AI use among UK firms is now common, but AI agents are not. The Office for National Statistics reported that in September 2026, 30% of UK businesses used at least one type of AI technology: 29% of businesses with fewer than ten employees and 51% of those with 250 or more. The government's Artificial intelligence adoption research 2026 found that only 2% of businesses with five or more employees were currently using agentic AI. Most small firms are still using chat assistants.
The next step, and the one owners now ask about, is agents: AI that does a piece of work end to end, using your own tools. The examples below suit small businesses because they are bounded, measurable and easy to supervise. None needs a large IT team, and each comes with the limit we would set on day one.
- About a third of UK businesses now use AI, but only 2% use agentic AI, so most small firms are starting from scratch.
- The best first agents handle repetitive, rules-based work with a clear start and finish: enquiries, bookings, calls, quotes, invoice chasing and inbox triage.
- Agents should connect to the systems you already use, not replace them.
- Keep a person approving refunds, payments, discounts and anything sent in your name until results are proven.
- UK GDPR and PECR still apply, including specific consent for automated marketing calls and safeguards for automated decisions about people.
What is an AI agent, in small business terms?
An AI agent is software that takes a goal, such as "reply to this enquiry" or "get this invoice paid", and works through the steps using tools you give it: your inbox, calendar, CRM, accounts package or shared drive. Unlike a basic chatbot, it can look things up and take actions. Unlike traditional automation, it can read messy, human input and decide what kind of request it is.
For most small businesses the safest design is the one set out in what is agentic AI: a workflow with fixed steps, where the AI does the reading, sorting and drafting, and a person signs off the parts that matter.
How many UK small businesses use AI?
Roughly three in ten, and the smaller the firm, the lower the figure. The government's 2026 adoption research, based on 3,503 telephone interviews, found that 32% of UK businesses with five or more employees used at least one AI technology, double the 16% recorded in 2025. Adoption rose with size, from 27% of micro businesses to 52% of large ones.
Base: 3,503 interviews with UK businesses with 5 or more employees, surveyed January to April 2026
| Business size | Using AI |
|---|---|
| 5 to 9 employees | |
| 10 to 49 employees | |
| 50 to 249 employees | |
| 250 or more employees | |
| All businesses surveyed |
Source: GOV.UK, Artificial intelligence adoption research 2026
Surveys differ because they ask different firms. The British Chambers of Commerce found that 54% of the firms it surveyed in 2026 were actively using AI, around 94% of them SMEs, while only one in ten SMEs was adopting deeper, bespoke AI. The bigger picture is that most UK businesses are very small: the official business population estimates count 5.6 million private sector businesses, and three quarters have no employees beyond the owners.
- 5.6 million private sector businesses in the UK at the start of 2026
- 74% of them do not employ anyone aside from the owners
- 99.85% are small or medium-sized enterprises
- 2% of businesses with five or more staff currently use agentic AI
Sources: GOV.UK, Business population estimates 2026; GOV.UK, Artificial intelligence adoption research 2026
What are good examples of AI agents for small businesses?
The strongest examples take one repetitive job off a person's desk and connect to systems you already pay for. These are the agent types we would consider first for a typical UK small business, with the guard rail that keeps each one safe:
| Agent | What it does | Systems it connects to | Sensible limit |
|---|---|---|---|
| Customer service agent | Answers common questions about opening hours, prices, delivery and policies, and logs the conversation | Website chat, email, your FAQ and policy documents | Hands complaints, refunds and anything it is unsure about to a person |
| Enquiry qualifier | Reads new enquiries, asks follow-up questions, scores fit and creates the lead | Website forms, email, CRM | Never promises prices or dates that are not in its approved information |
| Booking agent | Offers available slots, books, reschedules and sends reminders | Calendar or booking system, email or SMS | Follows your booking rules; escalates double bookings and special requests |
| Phone agent | Answers inbound calls, takes messages and books callbacks out of hours | Phone line, calendar, CRM | Passes urgent or upset callers to a person straight away |
| Quote drafter | Turns a job description or call notes into a draft quote from your price list | CRM, price list, document templates | Drafts only; a person approves every quote before it is sent |
| Invoice chaser | Checks overdue invoices and sends polite, escalating reminders | Accounts software, email | Stops on any dispute and never changes amounts or terms |
| Inbox triage agent | Sorts shared inboxes into sales, support, suppliers and admin, and drafts replies | Email, CRM | Drafts rather than sends until accuracy has been checked |
| Internal knowledge agent | Answers staff questions from your procedures, manuals and past jobs | Shared drive, wiki, document store | Read-only, limited to documents staff are allowed to see |
Most of these sit on top of a CRM, so the quality of your customer records matters as much as the AI. Our AI CRM solutions page covers that layer, and AI process automation covers the back-office jobs such as invoice chasing and reporting.
Which AI agents suit dentists, accountants, trades and other UK sectors?
