AI is already finding its way into small and medium-sized businesses. Sometimes it arrives through a planned initiative. More often, it starts with individuals testing tools to draft content, summarise documents, research a topic or speed up routine work.
That can create useful momentum. It can also leave a managing director with an untidy set of questions.
Where is AI already being used? Is it helping? What information is being entered into public tools? Which ideas deserve investment? Could the business move too slowly, or spend too quickly?
Recent YouGov research gives a useful view of how other UK SMEs are approaching those questions. The message is not that every business should rush to adopt AI. It is that leaders need a clearer way to separate practical opportunities from noise.
The sensible starting point is not “Which AI tool should we buy?” It is “Which part of the business needs to work better, and what would a useful result look like?”
What does the YouGov research tell us about AI adoption in SMEs?
YouGov surveyed 1,000 SME decision-makers in organisations with up to 250 employees through its B2B Omnibus. The findings were published on 7 August 2025.
The research found that 31% of the SMEs surveyed were already using AI tools, while a further 15% planned to. Awareness was much higher, with 86% saying they were familiar with AI.
That gap matters. Knowing about AI is not the same as knowing where it belongs in your business.
Among SMEs using or planning to use AI, the most common reason was task automation at 54%. Marketing and advertising followed at 45%, with product or service development at 37%, customer service at 31% and operations or logistics at 28%.
These figures show that SMEs are finding practical uses for AI. They also suggest that adoption often begins in individual functions rather than through one business-wide strategy.
For a founder or managing director, that creates a straightforward responsibility: understand what is already happening before deciding what should happen next.
Do not confuse AI activity with AI progress
A team using several AI tools may look further ahead than a business using none. That is not always true.
Activity can grow without clear ownership, agreed measures or a shared view of risk. Different departments may pay for similar tools. Employees may enter business information into services that have not been reviewed. A pilot may produce an impressive demonstration but fail to fit the real process.
Progress looks different.
It means the business can explain:
- the problem being addressed
- the outcome that should improve
- the people affected by the change
- the information the tool needs
- the decisions that must remain with a person
- the cost and effort involved
- how value, quality and risk will be measured
A smaller number of well-chosen uses will normally tell you more than a long list of disconnected experiments.
Where should an SME start with AI?
Start with a business process that is causing a recognisable problem.
That could be repeated administration, inconsistent records, slow access to information, avoidable rework, poor handovers or a service that depends too heavily on knowledge held by one or two people.
Do not begin by asking teams to suggest exciting AI ideas. Ask them where work is getting stuck.
A useful first conversation might cover:
- What is taking too long or producing inconsistent results?
- Who performs the work and who depends on the outcome?
- Which parts require judgement, expertise, empathy or a trusted relationship?
- What information is used, and how sensitive is it?
- What has already been tried?
- What would count as a worthwhile improvement?
This keeps the discussion tied to the business rather than the novelty of the technology.
AI may be part of the answer. The better answer may also be a clearer process, improved data, better use of an existing system or more consistent ownership.
Is AI worth it for a small business?
AI is worth considering when it can improve a meaningful outcome at a proportionate cost and risk.
That sounds obvious, but the YouGov research found that 30% of businesses not planning to use AI simply did not see the value. That should not be dismissed as a lack of ambition. In many cases, the value has not been explained in business terms.
A founder should be able to describe the expected benefit without relying on broad promises about productivity.
For example:
- reducing the time needed to prepare a weekly operational report
- helping staff find approved information without searching several systems
- improving the consistency of customer or service records
- highlighting exceptions that need a manager’s attention
- reducing repetitive document preparation while retaining human approval
The expected value should then be compared with the full cost of making the change work. That includes software, integration, staff time, training, governance, support and ongoing review.
A low monthly licence fee does not make an idea low-cost if the process around it is unclear.
Privacy and security concerns are not a reason to avoid the question
Among the businesses in the YouGov study that were not planning to use AI, 49% cited data privacy and security concerns. Ethical concerns were raised by 19%.
