AI for small business: a practical guide.
Where AI genuinely helps a small UK business, which tools to try first, the risks, and a simple plan for your first 30 days.
The short answer.
Start with one job that eats time every week. Try it with the AI already in the software you pay for, keep a person checking anything that goes to a customer, and only pay for something custom once you know it works. For most small businesses that means a general AI assistant for writing and summarising, the AI inside Microsoft 365 or Google Workspace, and simple automation between the apps you already use.
Where does AI help a small business?
Mostly in the office work around the real job: reading, writing, sorting and re-typing. In September 2026, the ONS found 30% of UK businesses used at least one kind of AI, and 29% of those with fewer than ten staff. If you haven’t started, you’re in the majority.
| Job | What AI can do | Try first | A person still |
|---|---|---|---|
| Emails and enquiries | Drafts replies, sorts a shared inbox and summarises long threads. | Copilot in Outlook or Gemini in Gmail, or a general assistant. | Reads and sends anything that goes to a customer. |
| Quotes, letters and documents | Turns rough notes into a first draft, rewrites in your tone, pulls the key points out of a contract or tender. | A general assistant such as ChatGPT, Claude, Copilot or Gemini. | Checks prices, scope and terms before anything is sent. |
| Meetings and calls | Writes the transcript, the notes and the list of actions. | The transcription and note-taking in Teams or Google Meet, depending on your plan. | Asks permission before recording and checks the actions. |
| Admin between systems | Moves details from emails and forms into your job system, accounts or spreadsheets, so nobody types them twice. | Zapier, Make or Power Automate, often with an AI step for reading the email. | Approves anything that changes money or a customer record. |
| Reports and numbers | Summarises a spreadsheet, explains what changed and drafts the weekly update. | Copilot or Gemini alongside your spreadsheet, or a general assistant given the file. | Checks the figures against the source. |
| Answers for your team | Answers questions about your own procedures, price lists and policies. | A project or knowledge feature in your assistant, loaded with your documents. | Keeps the documents current and owns the answer. |
| Marketing | Drafts posts, website copy, newsletters and image ideas. | A general assistant, plus the AI tools in Canva if you use it. | Makes sure every claim is true. |
| Scheduling and planning | Suggests options and spots clashes, but rarely decides well on its own. | Your job or scheduling software’s own features, before any AI. | Decides who goes where. |
Which AI tools should you try first?
In this order. Each step costs more effort than the last, so only move on when the one before stops being enough.
1. The AI you already have.
Microsoft 365 business plans include Copilot Chat, and Google Workspace business plans include Gemini. Signed in with a work account, both come with business data protection. Our Copilot guide covers what you may already have.
2. A general assistant.
ChatGPT, Claude, Copilot or Gemini, on a business plan rather than personal accounts. Use it for drafting, summarising and thinking a problem through. Our ChatGPT guide covers setting it up for a team.
3. AI in your other software.
Many accounts packages, CRMs and job management systems now have AI features built in. Switch them on and test them before adding another subscription.
4. Automation between apps.
Zapier, Make and Power Automate move information between the tools you use, and can add an AI step to read an email or a form. This is where the time savings usually are, and where workflow automation starts.
What we’d leave until later: a chatbot answering customers on its own, AI agents that take actions without a person approving them, and anything custom. They can be worth it, but only once you know which job they’re for.
AI won’t fix a muddled process.
If nobody agrees how a job should run, AI just does the muddle faster. Write down the steps first, fix the obvious gaps, then add AI where it takes a step off someone’s plate. That’s also why we start any project by looking at how the work actually runs.
Is it safe? The risks, and what to do.
None of these is a reason not to start. For the detail, the ICO publishes guidance on AI and data protection.
Customer and staff data
Personal accounts on free tools may use what you type to improve their models, and UK GDPR still applies to anything personal you put in.
What to doUse business plans with clear data terms, write down which tools hold what, and update your privacy notice if AI handles customer data.
Confident wrong answers
AI tools can state something false in a perfectly reasonable tone, including figures, dates and what a contract says.
What to doA person checks anything that goes to a customer, involves money, or touches safety or the law.
Decisions about people
Using AI to make significant decisions about individuals, such as hiring or an individual’s price or credit, comes with extra legal duties.
What to doKeep a person making those decisions, and read the ICO’s guidance before you automate any of them.
Staff using their own accounts
If there’s no agreed tool, people use whatever they have on their phone, with your data in it.
What to doGive the team one approved tool and a one-page policy: what can go in, what can’t, and who checks the output.
Subscriptions that pile up
Each tool looks cheap on its own; five overlapping ones, each half-used, add up.