The right first agent depends on where the hours go: phones and bookings for clinics, document chasing for accountants, quotes and missed calls for trades. Adoption also varies a lot by sector. In the government's 2026 research, 61% of information and communication businesses had adopted AI, against 25% in construction and 23% in hotels and catering, sectors where firms were more likely to see AI as irrelevant.
| Sector | Using AI |
|---|---|
| Information and communication | |
| Finance | |
| Business services | |
| Arts and other services | |
| Construction | |
| Retail and distribution | |
| Transport and storage | |
| Hotels and catering |
Source: GOV.UK, Artificial intelligence adoption research 2026
| Sector | Good first agent | What it connects to | Keep a person on |
|---|---|---|---|
| Dental and healthcare | Phone and booking agent for appointments, recalls and reminders | Practice management system, phone line, SMS | Clinical questions and anything involving health records |
| Accountancy and finance | Document chaser that requests missing records before deadlines | Practice software, client portal, email | Advice, tax positions and anything filed with HMRC |
| Trades and local services | Missed-call and enquiry agent that books site visits and drafts quotes | Phone line, calendar, job management software | Final prices and job scope |
| Restaurants and hospitality | Booking and FAQ agent for tables, allergens policy and opening times | Booking system, website chat | Allergy questions, complaints and large group bookings |
| Property and estate agents | Viewing booker and applicant qualifier | Property CRM, calendar, portal enquiries | Offers, referencing decisions and tenant selection |
Dental practices are a good illustration of why the limits matter: appointment admin is repetitive, but patient health information is special category data under UK GDPR, so the agent should see only what it needs. Practices that want bookings, reminders and patient records in one place can look at Circlemount Dental, our practice management platform.
How do AI agents help with customer service?
They answer the questions that repeat, at any hour, from your own approved information, and pass everything else to a person with the conversation attached. Customer service is a common target: 44% of UK businesses using or planning to use AI named it as an area, behind administration at 81%.
Base: businesses using or planning to use AI, 2026
| Business area | Share |
|---|---|
| Administration | |
| Marketing | |
| IT | |
| Operations | |
| Sales | |
| Customer service | |
| Finance |
Source: GOV.UK, Artificial intelligence adoption research 2026
Two design choices make the difference. First, give the agent a narrow, well-maintained knowledge base; out-of-date policies produce confident wrong answers. Second, decide in advance which topics always go to a person: complaints, refunds, anything legal or medical, and vulnerable customers. OpenAI's practical guide to building agents calls human intervention "a critical safeguard" and lists cancelling orders, authorising large refunds and making payments as high-risk actions. Our AI chatbot service is built on the same rule.
Can an AI agent answer the phone?
Yes. Voice agents can answer inbound calls, take messages, answer common questions and book appointments, which suits clinics and trades that miss calls while busy. Inbound calls are the straightforward case. Outbound calls and messages are where the Privacy and Electronic Communications Regulations (PECR) apply, and the ICO's telephone marketing guidance is explicit that "general consent for marketing, or even consent for live calls, is not enough" for automated calls.
| Type of contact | What PECR requires |
|---|---|
| Live outbound marketing calls | Do not call numbers on the TPS or CTPS without specific consent; say who is calling and display your number |
| Automated marketing calls | Specific consent to automated calls from you; each call must give your name and a contact address or freephone number |
| Marketing texts and emails to individuals | Specific consent, or the soft opt-in for existing customers |
Sources: ICO, Telephone marketing; ICO, Electronic mail marketing
Take advice before any outbound AI calling campaign. For inbound work, our AI phone agents page explains how calls hand over to staff.
How much do UK small businesses spend on AI?
Less than many owners expect. In the government's 2026 research, the mean spend on AI-related activity in 2025 among businesses using AI was £8,000, rising from £3,500 for micro businesses to £25,000 for large ones. These are averages across all AI activity, not the price of a single agent, and most small firms start with off-the-shelf tools.
Mean spend on AI-related activity, calendar year 2025
| Business size | Mean spend |
|---|---|
| Micro (5 to 9 staff) | |
| Small (10 to 49 staff) | |
| Medium (50 to 249 staff) | |
| Large (250 or more staff) |
Source: GOV.UK, Artificial intelligence adoption research 2026
Cost is rarely the main obstacle. The same research found limited AI skills affected 67% of adopters, and 65% of non-adopters cited the lack of a clear business case. That is why the first agent should target one task you can measure.
Which tasks should not be given to an AI agent?
Anything irreversible, regulated or with a significant effect on a person should stay with a human, at least at first:
- Decisions with significant effects on people, such as rejecting a job applicant, a tenant or a credit request. Under UK GDPR, as amended by the Data (Use and Access) Act 2025, solely automated decisions like these need safeguards: the ICO's draft guidance lists information about the decision, the chance to make representations, human intervention and the right to contest it.
- Anything irreversible without a check, such as payments, refunds, cancellations and deleting records.
- Regulated advice, including financial, legal and medical advice.
