AI Tools for UK Business: The Framework for Choosing What's Right for You
Chris Duffy
Certified Chief AI Officer • Updated • 7 Min Read
What AI tools should a UK small business actually use?
For most UK small businesses, a short stack beats a long one: one strong general assistant (we use Claude) for writing, analysis and research; a meeting note-taker; something for your existing admin and accounts; and one tool for whatever your core bottleneck is. Start there. Most businesses buy too many tools and use a fraction of them.
Read next: Book a call · All guides for UK SMEs
This guide covers the four general assistants most UK businesses weigh up (Claude, ChatGPT, Gemini and Microsoft Copilot), how to connect one to your own data, when a custom build is worth it, and a 90-day plan. For the head-to-heads, read Claude vs ChatGPT and Claude vs Copilot.
The short stack, at a glance
| The job | What to use | How to choose |
|---|---|---|
| Writing, analysis and research | One general assistant: Claude, ChatGPT, Gemini or Copilot | Pick by the work you do most (see the next table) |
| Meeting notes | A note-taker that works with the video-call software you use | Check where recordings and transcripts are stored before you switch it on |
| Admin and accounts | The AI features in the software you already pay for | Try these before you buy a separate tool |
| Your biggest bottleneck | One tool, or one build, aimed at that job | Time the job now, so you can tell whether it worked |
Claude, ChatGPT, Gemini or Copilot?
All four handle everyday writing and quick questions. They differ in where they sit and in the work they handle best.
| Assistant | Pick it if | Head-to-head |
|---|---|---|
| Claude | Your work is long documents, analysis and writing that goes out in your voice, or a job you repeat every week. Projects hold your business context, and skills turn a repeated job into a set routine. | vs ChatGPT · vs Copilot |
| ChatGPT | You need images, the widest range of third-party add-ons, or a big team of light users asking quick questions. | Claude vs ChatGPT |
| Microsoft Copilot | Your work lives in Outlook, Teams and SharePoint. Copilot can see your mail, files and calendar without being handed them. | Claude vs Copilot |
| Gemini | Your business runs on Google Workspace. It deserves a trial alongside the others. |
We have left prices and plan limits out of this table. They change often, so check each provider's pricing page before you buy. We take no commission on licences of any kind.
Our recommendation: pick by the work you do most, then run one tool properly for 90 days before you add a second. Many businesses end up with two tools for different jobs.
Claude or ChatGPT: the short version
For most UK small businesses doing written and analytical work, Claude is the better daily tool. ChatGPT is stronger on images and fine for quick everyday questions. What separates two businesses using the same tool is whether anyone set it up around how the business runs and trained the people using it. The full task-by-task comparison, and how to switch from ChatGPT without starting over, is on Claude vs ChatGPT.
Switching from ChatGPT?
Don’t want to start from a blank page? Start from our foundations.
Moving to Claude usually means rebuilding your prompts and habits from a blank box. The Founders plugin skips that. It arrives built around how a small business runs: your context, your voice, and your weekly jobs on a timetable. You run /start and it walks you through one question at a time, then you get 2 hours with a trained practitioner. For a business of one or two people, £2,500 exc VAT (as at November 2026).
Connecting an AI assistant to your business data
An assistant gets more useful once it can read the systems your work sits in: email, files, your CRM, your accounts. Claude and ChatGPT both connect to common business apps through connectors, so check the ones you use are on the list. Copilot already sees your Microsoft 365 mail, files and calendar. Automation platforms such as Make.com and Zapier join apps that have no direct connection, and suit simple, rule-based hand-offs between systems.
Set the rules first, then switch the connections on. The rules say which data can go into which tool, who checks the output, and what the assistant may never do on its own, such as send an email. On our Claude plugin, governance is what switches web access and connectors on: the plugin moves to Governed once your AI Manifesto, DPIA screening and AI Policy are in. How AI governance switches connectors on.
Off the shelf, set up for you, or built from scratch?
| Route | What it means | Choose it when |
|---|---|---|
| Off the shelf | A business plan of a general assistant, used as it comes | The work is standard: drafting, summarising, research |
| Set up around your business | The same assistant, configured with your context, your rules and skills for your repeated jobs, and connected to your systems | Several people do the same jobs and you want them done the same way. Most UK SMEs need this one. |
| Custom build | Software built for one job, on your data | A job runs at high volume to a fixed standard, or the data cannot leave your own systems |
Commissioning custom software and configuring an existing platform around your processes often get sold as the same product. For most businesses of this size the second is what they need, and it costs much less. Bespoke AI solutions for business, and what the word should actually mean →
Decide in five questions
| Question | Off the shelf | Set up | Custom build |
|---|---|---|---|
| Is the job standard, or unique to you? | Standard writing and analysis | Your own jobs, done the same way by several people | Unique logic at high volume |
| Can the data leave your systems? | Yes, under a business plan's terms | Yes, with rules on which data goes where | Must stay in your systems |
| Does it need to talk to other systems? | Stands alone | A few: email, files, CRM | Old systems with no modern connection |
| Who will run it? | Anyone | Your team, after training | Developers |
| How soon do you need results? | Days | Weeks | Months |
Set the rules before you pick the tool
The common failure is choosing a tool, rolling it out, then finding there are no rules for what data goes in, who checks what comes out, or how anyone will know it is working. Answer four questions first, in a one-page AI Manifesto:
- What data must never go into an external AI tool? Client personal data, financials and trade secrets are the usual red lines.
- Who checks AI output before it leaves the business?
- Who owns AI in the business: the budget, the risk and the results?
- Which two or three numbers will tell you it is working? Hours saved on a named job is the easiest to track.
We write those rules with you as part of AI governance for UK SMEs: an AI Manifesto, DPIA screening and an AI Policy, from £3,250 exc VAT (as at November 2026).
Two real examples
Set up around the business
A UK luxury travel company
A luxury travel company of about 20 staff runs Claude on a Team plan, with connectors to Microsoft 365. Staff agreed and signed the AI Manifesto and AI Policy before the wider rollout. Claude never sends email or books anything, and a person signs off anything that goes to a client.
17 of its 19 seats (89%) used Claude every week over the first 90 days. More case studies
Custom build
Harts of Stur
Harts of Stur, a premium kitchenware retailer in Dorset, adds 6,000 to 7,000 new products a year, and each listing took 45 to 60 minutes to write. The job needed a separate brand voice for each product category and had to work from inputs ranging from a single photo to a long technical specification.
At that volume we built a multi-agent system for that one job, with the editorial team signing off each launch run. Time per listing fell by 94%. Read the Harts of Stur case study
A 90-day starter plan
Pick the one job that costs the most hours and time it. Write your one-page AI Manifesto.
Choose one assistant for that job, using the table above, and start with one team.
Train the team on real jobs from their own week. Turn the jobs they repeat into skills. Log the hours saved each week.
Compare the hours with your starting figure. If it worked, add the next team or the next connection. If it did not, find out why before you spend more.
Chosen Claude? Get it working on your own work.
For a team: a Claude training day for a room of up to 10 is £2,500, plugin included. In the afternoon each person builds a skill for the job they do most. See Claude training for teams
For a business of one or two people: the Founders plugin gives you best-practice foundations, then two hours with a trained practitioner, £2,500. See the Founders plugin
Before you connect Claude to your data: governance from £3,250 sets your rules and switches the plugin to Governed. See AI governance
Prices exclude VAT and are as at November 2026. An organisation-wide set-up is available separately.
See prices for your team size