Why Every Business Needs an AI Policy
26 October 2026, 9:43 am
Posted in: AI
Most of the businesses I speak to are now openly embracing AI, and rightly so. They use it to draft emails, summarise documents and get through the admin faster. What some people don’t realise is how much AI they are already using without ever choosing to. It’s built into Microsoft 365 and Google Workspace, CRM systems, accounting software and video calls that write up meeting notes.
AI is everywhere, whether or not anyone has decided to use it.
That’s exactly why an AI policy isn’t just for tech companies. If AI is already part of how your business works, and it almost certainly is, you need some agreed rules on how it’s used. In most businesses those rules haven’t been written yet, so the risk goes unnoticed.
Confidential information can leave the business without anyone noticing. If someone pastes a client contract or anything covered by an NDA into a free AI tool, that information is no longer in your control. I tell people to treat the AI box like an email to a stranger. If you wouldn’t send it to someone outside the business, don’t paste it in.
You may not be on the version you think you’re on. Free and business versions of the same tool can have very different terms, and free versions are often allowed to use what you type to train their models. Few businesses have checked, and personal accounts used for work are a complete blind spot. Even with an enterprise licence, you may need to actively choose the setting that stops tools using your data for training. It’s an extra layer of protection, and an easy one to miss.
If AI gets it wrong, the mistake is still yours. If something inaccurate goes to a client under your name, “the AI did it” won’t carry much weight with that client, a court or a regulator. Some lawyers have already found this out the hard way after citing cases that didn’t exist.
Copyright is still unsettled. AI can produce content very close to someone else’s work without warning you, so anything going out externally should be checked first.
The EU AI Act reaches further than people expect. It requires people to be told when they’re dealing with AI or content it has produced, and being based in the UK doesn’t take you out of scope if you have EU clients or customers.
A good policy has to reflect how your business actually works: the tools your team uses, the data you handle, the sector you operate in and what your clients expect. That’s why a generic template, or one produced by an AI tool, rarely does the job. It may look the part, but it won’t know where your real risks sit, and a policy nobody can follow offers very little protection when something goes wrong.
It also needs to give you genuine control over which AI tools are used in your business and by whom. Clients and customers are increasingly asking suppliers exactly that, and those questions will only become more common if the UK follows the EU with its own transparency rules. It makes sense to prepare now.
And it needs to be honest about how AI output is checked. “A human will review everything” sounds responsible, but if in practice does not always happen then a policy that promises more than the business can deliver creates its own problems.
Most contracts were written before AI was part of everyday work.
If your AI provider’s tool produces something that infringes someone else’s rights, who pays? On many free and consumer terms, the answer is you. Do your client terms say anything about how you use AI and who is responsible if it gets something wrong? Are your own suppliers using AI on your data without telling you without their own robust AI policy?
As a consultant with NBRG, I work with businesses of all kinds, not just tech firms, to get this right without overcomplicating it. I can:
None of this is about stopping your team using AI. It’s about making sure you stay in control of how it’s used.
If you’re not sure where you stand, please get in touch and we can talk it through,. You can contact me on emma@nbrg.co.uk or 0800 069 9090.