- Tasks without a clear right answer that nobody can check quickly.
For these, an agent can still prepare the work, gathering documents, summarising and drafting, while a person decides.
How should a small business start with AI agents?
Start with one agent, one process and one measure of success, and run it in draft mode before it acts alone.
- Pick one repetitive, rules-based task
- Write down the rules and exceptions
- Connect only the systems it needs
- Run in draft mode for two to four weeks
- Measure time saved and errors
- Record the personal data it uses
First replies to enquiries are a common first task. Write down what a good member of staff does today, including the exceptions. Keep access tight: read access to the calendar is not write access to the accounts, and the Model Context Protocol is a common way to connect agents to your systems with those limits. Have a person approve every output at first, then measure response times, hours saved and errors before deciding what can run without approval. Finally, update your privacy information to cover what the agent does.
If you have developers, the same approach applies to engineering work, where coding agents are now the most mature category; see Claude Code for business.
What do UK data protection rules mean for AI agents?
UK GDPR applies to any personal data an agent handles, just as it does to staff. In practice: use the minimum data needed, tell people in your privacy information how you use AI, keep records secure, and make sure people can reach a human. The Data (Use and Access) Act 2025 broadened the lawful bases for significant automated decisions, provided safeguards are in place, but special category data such as health information remains more restricted. The ICO's guidance on AI and data protection says it is "under review and may be subject to change", so check it before you design anything that decides about people.
Where can UK small businesses get help with AI agents?
Start with free training for yourself and your team, then bring in a specialist once an agent needs to touch customer data or connect several systems. Under the government's AI Skills Hub programme, every adult in the UK can take free, benchmarked AI courses, with a target that includes at least 2 million SME employees.
Location matters less than it used to: the 2026 research found no significant regional differences in adoption, with the West Midlands at 30% against 32% for the UK. Most agent work can be done remotely, but mapping a process is easier in the room. Circlemount is based at St James House in Telford, Shropshire, and our Birmingham, Manchester and London pages explain how we work in each area. When you are ready to build, our AI agent development team designs agents around the systems you already use.
Frequently asked questions
What is an example of an AI agent for a small business?
A common example is an enquiry agent that reads new website enquiries, asks the customer a few follow-up questions, checks availability in your calendar, creates the lead in your CRM and drafts a reply for a person to approve. It saves time on every enquiry while keeping you in control of what is sent.
How many UK small businesses use AI agents?
Very few so far. Government research published in October 2026 found that 32 per cent of UK businesses with five or more employees used some form of AI, but only 2 per cent were currently using agentic AI. Most small firms use chat assistants and AI features built into software they already have.
Can AI agents replace customer service staff?
For most small businesses, no. AI agents handle the repetitive questions well, such as opening hours, delivery and booking changes, which frees staff for complaints, sales and complex problems. Customers should always be able to reach a person, and anything involving refunds, disputes or vulnerable customers should go to your team.
Are AI agents legal to use in the UK?
Yes, but the usual rules apply. UK GDPR covers any personal data the agent handles, so you need a lawful basis, clear privacy information and good security. Significant automated decisions about people need safeguards, including human intervention. Automated marketing calls need specific consent under PECR, as the ICO explains in its guidance.
What systems can an AI agent connect to?
Most agents connect to the tools a small business already uses: email, calendars and booking systems, CRMs, accounts software, shared drives, phone lines and website chat. Connections are made through each system's API, an automation platform or an MCP server. The safest setups give the agent only the access it needs for its task.
How do I know if an AI agent is working?
Decide on the measure before you start, such as time to first reply, hours saved each week or errors per hundred tasks. Run the agent in draft mode with a person approving its work for a few weeks, compare against how the task was done before, then widen what it can do on its own.
Do AI agents for small businesses need custom development?
Not always. Simple agents can be built on automation platforms or the AI features already inside tools you use. Custom development makes sense when the agent must follow your specific rules, connect several systems, handle personal data carefully or keep an audit trail, which is common once an agent touches customers.
Sources
Checked on 16 November 2026. Product features and figures change quickly, so check the current source before you rely on a detail.
- GOV.UK: Artificial intelligence adoption research 2026 (full report, October 2026)
- GOV.UK: AI adoption survey report 2026, executive summary
- ONS: Business insights and impact on the UK economy, 8 October 2026
- GOV.UK: Business population estimates for the UK and regions 2026
- British Chambers of Commerce: Half of SMEs using AI, March 2026
- ICO: Telephone marketing (PECR)
- ICO: Electronic mail marketing (PECR)
- ICO: What are the ADM safeguards? (draft guidance, March 2026)
- ICO: Guidance on AI and data protection
- OpenAI: A practical guide to building agents (PDF)
Want an AI agent working in your business?
Circlemount designs and builds AI agents for UK businesses, connected to the systems you already use, with human approval where it matters. Join the Circle for a free business analysis and proposal.





