Is it working?

Six months ago the company bought AI licences for everyone and brought in someone to show staff how to use them. The sessions went well. People use it every day, and some can’t imagine going back. Then, at a leadership meeting, the CEO asks a simple question: where could we be doing better? Nobody can answer. Not because the rollout failed, but because nothing was set up to tell them.

This is the most common AI story in smaller organisations right now. They aren’t building models. They’re adopting them: AI in the tools staff use every day, personal assistants, enterprise platforms rolled out across teams. It’s a real investment, in licences, in training and in people’s time, and it usually arrives without any way of knowing what it returns.

That’s what governance is for. Is it working? Are you getting the return you expected, and where? Has it increased your exposure, through data going places it didn’t before or decisions resting on outputs nobody checks? What’s the effect on customers? How do staff feel about it: more capable, or quietly worried? And beyond that, what effect does it have on the people and communities your organisation serves?

Buyers and regulators will increasingly ask some of these questions, and meeting their expectations matters a great deal. But that isn’t the main reason to ask them. Organisations that can answer them compete better, stay safer and actually get the gains they paid for.

ISO 42001 is built around this. At its core is a loop: plan, act, monitor, review and improve (clauses 9 and 10). It asks you to decide what to measure, measure it, look at the results and change something. That loop is the part of governance that pays for itself.

Anyone who has built an AI agent will recognise the shape. An agent is a model in a loop. It acts, observes the result and adjusts, and it only gets better if it collects the right information along the way. An organisation adopting AI works the same way. If you aren’t collecting the right data as you go, about usage, quality, cost, errors and how people feel, the loop has nothing to work with, and the CEO’s question goes unanswered.

Where you start matters less than you might think. A modest beginning with a working loop beats an impressive rollout with no feedback, because within a few cycles you’ll know where you’re doing well and where you aren’t.

If you’d like help setting up that loop and getting more from what you’ve already invested, that’s what fractional AI leadership is for. If you’d like to see where you stand first, start with a readiness assessment.