Why AI strategies fail when leaders delegate them, according to Pathfindr CEO

AI strategies often stall because the people expected to lead them are too far from the technology itself.

When AI gets handed off to a project team or parked inside the technology function, business leaders stay disconnected from where it actually creates value.

In this ADAPT Insider conversation, Dawid Naude, CEO at Pathfindr, talks about how AI behaves less like a fixed enterprise system and more like a capability leaders have to use, test, and learn from directly.

Listen to the full episode on Apple Podcasts, Spotify, and YouTube.

Key takeaways:

• Leaders need direct experience with AI if they expect it to create value across the business.
• The most useful AI applications often emerge through hands-on experimentation rather than narrow upfront use cases.
• Cost, data, and oversight still matter, but they need to be managed in ways that support adoption rather than delay it.

Leaders have to make AI their own

Dawid’s point is that AI cannot be treated like another transformation program that gets delegated and reviewed at a distance.

Even a well-written strategy can fail if the leaders expected to act on it do not feel that it belongs to them.

That is what changed in Pathfindr’s own work.

Earlier strategy engagements often produced strong recommendations but weak follow-through.

The shift came when AI stopped being the strategy itself and became an input into each leader’s own strategy.

Once leaders started using the tools directly, they could see where AI helped them think, test ideas, and solve problems in ways that felt practical rather than abstract.

AI value shows up through use

Dawid also pushes back on the way many organisations still frame AI through narrow use cases.

His view is that AI behaves more like a general capability, where the highest value often only becomes clear once people start using it seriously.

That is why he compares it to tools like Excel or the internet.

The value did not come from defining every use in advance but from putting the capability in people’s hands and seeing where it unlocked better ways of working.

In AI, that means leaders using it to prototype ideas, pressure test assumptions, and work through more complex problems than a standard business case would surface.

Cost, data, and oversight need better judgement

He is equally clear that none of this removes the need for judgement. Token costs can rise quickly, so organisations have to link usage back to value.

Data still matters, but it should improve alongside adoption rather than hold it up.

And human oversight does not disappear, it changes shape as people and AI work back and forth across a task rather than through one static checkpoint.

The challenge is to sequence all of this properly.

Organisations need enough freedom to learn where AI is useful, enough discipline to stop waste, and enough human judgement to know where oversight still matters most.