Those concerns are reasonable. They are also more useful when turned into specific questions.
Before approving an AI use case, an SME should understand:
- what information the tool needs
- what information can be removed, masked or excluded
- who can access the input and output
- where data is stored and processed
- how long it is retained
- whether the provider can use it to train its models
- how information is deleted
- what happens if the tool produces an incorrect or harmful output
- who remains accountable for the final action or decision
The answers will depend on the use case, supplier, configuration and delivery design. A public tool used to rewrite non-sensitive marketing copy presents a different risk from a system handling employee, customer, financial or service-user information.
The aim is not to create the heaviest possible governance. It is to use controls that match the potential impact.
Protect the qualities that make the business valuable
The research also shows why AI adoption cannot be treated as a software decision alone.
Among adopters, 48% were concerned about the effect of AI on employees’ critical-thinking skills. Legal risks concerned 42%, job losses concerned 34%, and around three in five were worried that over-reliance on AI could reduce business creativity.
For an SME, this is especially important. Smaller businesses often compete through experience, responsiveness, judgement and close customer relationships. Weak AI adoption can dilute those strengths if teams begin accepting outputs without challenge or remove valuable human contact from the wrong part of a service.
The answer is not to keep people outside the process. It is to define their role properly.
For each use case, decide:
- what the system may draft, recommend or identify
- what a person must check
- who can approve or change the output
- when an issue must be escalated
- how staff can challenge an incorrect result
- what knowledge or skill could be weakened through over-reliance
This is sometimes called keeping a human in the loop. In plain English, it means a person reviews, approves or can change an AI-supported output before it affects an important decision or action.
Five questions to answer before spending money on AI
You do not need a lengthy AI strategy before exploring one useful opportunity. You do need enough discipline to answer five questions.
1. What problem are we trying to solve?
Name the process, the people involved and the outcome that needs to improve. “We need to use AI” is not a business problem.
2. What would measurable value look like?
Choose evidence that can be observed. Time saved may matter, but so might quality, turnaround time, service consistency, workload, customer experience or reduced risk.
3. What information would be required?
Identify the data source, owner, sensitivity, quality and access requirements. Do this before choosing a provider.
4. Where must people retain control?
Clarify who reviews outputs, handles exceptions and remains accountable. The more important the decision, the clearer the human role should be.
5. What would make us stop?
Agree the conditions under which the idea should be paused or rejected. Weak accuracy, poor adoption, unacceptable data exposure or a business case that no longer holds are valid reasons to stop.
A stop decision is not a failed AI programme. It is evidence that the business has avoided a poor investment.
What should a managing director do in the next 30 days?
useful first month does not require a large programme.
Week 1: Find out what is already happening
Ask team leaders where generative AI or automated decision support is already being used. Include approved tools, free public services and AI features built into existing software.
The purpose is visibility, not punishment. People are more likely to be honest when the discussion is about learning and safe use rather than catching them out.
Week 2: Choose one or two processes
Look for work that is repeated, information-heavy or inconsistent. Avoid choosing a use case simply because it will produce an impressive demonstration.
Week 3: Test value and hazards
Review the likely benefit alongside data, privacy, security, accuracy, people, systems and ownership. Write down what is known and what still needs evidence.
Week 4: Make one clear decision
Proceed to a fuller assessment, prepare the foundations, pause the idea or stop it. Do not leave the business with a vague instruction to “explore AI”. Give the next step an owner and a reason.
When should an SME bring in external AI support?
External support is useful when the business needs independent challenge, specialist knowledge or capacity that it does not have internally.
YouGov found that 67% of AI-using SMEs said they had staff with the right expertise in-house. It also found that 13% already used external consultants and 26% were likely to bring in outside support.
This does not need to be an either-or decision. Internal teams understand the business, customers and processes. A suitable external adviser should add structure, challenge assumptions and fill specific gaps without displacing that knowledge.
