Bill Lewis Linacre Capital

What AI says.
What AI does.
What AI builds.
Who controls it?

AI is talking about your business when you are not in the room.

AI is running inside your business in ways you don't know.

AI is building your products while your team learns it as it goes.

Who is watching?

These are not three separate problems.

They are three glimpses of one shift. AI has become an operating layer that impacts your business — judging it from the outside, acting inside its systems, shaping its work, building parts of what comes next.

Many would say the explosion of AI is outpacing its governance. Accountability for AI layers that cannot be seen?

Is what you feel about AI right now an unease you cannot name?

This is board-level AI visibility and control: what AI says, what AI does, what AI builds, and who controls it.

The verdict that arrived before you did

Someone is deciding about you. A buyer weighing a contract. A partner considering the deal. A client about to place the commission. An investor running the numbers. An acquirer taking early eyes over you. A candidate choosing whether to join.

Before the meeting that matters, they ask AI about you and your company. It searches whatever it can reach, weighs what it finds, and hands them a verdict. The record is thin. The story is inconsistent. Or it finds almost nothing at all.

They read it. They quietly close the file. Nobody tells you.

You assume the timing was wrong. The budget moved. They went with someone else. You move on, never knowing the decision was shaped in a room you were never in.

The agents no one approved

Now look inside your own business.

AI rarely arrives as a board decision. It arrives in a software upgrade. It grows inside workflows already trusted. It appears in tools your people use every day.

Then, one day, the question occurs to someone.

What can it reach? What is it doing? Who switched it on? And who answers if it gets something wrong?

Someone will tell you they have it covered. Ask them to count. The agents multiply faster than anyone keeps track of — across suites, sites, no-code builders, partner tools, vendor updates that arrived without a decision.

No one can show you the whole picture. That is the exposure.

The build that runs in your own hands

And closer still.

You used AI to build the thing. The code ran. The demo worked. Then the trouble started.

One fix created another problem. One session forgot what the last one knew. One confident answer broke something that had taken weeks to get right.

AI can write the code. AI cannot govern itself.

The whole of it — the design, the code, the decisions — runs with no hand on the wheel. It works, so no one asks whether it was controlled. The reckoning comes later, in ways no one can trace back.

Someone has to take charge of the build. That someone is you. The only question is whether you know how.

Those are three rooms. Do not mistake them for categories. They are examples — three places the shift becomes visible. There are many more.

AI is now at work wherever it has been let in — and it has been let in almost everywhere. In finance. In hiring. In how your customers are answered. In what your suppliers are told. In the papers that reach your own board.

Whatever you run, in whatever sector, AI is already saying something, doing something, or building something inside it.

You cannot see most of it. Not because you are careless. Because it acts where you are not, at a speed you cannot match, and it leaves no note on your desk.

So the real question is not which of these three is yours.

It is who, in your business, can see across all of it — and answer for it. For most organisations, the honest answer is no one. That is not an IT gap. It is a board-level blind spot, and it is widening every month.

I am not selling you the thing I am warning you about

Most people who will talk to you about AI have something to sell — a platform, a transformation, a scan, a subscription. That is precisely why they will not point you to where it costs you.

I have none of that to sell. Which is why I can tell you where it will hurt.

I saw the agent problem before it had a name

In late 2025 I watched AI agents arrive inside enterprise software while most boards had no idea they were there.

I wrote what was missing — The Emerging Risk of AI Agent Networks, a board-level risk assessment published under Linacre Capital in February 2026, built around the one thing most organisations still cannot produce: a reliable count of their own agents.

Ten weeks later Gartner gave it a market name: agent sprawl. What I had been warning about was now recognised as enterprise risk.

I am inside this, not commenting on it

I am building with AI now — a complex, real-time AI platform of my own (Voxa), using a methodology I had to develop to do it (the Conductor Protocol). I know where the build turns dangerous because I have been in it.

I have also spent a career where technology, capital and board accountability meet. I know what it looks like when a tool quietly becomes a liability — and how late that is usually seen.

Today I advise Chairs, CEOs and Founders on AI where it meets board accountability — what it is saying about you, what it is doing inside your business, and what you are building with it.

One conversation.

No theatre. No pitch. No transformation programme you did not ask for.

Three questions to begin with.

What is AI saying about your business?

What is it doing inside it?

What is it building for you — and who is in charge of that?

You will leave the conversation knowing three things, or more. What AI is saying about your business right now. What is operating inside it that no one has yet counted. And — if you are building with AI yourself — whether what you are building is being governed, or just running.

Just the questions your business should already have asked.

Who controls it? That is the test. Can you answer it — or only assume someone has it covered? Most leaders have never asked.

Every week the agents multiply. Every quarter another deal closes in a room you were not in. Every month another decision is made about you, by AI, without you. None of this is theoretical. Have the conversation now. The cost of waiting is not paid by AI. It is paid by you.