A practical guide for UK businesses. What AI is good at, what it is bad at, what you can safely put into it, and who should be in charge.
Artificial Intelligence is moving fast. Your team is probably already using ChatGPT or Copilot, whether you know it or not. The question isn’t whether to engage with AI. It’s how to do it safely and strategically.
We have been helping UK businesses through technology change since 2006, so, we understand both the opportunities and the risks. We are not here to sell you technology. We are here to help you work out what matters and what to ignore, so your business stays protected while you find out what is possible.
Whether you’re a construction firm needing to maintain Cyber Essentials compliance, an engineering company looking to streamline estimating, or any mid-sized business trying to work out where AI fits, we can help you cut through the noise and make informed decisions.
That is what our AI Manager service does.
This guide covers a lot of ground. If you’re feeling overwhelmed, start here. These two actions will get you moving safely while you work out your longer-term approach.
Both downloads are ours and both are free. If you would rather someone else kept on top of all this, that is what an AI Manager does.
Forget the Hollywood robots. For your business, Artificial Intelligence is simply a set of tools that can automate repetitive tasks, find patterns in your data, and help your team work faster on routine jobs.
What matters: Focus on specific problems AI can solve rather than trying to understand every technical detail. The value lies in what AI can do for your business, not how it works.
The most effective approach is human-AI collaboration: AI handles the tasks it does well, your team provides judgement, oversight, and expertise. AI augments your people. It doesn’t replace them.
Before picking any AI tool, ask your team: what tasks eat up time that shouldn’t? Common candidates:
You don’t understand exactly how your kettle heats water. You just know: water goes in, button pressed, hot water comes out. AI works the same way.
You don’t need to understand neural networks or machine learning algorithms. You need to understand: what goes in, what comes out, and whether the output is good enough for your purposes.
Focus on inputs and outputs. Test the results. Refine your approach based on what works. Leave the technical details to the people who build these tools.
Don’t try to transform everything at once. Pick one specific, low-risk task. Try it. Learn from it. Then expand.
Good first projects:
Save these for later:
Someone needs to own this. Not necessarily technical. They need curiosity, common sense, and protected time to experiment. Their job:
If you already pay for Microsoft 365, start with the Copilot you already have. It costs nothing extra and your information stays somewhere you control.
Most businesses do not know this exists. Microsoft 365 Copilot Chat is included at no extra cost with almost every business Microsoft 365 plan. It is proper work AI, covered by business terms, and it is already sitting in your subscription. People who assume the company has not bought anything go and use the free public version instead, which is the one arrangement that puts your work where you cannot see it.
So before anyone spends anything: find out who already has it, and tell them.
If you then want Copilot working inside Word, Excel, Outlook and Teams, that is the paid add-on, and the price depends on your size.
| What you buy | Who it is for | Per user per month |
|---|---|---|
| Microsoft 365 Copilot Chat | Everyone on a qualifying business plan | Included, free |
| Microsoft 365 Copilot Business | Up to 300 staff | £13.80 (list £16.10, paid yearly) |
| Microsoft 365 Copilot | Enterprise plans, or over 300 staff | £23.10 (paid yearly) |
The bottom two are the same product. A business under 300 people buying the enterprise one pays around £110 per person per year more than it needs to, so it is worth checking your invoice.
Prices are UK, excluding VAT, correct as at July 2026.
You do not. You give them a safe way to do the same job, and you tell them which one it is.
Your people are already using it. Not because you asked them to. Because it saves them hours. And even if you banned it today, most of them would find a way round it by tomorrow.
Here is the version that actually happens. Someone is three days from a tender deadline and does not understand a clause in the main contractor’s contract. So they paste the contract into ChatGPT and ask what it means. They get a decent answer, they get on with the job, and nobody mentions it because it did not feel like using software. It felt like using Google.
That contract is now sitting in a system you have no agreement with. Same story with a pricing summary pasted in to tidy up a covering letter, or an accident report pasted in to make it read better.
Nobody did anything wrong. They were trying to get the work out.
What to do about it, this week. Give people one approved tool, on a company account, and say out loud which one it is. If you are on Microsoft 365 you almost certainly already have one and are paying nothing extra for it. Then write down the three things that must never go into it: client contracts, pricing, and anything with a person’s name on it.
That is it. The safe route has to be the easy route, or people will keep going round it.
This isn’t scare tactics. It’s about understanding what happens to your data so you can make informed decisions.
When your team uses AI tools, data often leaves your control. Some tools use inputs to train their models, meaning confidential information could influence responses given to other users. Some store data indefinitely. Some are subject to laws in other countries.
For UK businesses, there are real UK GDPR implications. If someone pastes customer data into a public AI tool, you may be breaching data protection regulations.
