AI And The Risks Of Slow Ownership

The biggest AI risk for most businesses is not slow tooling. It is slow ownership.

That sounds counterintuitive, because most AI conversations still begin with tools. Which model should we use? Should we try Codex or Claude Code? Is there an agent for that?

Those questions matter. And the experimentation matters too. In many businesses, the best AI ideas will not start in a boardroom. They will come from the people closest to the work: the salesperson trying to speed up proposals, the support team dealing with repeated questions, the marketer testing new content workflows, or the operations person quietly automating a painful admin task.

I see that curiosity as a strength. I want people trying things. I want them spotting friction. I want them to find better ways to do the work.

I’ve seen this again and again. The breakthrough rarely comes from a set-piece strategy session. It comes from exploration. It comes from irritation. It comes from someone saying, “There must be a better way to do this.” It comes from someone saying, “I spend hours every week doing this,” “customers keep asking us the same question,” or “we always lose time waiting for that document to be reviewed.”

Those moments matter because they point to the real bottlenecks. They are where AI moves from interesting to useful.

But here is where I think many businesses are about to run into trouble: the learning stays local. Sales grow a little faster, but nobody knows whether the proposal process actually improved. Support tries an assistant, but escalation rules are unclear. Marketing produces more content, but no one is asking whether quality, consistency or conversion changed. Operations builds a workaround, but it sits in one person’s head.

It is the old silo problem, familiar to all of us who have spent years working in digital transformation or trying to move a business forward as a whole. Gently, part of the consultant’s role has often been to give departments good and beneficial reasons to work together, rather than fight over budgets, ownership or credit. Except this time, the silo may be only one person deep. And that person may feel they have more to gain, and more to lose, by keeping it to themselves.

I’m also noticing something more human underneath this. Increasingly, when I talk to businesses and tease out what is really happening internally, it becomes clear that some people feel nervous about sharing how they use AI. They worry it will be seen as inauthentic, or as if they have taken a shortcut.

Others know AI is saving them time, especially when working remotely. And why would they rush to tell everyone? If the old bargain was about being seen to put the hours in, and AI is helping them get some of their life admin done, or simply giving them a little more breathing room, the incentive to share is not always obvious.

And I think that is the part many leaders need to understand. Some of the most useful AI learning inside a business may be happening quietly and invisibly. Someone has found a better way to summarise calls, draft follow-ups, prepare reports, search internal knowledge, or get through repetitive admin. They may even have built their own little system around it.

But if the culture makes people feel exposed or disadvantaged for saying so, the business never learns from it.

The result is not just scattered activity. It is hidden progress. And hidden progress is hard to improve, govern, scale or celebrate.

That is what I mean by slow ownership. I do not mean that people’s experiments are unimportant. I mean the business has not yet created a safe and useful way to absorb what is being learned, decide what matters, and turn the best experiments into better workflows.

A few weeks in, I think a founder or business leader should be able to ask: what changed for the customer? Which workflow improved? What should we stop, standardise or scale? What risks have we uncovered? Who owns the next decision?

If the answers are vague, I would not read that as a failure of curiosity. I would read it as a signal that the operating model has not caught up.

AI makes this more visible because it is unusually easy to start and unusually hard to integrate well. A team can open an account, test a prompt, automate a task or build an agent in an afternoon. That low barrier creates momentum, but it can also create sprawl.

Before long, different teams are using different tools, writing different prompts, making different risk judgments, and inventing different versions of “best practice”. Not because they are doing anything wrong, but because nobody has connected the dots yet.

At that point, the issue is no longer whether the business has access to AI. It is whether AI is being worked into the way the business actually runs.

So I think we need to be asking repeated better questions. Where is the real bottleneck? Where is customer friction showing up? Which decisions need human judgment? Which workflows can be assisted or automated? What have customer-facing teams already learned? How do we turn that learning into something repeatable?

AI does not become valuable because more people use more tools. But it also does not become valuable by shutting down experimentation. It becomes valuable when individual curiosity is connected to shared workflows, better decisions, practical guardrails and continuous learning. From my experience, the leader who takes the time and has emotional understanding to shape the culture to make that happen is the one that introduces lasting change.  

That work is less glamorous than another new tool, but it is where the real progress happens.

The question is no longer, “Are we using AI?” Most businesses are, in some form.

The better question is, “Are we learning from AI in a way that changes how the business works?”

If the answer is vague, that is the risk. Not slow tooling, but slow ownership.