Who is still in control?

Every week brings a new headline. AI now writes emails, screens applicants, answers customers, and drafts forecasts. Somewhere between the first cautious pilot and the rollout across the whole company, a quieter question tends to surface at the back of the room.

Who is actually running this company now?

It sounds like fear. We think it is the most useful question a business owner can ask right now, because, unlike the headlines suggest, the answer is not fixed. Control over AI is not something you simply have or lose. It is something you design. Decision by decision, you decide how much of it stays in your hands.

Control over AI is not something you have or lose. It is something you design.

The false choice

Two stories dominate the conversation, and they contradict each other. The first is adopt or die. Move fast, automate everything, or watch a competitor who did eat your lunch. The second is that the machines are taking over. Hold the line, keep people in charge, resist the black box before it swallows your judgement.

Both stories share a hidden assumption: that control is binary. Either you are in charge or the algorithm is. For a smaller business, that framing is not just wrong. It is expensive. Fear of losing control leads companies to reject tools that would genuinely help them. Fear of being left behind leads them to adopt tools they do not understand, cannot audit and did not choose on purpose.

Neither is control. Both are reactions.

Control is distributed, not surrendered

In practice, control over AI in your business is not one lever. It is spread across a handful of concrete, answerable choices, and each one is yours to make.

Five dimensions of control

1. Your data. Do you know where it goes when your team pastes it into a tool? Whose servers, whose jurisdiction, whose terms?
2. Your decisions. Which calls does the AI get to make, and which ones does it only get to inform? Where is the person, and are they actually looking?
3. Your dependencies. If your key vendor changed their price or their model tomorrow, how trapped would you be?
4. Your people. Have you led your team through this change, or handed them a login and hoped?
5. Your obligations. Do you know what the law now requires of you, before a regulator explains it to you?

Notice that none of these is a question about the technology. Every one is a question about how you choose to run your business. That is the whole point. The company that can answer these clearly is in control, whatever tools it uses. The company that cannot is not in control, however cautious it has been.

Why this question, and why now

There is a reason to have this conversation now rather than next year. Since 2 August 2026, the transparency obligations of the EU AI Act, set out in Article 50, have been in force (European Commission, 2026). If your business puts AI in front of people, through a chatbot or through content it generates, you are now expected to say so plainly. These duties reach any organisation serving users in the EU, and failing to meet them carries real penalties (Stibbe, 2026). This is not a distant problem for large technology firms. It is a present obligation for ordinary Dutch SMB businesses that quietly added an AI tool over the past year.

Chatbots must identify themselves, and AI generated content must be marked as such. Penalties reach up to 15 million euros, or 3 per cent of worldwide annual turnover, whichever is higher (Stibbe, 2026).

Regulation is only the most visible marker. The deeper one is competitive. The businesses pulling ahead with AI are not the ones adopting it fastest or resisting it hardest. They are the ones adopting it deliberately, who chose where their data lives, kept a person on the decisions that matter, avoided lock in, brought their team with them and did the compliance work before it became urgent.

Being in control is not the opposite of adoption. It is the discipline that makes adoption safe.

What comes next

Over the coming weeks, this series takes each dimension of control in turn and makes it practical. Not should you worry about your data, but how to know where it is and what to do about it.

1. Your data. Sovereignty in plain terms, and how to find out where yours actually lives.
2. Your decisions. Keeping a real human in the loop, not a rubber stamp.
3. Your dependencies. Building optionality so no single vendor owns your future.
4. Your people. Why adoption, not software, is the real bottleneck.
5. Your obligations. What the EU AI Act asks of you now, and how to meet it without panic.

The goal is not to make you more afraid of AI, or more reckless with it. It is to put the question back where it belongs.

Who is still in control? You are, if you decide to be.

Next in the series: Your data, where it goes, who can see it, and how to get it back.