My whole career trained me to start with strategy. Vision, use cases, investment plan, execute. When AI arrived on our agenda, every instinct said do the same.
We didn’t. We experimented first. And I now think that decision, more than any other, is why we’re in a strong position today.
Here’s what would have happened had we gone strategy-first, because I’ve watched it happen elsewhere. We would have produced an impressive catalogue of use cases. Most would have required IT to build or embed AI platforms. The plan would have quietly assumed things that weren’t always true: that people would embrace ownership of AI outputs to ensure they are correct. That our processes were mapped, owned and actively governed. That our data was AI-ready. That our platforms could securely integrate with large language models. We’d have quickly discovered that the AI Use Case looked good on paper, it would have made a good pilot and demo’d well, but safely scaling it into operations and realising value. Well, let’s just say a lot of time and money is being sunk by many organisations going after ‘cool AI use cases’ that are not solving high priority, real business problems, and the cost to operationalise them is way more than they first estimated.
Instead we started small and deliberate. We educated the executive and senior leaders first. We brought our people team in early, because this was always going to be a change program wearing a technology costume. We ran a pilot with fifty people, learned, then opened it to the whole organisation, learning again at each step.
What experimentation surfaced, no strategy workshop could have. We learned where trust breaks down. We learned which processes only worked because someone had been quietly compensating for years. We learned where our data quality was not quite AI-ready and where it was. We learned who our natural champions were, and they were rarely who we thought they would be.
Now we’re writing strategy, and it’s grounded. We can tell the difference between where investment makes sense, where we need to lift basic maturity first, how we need to govern.
Experimentation is not a substitute for strategy. It’s a prerequisite for a strategy that describes the organisation you actually have, rather than the one you wish you had. If your AI strategy was written before your people had real experience of using AI, it’s worth asking what it’s actually based on.