W
Reviewed by Jacob Whitmore, Whito · Fact-checked for accuracy

Last Updated on November 9, 2026

To introduce AI into your business, start with one repeated job, choose one tool that fits software you already use, and set a few clear rules for staff before anyone starts. Protect customer data from day one, trial it for a fixed period, measure the time taken before and after, then decide on evidence. Everything else can wait until that first job works.

In short

  • One job, one tool, one trial period. Do not roll out five tools at once.
  • Write a short AI use policy before staff start, not after something goes wrong.
  • Keep customer personal data out of AI tools unless you have checked the provider’s terms. UK GDPR still applies.
  • Measure the time taken, including checking, before and after. Decide on what you measured.
  • AI drafts, a person approves. That rule does not change as you grow.

The mistake most businesses make

Most businesses introduce AI backwards. Someone signs up for a tool, staff start using it in different ways, and months later nobody knows what it is used for, what it costs or what data has gone into it.

That is a tactic without structure. The fix is to treat AI like any other new process: decide the job, set the rules, test it, then decide.

Step 1: pick one job

Choose a single task that is repeated, written or admin-based, and low risk if a draft is wrong. Good first jobs include:

  • Replying to common customer enquiries.
  • Writing social media posts from notes.
  • Summarising meeting notes into actions.
  • Tidying up job descriptions or quotes before a person checks them.

Avoid first jobs where an error costs money or trust: final prices, legal wording, tax figures, medical or safety advice. They can come later, with stronger checks, or not at all.

For example, imagine a three-person lettings agency. Its best first job is probably drafting replies to viewing requests, not writing tenancy terms.

Step 2: choose a tool

Start with what you already pay for. It is cheaper, staff already log in, and data stays inside an account you control.

  • If you use Microsoft 365: Microsoft says Copilot Chat is included at no extra cost with eligible Microsoft 365 plans, and its paid Copilot plans work inside Word, Excel, PowerPoint and Outlook.
  • If you use Google Workspace: Google says all Workspace plans include Gemini chat, with Gemini inside apps such as Docs on Standard plans and above.
  • If you use neither: try a free chat assistant first. Our guide to AI chatbots in the UK compares the main options.

Before staff put any work information in, check which plan you are on and what the provider does with your content. OpenAI’s pricing page, for example, says content on its Free, Go, Plus and Pro plans is used to train its models, with an opt-out available, while its help centre says it will not train on a ChatGPT Business workspace’s data. The plan matters.

Check the live pricing page before you buy anything. Prices and plan names change often.

Step 3: set simple rules for staff

Staff need to know three things: what they can use AI for, what they must never put into it, and who checks the result. Write it down. One page is enough.

A short AI use policy should cover:

  1. Approved tools. Which tools and which accounts staff may use. Personal accounts on free apps are the usual weak spot.
  2. Approved jobs. What AI may be used for, such as drafting emails or summarising notes.
  3. Banned information. What must never be entered: customer personal data, passwords, bank details, anything confidential.
  4. Human checking. Every piece of AI output is read and approved by a named person before it is used.
  5. Honesty with customers. No AI-made images or text presented as something it is not, such as a fake photo of completed work.
  6. Reporting problems. Who to tell if something goes wrong, such as data entered by mistake.
  7. Review date. When the policy will be looked at again.

A sample you can adapt is further down this page.

Step 4: protect customer data

UK GDPR (the UK General Data Protection Regulation, the law on handling personal information) applies to AI tools just as it does to email or a spreadsheet. If staff paste personal data into an AI tool, you are sharing it with that tool’s provider.

Practical steps:

  • Default to no personal data. Staff can usually get the same help by writing “the customer” instead of a name, and leaving out addresses and account details.
  • Use business accounts, not personal ones. You can see and control a business account. You cannot control a staff member’s personal app.
  • Read the provider’s terms for your plan. Check whether your content is used for training, how long it is kept, and whether you can turn that off.
  • Update your privacy notice if needed. If you do start using AI with customer information, your privacy information may need to say so.

The Information Commissioner’s Office (ICO), the UK data protection regulator, has a section on artificial intelligence, including guidance on AI and data protection and a risk toolkit. It notes the guidance is under review following the Data (Use and Access) Act, so check for updates. The ICO also runs a section of advice for small organisations.

If you work in a regulated profession, such as accountancy, law or estate agency, check your professional body’s guidance before using AI with client work. This guide is not legal advice.