Before using any AI tool, classify the data involved:
Public: Information you’d happily put on your website. Safe for any AI tool.
Internal: Business information that isn’t secret but shouldn’t be public. Use with caution. Check the tool’s data policy first.
Confidential: Client contracts, pricing, drawings, HR files, financial detail. Never put this into public AI tools. Only use AI services with proper data protection agreements.
Restricted: Passwords, bank details, trade secrets, anything you would ring a solicitor about. Never use with any AI tool.
For businesses holding Cyber Essentials or working towards ISO 27001, AI tool usage needs to fit within your existing information security policies. This is an area where proper planning upfront saves significant problems later.
That depends on the account it was signed into, not on which tool it is.
The same product behaves completely differently depending on how you reach it. A personal ChatGPT account and a company one are not the same thing. Microsoft’s free Copilot and Microsoft 365 Copilot share a name and very little else.
Verified 19 August 2026.
| Tool | Personal or free account | Company account |
|---|---|---|
| ChatGPT | Trains on what you type by default. You can turn it off in settings | Team, Enterprise and Edu do not train by default. You control retention, deleted chats go within 30 days |
| Claude | Trains by default since September 2025. Five years if you allow it, 30 days if you opt out | Claude for Work and Enterprise do not train by default |
| Google Gemini | People at Google review a sample of chats. Reviewed ones are kept up to three years, and deleting your history does not remove them | Gemini for Workspace does not train on your data and is not reviewed outside your business |
| Microsoft Copilot | The free version on a personal account is outside your business entirely | Microsoft 365 Copilot on a work account is covered by business terms and is not used for training |
| DeepSeek | Trains on what you type, processed in China | Nothing we would put a client’s work into |
Why this table has a date on it. The version that sat here for a year said Claude does not train on your data. That was true when it was written. The terms changed in September 2025 and nobody updated the page. Nothing on your screen changes when that happens, which is the entire problem.
Reviewed quarterly. Owner: Dave Lane. Last verified 19 August 2026.
Your staff are probably already using AI tools. The question is whether they’re doing it safely. A clear, simple policy protects your business and gives your team boundaries to work within.
The template covers the essentials. But as your AI usage matures, you’ll need to think about:
This is where data governance planning becomes important. The basic policy gets you started safely. Proper governance ensures you’re building on solid foundations as AI becomes more embedded in how you work.
The nine guides below go deeper than this page does, each on one question.
For most businesses, start with ChatGPT or Claude for general tasks. Both handle writing, summarising, and answering questions well. If you’re on Microsoft 365, explore M365 Copilot for integration with your existing tools and data. Start with low-risk internal tasks and build from there.
No. Modern AI tools are conversational. You type what you want in plain English. The skill is in asking clear questions and checking the results, not understanding the technology. See black box thinking.
You can start for nothing. If you are on Microsoft 365 you almost certainly already have a work Copilot included.
Beyond that, Microsoft 365 Copilot Business is £13.80 per user per month for businesses up to 300 staff, and £23.10 on enterprise plans. Prices are UK, excluding VAT, correct as at July 2026.
AI augments staff, it doesn’t replace them. The businesses getting results use AI to handle routine tasks so their people can focus on work requiring judgement, relationships, and expertise. Think of it as giving everyone a capable assistant.
It depends on the tool and how you use it. Some tools use your inputs to train their models. Others have stronger privacy protections. Check our data governance section and the data usage table. Never put confidential data into tools that don’t protect it.
If you’re processing personal data with AI tools, GDPR applies. You need a lawful basis, appropriate security measures, and potentially a Data Protection Impact Assessment. Our AI Policy template includes GDPR considerations. For complex situations, seek specialist advice.
AI tools need to fit within your existing security policies. Key considerations: data classification, approved software, access controls, and supplier assessment. If you’re certified or working towards certification, plan AI adoption carefully. This is exactly the kind of thing we help businesses navigate.
One is free and personal, one is a paid work product, and they share a name for no good reason.
The free Copilot is a general assistant like ChatGPT, and on a personal account it sits outside your business entirely. Microsoft 365 Copilot works inside your own SharePoint, Teams, OneDrive and Outlook, under your business terms. There is also Microsoft 365 Copilot Chat, which is free with most business plans and is the one most people do not realise they already have.
For the right use cases, yes. If your team spends significant time on document creation, meeting follow-ups, email management, or finding information across your organisation, it can deliver real time savings. But it needs proper setup. Your SharePoint structure, permissions, and data organisation all affect results.
If you have read this far, the next step is usually a conversation rather than another article.
If you would rather somebody else kept on top of all this, that is what an AI Manager does: one person accountable for what is approved, what data can go in, and what happens when the tools change.
If you would sooner just ask a question first, get in touch. We will answer it whether or not you ever become a client.