, an AI initiative is a team-dynamics exercise wearing a technical costume. Behind every struggling algorithm is a series of missed conversations, and under every stalled pilot is a group that never safely agreed what success looked like.

The practical move is to make trust an explicit, named part of the programme, rather than something you hope forms quietly in the background. That is the kind of thinking worth doing before you take an AI roadmap to your team. Pragma, our AI change advisor, is built for that first-draft thinking, and you can try it free, with no sign-up. If you want to see how it fits a live programme, the frequently asked questions cover the detail.

So before the next prompt workshop, ask the quieter question: do the people holding your best data actually trust what you will do with it…?

Mark works through all five dysfunctions under an AI lens, and why trust belongs in the technology budget, in the full article: the leadership concept behind why so many AI projects underperform.