AI initiatives don’t fail because of strategy. They fail in execution.

AI initiatives don’t fail because of strategy. They fail in execution.

In recent months, I’ve seen genuinely impressive AI strategies, even in the most conservative organisations. Dozens of slides, dense with market analysis, potential use cases and projected revenue uplifts. Nothing fundamentally wrong. Nothing obviously flawed.

But what I’ve seen far less often (maybe never): working products. AI ventures that made it past the workshop room. Initiatives that created real value, not just theoretical potential.

Execution breaks down, not because of one big problem, but because of many small ones. No decisions get made, stakeholders block each other, IT is hesitant, budgets vanish mid-project, technical skills are missing. Sometimes all of that, all at once.

So where does it go wrong? The default explanation tends to be: the people inside the organization. Supposedly, they resist change, cling to old processes, don’t want to let go. That may be true in some cases, but it’s rarely the core issue.

The actual problem often lies with those brought in to fix it. Consultants. Coaches. Experts. People who are excellent at presenting solutions but have never built anything themselves.

How is a management consultant supposed to help launch AI products when she’s never built anything that doesn’t start with “MS”? How is the agile coach going to set up new structures if he’s never worked in a real product team? And how can anyone demand innovation when the most disruptive moment in their career was switching from PowerPoint to Prezi?

If everyone’s giving talks, running workshops, or building slides – who’s doing the work?

AI projects don’t need more theory. They need people who’ve been in the engine room. People who’ve coded, designed, launched... and failed. Who’ve worked with legacy systems and still got something out the door.

In short: we need builders.
Not advisors. Not thought leaders. Not keynote speakers.
Builders.

Because AI doesn’t get implemented on slides.
It gets implemented by doing.