Step 5: try it for a set period

Pick a fixed trial, such as four weeks. Long enough to get past the learning curve, short enough that you will actually review it.

  • Name one person to lead the trial.
  • Limit it to the chosen job and the people who do that job.
  • Use free versions or the smallest paid option. Do not buy seats for the whole team yet.
  • Keep a simple shared note of good requests (often called “prompts”) that worked, so others can reuse them.

Step 6: measure time before and after

This is the step most businesses skip, and it is the one that tells you whether AI is worth paying for.

Before the trial, time the job the normal way for a week. Write down how many times it was done and roughly how long each one took.

During the trial, time the same job with AI. Include the time spent checking and correcting the output. A draft that takes one minute to write and fifteen to fix has not helped.

A simple tracking table is enough:

What to recordBefore AIWith AI
Number of times the job was done
Average minutes per job (including checking)
Number of drafts that needed major rewriting
Errors caught before they went out
Staff view: easier, the same or harder?

Also note quality. Did customers reply as they normally would? Did any wrong information slip through?

Step 7: decide

At the end of the trial, you have three honest options:

  • Keep it. The job takes less time including checking, quality held up, and staff found it easier. Consider a business plan for the people who use it, then pick the next job.
  • Change it. It helped in places but not others. Adjust the requests, the rules or the job, and run another short trial.
  • Stop it. It did not help. That is a useful result. Cancel any paid seats and try a different job later, rather than a different tool now.

Whatever you decide, write it down with the numbers. It stops the same debate starting again in six months.

Sample AI use policy

This is a starting point for you to adapt, not legal advice. Check it against your own contracts, your privacy notice and your professional body’s guidance if you have one.

[Business name] AI use policy

Version: 1.0. Owner: [name]. Review date: [date, for example six months from today].

1. Why we have this policy

We use AI tools to help with drafting and admin. This policy sets out how staff may use them safely, so we protect our customers, our data and our reputation.

2. Approved tools

You may only use these tools, signed in with your work account: [list tools and plans]. Do not use personal accounts or other AI apps for work without asking [owner] first.

3. What you may use AI for

  • Drafting emails, letters, posts and documents.
  • Summarising notes and documents that contain no personal or confidential information.
  • Rewording text to make it clearer or shorter.
  • Generating ideas and checklists.
  • [Add or remove jobs for your business.]

4. What you must never put into an AI tool

  • Customer or staff personal data, such as names, addresses, phone numbers, email addresses, health details or account histories, unless [owner] has approved that tool for it in writing.
  • Passwords, bank details or payment card numbers.
  • Confidential business information, such as contracts, pricing agreements or anything marked confidential.
  • Information a customer or supplier has asked us to keep private.

5. Checking AI output

  • AI can produce information that sounds right but is wrong. You must read and check everything before it is used.
  • Check every fact, date, price, name and figure against a reliable source.
  • You are responsible for anything you send, publish or act on, whether or not AI helped write it.
  • Do not use AI output for final prices, legal wording, tax figures or safety advice without approval from [owner].

6. Being honest with customers

  • Do not present AI-made images as real photos of our work, staff or premises.
  • Do not use AI to write reviews, testimonials or anything that pretends to come from a customer.

7. If something goes wrong

If you enter personal or confidential information by mistake, or AI output causes a problem, tell [owner] the same day. We would rather know early. We may need to act under our data protection duties.

8. Changes to this policy

We will review this policy by the review date above, or sooner if we start using a new tool or using AI for a new job.

I have read and understood this policy. Name: ________ Signed: ________ Date: ________

The takeaway

Introducing AI is a process change, not a software purchase. One job, clear rules, a fixed trial and honest numbers. If it works, extend it. If it does not, you have lost very little.

For more practical guides, visit AI for UK business or browse our guides.

Want the working kit?

The Whito AI Starter Toolkit is a 39-page PDF with a 30-day plan, a staff AI policy to adapt, a UK GDPR checklist and 40 copy-ready prompts. £29, instant download.

Prefer to keep learning for free? Get Whito Insights, our free email on what changed in AI and what it means for UK businesses.

More from AI for UK business

Sources

author avatar
Whito
Whito is an independent UK research publisher. We publish research on how UK businesses are found, trusted and priced: what they cost to market, how their sectors work, and who actually gets recommended. Studies are free to read, every figure carries the date it was checked, and nobody pays to appear.