A lot of small and medium-sized businesses are making AI harder than it needs to be. They hear that they need an “AI strategy”, see large companies talking about transformation programmes, policies, data platforms and multi-year roadmaps, and conclude that it sounds expensive, risky and not for them yet.
For most SMEs in Gloucestershire and Cheltenham, that is the wrong starting point. You do not need a grand AI strategy before doing something useful. You need one well-chosen project that solves a real problem: not a novelty, not a chatbot for its own sake, and not an experiment that impresses people for five minutes and is then forgotten.
Start with friction, not technology
The most common mistake is starting with the tool. Someone signs up for an AI platform and asks, “What can we use this for?” That usually leads to shallow experiments, a few people writing emails, and a conclusion that AI is interesting but not essential. A better question is: what work keeps getting done manually even though it follows a fairly repeatable pattern?
- Sorting incoming enquiries
- Summarising long email threads before a client meeting
- Turning meeting notes into actions
- Checking documents for missing information
- Drafting first responses to common questions
- Researching prospects before a sales call
- Preparing a weekly internal update from scattered sources
These are not glamorous, and that is the point. The first useful project usually sits inside ordinary admin, sales, operations or customer service work. If a task happens often, takes longer than it should and has a clear “good enough” output, it is a good candidate.
Small, contained and clearly bounded
Small does not mean pointless. It means contained. You should be able to say when the project starts, what information it uses, what it produces and who checks it. “When a new enquiry comes in, AI reads it, identifies the type, drafts a short internal summary, suggests the next step and prepares a reply for a human to review” is a project. “We want to use AI in customer service” is not. The first version does not need to connect to every system or make final decisions. It reduces the repetitive thinking and typing before a person does the important part.
Examples of a contained first project: a local accountancy firm starting with document intake, checking whether a client has sent the right files and producing a short summary for the team; a professional services firm starting with meeting follow-up; a trades or property business starting with enquiries, extracting location, urgency, contact details and likely next step; a marketing team starting with research and first drafts for review.
Keep the human in the loop
For most SMEs the safest first step is assisted work, not full automation. AI can draft, summarise, classify, extract and suggest, but a person reviews anything that matters before it reaches a client, supplier or the public. AI can be confidently wrong, misread context, or expose sensitive information if used carelessly. So the first project needs basic rules: what information the AI may use, what it must not see, who checks the output, what happens when it is uncertain, how mistakes are spotted and corrected, and how success is measured. This does not need to be a 40-page policy, but staff must know how the workflow should be used.
Measure something real
If the project is useful, you should be able to measure it. Pick one or two measures before you start, such as time saved per enquiry, fewer missed follow-ups, faster client-meeting preparation, fewer documents returned for missing information, lighter admin load on a senior person, or more consistent first drafts. A simple before-and-after comparison over a few weeks is often enough. The aim is not to prove AI is magical. It is to find out whether a specific workflow improves the business.
Score the options, then pick one
If you are not sure which job to start with, list what your team repeats every week, where people copy, paste, reformat or summarise information, and which tasks are important but safe enough to test. Then score each idea on four tests: value (would it save time, reduce errors or improve service?), frequency (does it happen often enough?), risk (can a human review the output before it affects a client?) and data (can the AI safely use the information?). The winner is not always the biggest opportunity. It is the one that gives a useful, low-drama first win.
A good first project fits in one sentence, has a named owner, has a before-and-after comparison, and has a clear rule for when the AI may suggest, when it may draft and when a person must take over. It should also produce evidence: did it save time, were the outputs good enough, what mistakes did it make, and did staff trust it more or less afterwards? That evidence is worth far more than another strategy discussion.
If you run an SME or professional firm in Gloucestershire or Cheltenham and want help identifying your first practical AI project, book a short AI consultancy call or get in touch and ask: “What could AI usefully do in our business?”