What to doUse what you already pay for first, and add a tool only when it does a job the others can’t.
Things that break quietly
An automation that worked in spring stops when a form or a supplier’s email changes, and nobody notices for weeks.
What to doGive every automation a named owner, and check it on a schedule.
Your first 30 days with AI.
One job, four weeks, a clear decision at the end. You need a business account and one person who owns it.
- Days 1 to 7
Pick one job.
List the jobs that eat time every week. Pick one that’s repetitive, mostly text, and low-risk if it goes wrong, such as first drafts of quote emails. Write down roughly how long it takes now.
- Days 8 to 14
Set it up properly.
Use a business account, not a personal one. Write a one-page policy on what data can go in and who checks the output. Try the job with the tool you already have, and save the instructions that worked.
- Days 15 to 21
Use it for real, with checks.
One or two people use it on the real job, every time, with a person checking each result. Note what goes wrong and adjust the instructions.
- Days 22 to 30
Decide.
Compare against your starting point. Keep it, change it or stop. If it stays, write down how it’s done and show the rest of the team, then pick the next job, or look at connecting it to your other systems.
When do you need help?
Plenty of small businesses never need outside help with AI. The ONS reports the most common barriers are working out where to use it, cost and a lack of expertise.
You can probably do it yourself when
- The job lives in one tool
- One or two people do it
- An off-the-shelf tool already does most of it
- A mistake is easy to spot and cheap to fix
It’s worth getting help when
- The work crosses several systems, such as email, a job system and accounts
- It needs to run reliably without someone watching it
- It touches customers, money or compliance at volume
- Nobody in the business has time to own it
- You’ve tried tools and nothing has stuck
That’s the work we do. A Fit check looks at how your business runs and the tools you already use, and gives you a written plan and price. There’s no commitment to build. Building and ongoing support are scoped and priced separately; how the cost works explains it. See also AI implementation, AI integration, AI strategy and what an AI automation agency does.
About this guide.
Written by Jacob White, founder of Propel. Jacob runs operations at an agency and built the internal system it runs on for planning work and team capacity.
We checked the facts against the sources below on 3 November 2026. AI tools change quickly, so check the provider’s own pages before you buy, and we’ll update this guide when they change. We’re independent of the companies named here.
Questions owners ask.
More about working with us in the full list of questions.
What is the best AI for a small business?
For most small businesses, the best first tool is the AI already in the software you pay for: Copilot in Microsoft 365 or Gemini in Google Workspace. Add a general assistant such as ChatGPT or Claude on a business plan if you need more. The right tool depends on the job, so pick the job first.
How can a small business start using AI?
Pick one repetitive, low-risk job, such as drafting replies to enquiries. Use a business account, agree who checks the output, run it for real for two weeks, then compare it with how long the job took before. Keep it, change it or stop, then pick the next job.
Is it safe to put business data into AI tools?
It can be, on the right plan. OpenAI, Microsoft and Google all say their business plans don’t use your business data to train their models by default, but free personal accounts can be different. UK GDPR still applies to personal data, so check each tool’s data terms and keep a record of what goes where.
How much does AI cost a small business?
Off-the-shelf tools are usually a monthly subscription per user, and many have a free tier or trial, so check the current price on the provider’s site. A custom build is a one-off project priced on its scope, plus ongoing support if you want it. Start with what you already pay for before adding anything.
Will AI replace my staff?
In most small businesses, AI takes over parts of jobs, such as typing, sorting and first drafts, rather than whole roles. The ONS found that in September 2026, 6% of AI-using UK businesses with 10 or more staff reported a reduced headcount as a result, and administrative and clerical roles were the most affected.
Do I need an AI consultant?
Not to get started. One person can try off-the-shelf tools on one job. Help is worth it when the work crosses several systems, needs to run reliably without someone watching it, touches customers at volume, or nobody in the business has time to own it.
What’s the difference between AI and automation?
Automation follows fixed rules, such as copying a new order into your accounts. AI handles the messy parts rules can’t, such as reading an email and working out what the customer wants. Most useful setups combine the two, with a person approving anything that matters.
Where next.
- ChatGPT for businessSetting it up safely for a team.
- Microsoft Copilot for businessWhat you may already have in Microsoft 365.
- AI assistants and agentsAI that drafts, sorts and summarises, with a person checking.
- Workflow automationDetails entered once, not three times.
- AI consultancyHelp working out where AI fits your business.
- Who we helpWhere the time goes in businesses like yours.
Book a free 30-minute call.
Tell us about one job that takes more effort than it should. We’ll tell you honestly whether AI is the answer.
Pick a timeOr email jacob@propelai.co.uk