AI governance sounds like something for big companies with a compliance department. It is not. It means knowing what AI your team uses, who decides what is allowed, and who answers if it goes wrong. This guide is for UK organisations of any size and sector, and shows how to start with your team.

Not sure what AI your team already uses? Start with our guide to finding out what AI your people really use.

What AI governance means

It is three plain questions. Who decides how AI is used here? Who is accountable when something goes wrong? What do we check before and after?

That is all. It is not a policy document, a committee or a piece of software. Those can come later.

Why it matters now

Your team is already using AI, whether you have said yes or not. Our first guide shows how much of that use leaders never see. Without governance, nobody decides what goes into these tools, and nobody checks what comes out.

Who should own it

Pick one named person. In a small organisation, that is usually the owner or managing director. In a larger one, it is often the COO, the head of operations or the head of IT, with a senior sponsor behind them.

One person is enough to start. What matters is that your team knows who to ask, and that someone is accountable.

No one obvious to name? A fractional Chief AI Officer can own it part-time until you have someone.

The most important things to have in place

There is more you can do over time. These six matter most, so start here.

  1. A list of the AI tools in use. Company accounts and personal ones. You cannot govern what you cannot see.
  2. A named owner for each one. Every AI tool, every agent that acts on its own, and every process that relies on AI should have one person who answers for it. When something changes or goes wrong, everyone knows who to ask.
  3. Rules on what data can go in. Customer details, financial figures and HR matters need clear answers: yes, no, or only in this approved tool.
  4. An approver. Someone who says yes or no to a new tool or a new use, quickly, so people do not go around them.
  5. A way to review outputs. Anything AI drafts that reaches a customer or a decision should be checked by a person first.
  6. Training for your team. Short and practical: what is allowed, what is not, and how to check an answer.

The wording of the rules belongs in a policy. This guide is about who decides and what to check.

What UK law, the ICO and the NCSC expect

The UK has no single AI law for organisations. Existing law still applies, and the regulators have said so.

  • Data protection. If personal data goes into an AI tool, UK GDPR and the Data Protection Act 2018 apply. The ICO expects you to know what data is used, why, and that you have a lawful basis. Check what your tool provider does with the data.
  • Fairness. If AI helps make decisions about people, such as hiring, the ICO expects you to be able to explain those decisions and to have a person involved.
  • Security. The NCSC publishes guidance on using AI securely. In plain terms: know your suppliers, protect your accounts, and treat what you put in as something that may leave your control.

This is a plain-English summary, not legal advice. If AI touches sensitive decisions or special category data, take advice.

A simple first 30 days

  1. Week 1. Name the owner. Ask your team what AI they use, anonymously.
  2. Week 2. Agree what data can and cannot go into AI. Tell everyone, in one page.
  3. Week 3. Approve one tool for company use, name its owner, and name who approves the next.
  4. Week 4. Run a short session for your team. Agree who checks AI output that reaches customers.

Then review in three months. Governance is a habit, not a project.

Governance is one of six areas in an AI maturity assessment.

Sources