Four questions to stop AI pilots multiplying without a plan

AI leaders should choose projects by the business outcomes they can deliver, not by whichever model or platform is making the most noise. Patrick Pugh, a PwC executive, offers a practical four-question test for deciding what to scale, change or stop.

Watch Desk analysis

What happened

In a Forbes Technology Council essay published on 19 October, Pugh argues that companies can spread money, staff and attention too thinly by turning every promising AI idea into another pilot. His proposed filter is straightforward: does it solve a business problem that matters; can existing technology meet the need; is the organisation ready to scale it if it works; and what evidence would trigger a bigger commitment, a change of course or a stop?

Pugh writes that leaders should connect investments to desired outcomes and consider how each choice fits their data, workflows and existing technology partners. The aim is not to predict which technology wins, but to decide what is worth backing while leaving room to adapt.

Why it matters

The advice gives organisations a way to make AI experimentation answerable to a real decision. A pilot is useful if it helps establish whether to invest, change direction or walk away. Without that next step, “learning” can become a tidy label for a growing pile of projects competing for the same people and budget.

The four questions are Pugh’s recommendations, not evidence that one approach guarantees better returns. But they make the choice more concrete: assess the problem, what is already available, the ability to scale and the conditions for stopping.

Our read

This is a useful antidote to the idea that every new model deserves its own pilot and a small committee. Pugh’s strongest point is that leaders should define what success and failure look like before a project gathers momentum. Start with the business problem, then ask whether a new AI tool is actually needed. Novelty is not a strategy, even when it arrives with a very polished demo.

What to watch

• Whether organisations define measurable outcomes before starting AI pilots.
• How often existing systems can meet the need without another vendor or tool.
• Whether teams have clear criteria for scaling, changing or stopping a project.

Discussion spark: Should companies require every AI pilot to have pre-agreed scale-or-stop criteria, or would that risk shutting down useful experiments too early?

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.