Consider outside support when:
- several departments have competing ideas
- no one owns AI use across the business
- privacy, security or supplier questions are unresolved
- a pilot has stalled before normal use
- leaders need a clearer business case
- internal teams are too close to a preferred tool or solution
- the opportunity requires process, people and delivery expertise as well as technical knowledge
Ask any adviser how they decide when AI is not the right answer. Their response will tell you a great deal about how they work.
A practical route from interest to evidence
AI Terrain helps SMEs improve how work gets done, test where AI may add value and move viable opportunities towards responsible delivery.
We start with the process, the people and the outcome. We then examine value, data, privacy, security, accuracy, ownership and practical delivery before recommending further investment.
Start with The Recce
The Recce is a free, one-hour AI readiness conversation. It reviews current AI use, sanctioned or not, explores one or two candidate processes and applies a standard hazard check.
The aim is simple: establish whether there is enough signal to plan the path. A 15 to 20-minute fast-track option is available for contacts who are already fluent in the problem and context, but the hazard check is never skipped.
[CTA button: BOOK THE RECCE]
Test the opportunity properly
Where an idea deserves a closer look, an AI feasibility assessment can test the business outcome, process, people, data, privacy, security, systems, governance and delivery requirements in more depth.
The result should be a clearer decision about whether to proceed, prepare, reshape or stop.
[CTA button: Learn what an AI feasibility assessment covers]
Move viable ideas into delivery
Where the opportunity is viable, AI Terrain can help define the target process, requirements, controls, human role and delivery plan. We remain solution-agnostic and work with existing teams, suppliers and appropriate partners rather than forcing a preferred platform.
[CTA text button: Explore AI strategy and delivery]
The point is not to adopt AI first
The YouGov research describes a market that is interested, active and still cautious. That is a sensible place for SMEs to be.
The strongest businesses will not be the ones that collect the most AI tools. They will be the ones that understand where AI improves the work, where it creates new risk and where human judgement remains essential.
For a founder or managing director, the next move is not to write a grand AI strategy or approve a large technology budget.
Find one worthwhile problem. Understand the process. Test the evidence. Then decide.
Frequently asked questions
Where should a small business start with AI?
Start with a specific business process that is slow, repetitive, inconsistent or difficult to manage. Define the desired outcome, understand how the work currently happens and identify where human judgement must remain before considering a tool.
Is AI worth it for SMEs?
AI can be worthwhile when it improves a meaningful outcome at a proportionate cost and risk. The business should test the full case, including implementation, staff time, training, data controls, support and ongoing governance, rather than judging value from a software licence alone.
How can an SME use AI safely?
Use AI safely by controlling what information is entered, checking supplier terms, restricting access, testing accuracy and defining who reviews and approves outputs. The controls should reflect the sensitivity of the data and the impact of the decision or action.
Does a small business need an AI strategy?
A small business does not always need a large strategy document. It does need clear priorities, ownership, rules for data and acceptable use, a way to assess opportunities and evidence for deciding whether to invest further.
What is shadow AI?
Shadow AI is the use of AI tools or features without formal approval, visibility or oversight from the organisation. It may involve useful experimentation, but it can also create data, security, quality and accountability risks if it remains unmanaged.
Will AI replace jobs in an SME?
AI may change tasks and responsibilities, but job removal should not be treated as the default outcome. A responsible approach looks for avoidable work that can be reduced while protecting judgement, creativity, expertise, relationships and accountable decision-making.
Source note:
This article interprets findings from YouGov’s article, “We polled UK SME leaders about AI adoption. Here’s what they said”, published 7 August 2025. The research surveyed 1,000 SME decision-makers in organisations with up to 250 employees through YouGov’s B2B Omnibus.
Source: YouGov SME AI adoption research
The findings provide market context. They do not represent every UK SME and should not be treated as proof that a particular AI use case will succeed.