Teams + Copilot + Claude Chat does not make your business AI-mature

“How was your weekend?”

“Good. I built an app for my cycling club and designed some merch for the next trip. You?”

“Something similar. Built an app for my workouts, to get the plan out of my head and into something I can use.”

“You two are so lame. I watched a movie and hung out with my kids. Booked the December holiday. Although, there was this thing Claude did when I went to book the flights…”

That was our virtual watercooler chat on Monday, and I gather it was similar to others too. It seems no social occasion is immune anymore. The conversation swings from awe at what the tools can do to something closer to panic at how fast they’re changing, and back again before the coffee/glass of wine/[insert beverage of choice here] is finished. What did we talk about before these two letters filled every silence?

And yet, for all the talk, the line I hear most from people running businesses isn’t about agents or security. It’s this: “I want to use it across the business. I know I should be. I just don’t know where to start.” At dinner on Friday, a friend offered a fix to the person asking the question: just ask the AI what it suggests you do to use AI effectively in your business. The problem is context. The tool knows nothing about how your business runs, and where there’s no context, a confident answer gets expensive. It is just a tool after all.

So why is the leap from dabbling to depending on it still so hard?

It isn’t ability. It’s access.

Access to tools. Businesses have, for good reason, locked down which models and tools their people can use. Risk and cost both argue for it. But I’m starting to wonder whether some have overcorrected, and locked things down so far that the teams who would get the most from these tools can’t reach them. If the only sanctioned use is a Copilot prompt to summarise your inbox, that’s where the use will stay.

Access to data. A company’s data is its most valuable asset: clients, deals, decisions, financials, all of it confidential. If the tools on offer don’t guarantee the privacy of that data, or nobody has explained how they do, then being terrified to put anything real into them is the rational response, not the timid one. Of everything I see keeping AI stuck at the personal level, the unanswered data question is the biggest.

Access to know-how. The everyday dabbling is easy. Using AI to run part of the business is unfamiliar and needs new habits, starting with opening Claude, ChatGPT or whichever tool as the first thing you do in the morning rather than the last resort. The pace of change doesn’t help. Keeping up with every release is impossible for anyone, and when something makes you feel behind every single day, putting your head in the sand and leaving it there is an understandable choice.

Where to start: some suggestions from our own learnings

It is possible to use AI across the business, and to change the business with it. We don’t claim to have finished. We’ve been doing it inside our own company and building it into what and how we deliver for clients, and the lesson that keeps repeating is that the hard part isn’t the technology. It’s knowing where in the business to apply it, how to make sure it is governed and what has to be done first. What follows isn’t a methodology. It’s the handful of ideas that have told us the most about where a business stands and what might be holding it back.

A view on AI maturity. This is not a score on your change programme. A plain understanding on where the business is: what AI policies and governance exist, how well they’re understood by the people expected to follow them, whether they include practical use cases and guardrails or just prohibitions, and how often anyone looks at them again. A policy nobody can find is a policy nobody follows.

A view on the data ecosystem. Which system is the golden source for which data. Who can access what, and for what purpose. What’s private, what’s public, and the awkward middle. How and where it’s stored. Most of the fear in the previous section dissolves once these questions have answers, because “can I put this in?” becomes a question someone can answer.

A view on processes. Which data should be available at which stage of your business lifecycle, how accessible it is in practice, and the question that matters more than any tool: do you trust it? How decisions get made, and how the data behind them changes. An AI pointed at numbers nobody trusts simply produces faster versions of the same result, and does not change the business.

Leadership driving it. None of the above happens because a policy says so. It happens when the people across the business use the tools themselves and ask, out loud, where else they’re being used. That is what makes it safe for everyone else to try. The businesses I see making progress have a leader who opened the tool first, not one who commissioned a strategy about it.

There are contextual pieces missing from these four, and every business will have its own. But taken together they explain most of the gap between “we’ve all got Copilot” and a business that runs differently.