Most small business owners come to AI one of two ways. Either they've been burned — they tried something, it didn't stick, and now they're sceptical. Or they haven't started yet, and they're not sure where to begin without it feeling like a gamble.
This guide is for both groups. What follows is the method I actually use with clients to find a first use case that works.
Jake Whitford
Founder, Magnetic AI
Jake has run a manufacturing business and now helps UK SMEs adopt AI practically — process first, tools second. He works directly with every client from his Plymouth base.
What counts as a good first AI use case?
A good first AI use case has three things going for it: it's repetitive, it's document or language-based, and it's low risk — if the AI gets it slightly wrong, a human catches it before anything bad happens.
Good candidates
- Drafting routine emails and proposals
- Summarising meeting notes
- Writing first drafts of job descriptions
- Pulling key information from supplier documents or contracts
- Creating social media content from existing material
Not good candidates (yet)
- Anything that makes a financial decision autonomously
- Anything that goes directly to a customer without human review
- Anything where data is sensitive and governance isn't yet in place
The seven steps
Find the tasks your team quietly dreads
Ask one question: "What do you do regularly that takes longer than it should?" Then ask: "Could a clever person do this if you briefed them properly?" If the answer is yes, you have a candidate. Common answers: writing proposals, responding to enquiry emails, preparing meeting summaries, writing job adverts.
Map where your time actually goes
Spend one week keeping a rough log of where your time goes. Pay attention to anything where someone copies information from one place and pastes it somewhere else — every copy-paste is a symptom. Also look for tasks where someone produces a similar output repeatedly but starts from scratch each time.
Ask whether the task is language-based or logic-based
Current AI tools are exceptionally good at language tasks — writing, rewriting, summarising, classifying, and extracting information from text. They're less suited to precise numerical calculation or complex multi-step logic without integrations. If the task involves working with words and documents, you're in good shape.
Test before you commit
Before you buy anything or involve your IT team — just try the task. Take a real example, give the AI a clear brief, and see what comes back. Do this five or ten times with real examples, not hypothetical ones. You're not looking for perfection. You're looking for whether the output is a useful starting point.
Calculate the time saving honestly
How often does this task happen? How long does it currently take? How much time would the AI-assisted version take? Multiply the saving by how many times it happens per year. A couple of hours a week is roughly 100 hours a year — meaningful money, and time your team can spend on work that moves the business forward.
Get one person to own it
Give one person responsibility for the trial. They write the prompts, review the outputs, track whether it's saving time, and flag problems. In most SMEs, the right person is whoever does the task most often. Give them a month with real examples — at the end you'll have genuine evidence, not a vendor pitch.
Document what works before you scale
Before rolling out to the wider team, write down what works. The prompt you use, the context you give, the things you always change in the output. A single page or shared note is enough. This means the next person doesn't have to figure it all out again — and it's the foundation everything else builds on.
Common questions UK SME owners ask
Do I need a big budget to start with AI?
No. The tools that will cover most first use cases — Claude, ChatGPT, Microsoft Copilot — cost between £15 and £30 per user per month. A sensible first pilot costs very little, and the time saving almost always justifies it within weeks.
Do I need a technical person to implement AI?
For most first use cases, no. If the task is writing or document-based, you can get meaningful results using an AI tool directly, without any integration or development work. Technical capability becomes more relevant when you want to connect AI to your existing systems.
Is my data safe if I use ChatGPT or Claude?
It depends on the tool and the plan. Consumer-tier tools may use your data to improve their models by default. Business and enterprise tiers typically don't. For anything involving personal data, client information, or commercially sensitive material, check the data processing terms before you start — or use a tool that offers a data processing agreement. A genuine consideration under UK GDPR.
What if my team is resistant to using AI?
Resistance almost always comes from fear that AI will replace their job, or frustration that a previous tool didn't help. The best response is to start with a use case that clearly helps the person doing the task. If AI saves your team members an hour a week on something tedious, they'll come round quickly.
How do I know if AI is right for my specific business?
There's no business I've worked with where AI has had nothing to offer. The question isn't whether there's a use case — it's which one to start with, and in what order.
The summary version
Find the task someone dreads. Check it's language-based and low risk. Test it with a free or cheap tool using real examples. Measure the time saving honestly. Give one person ownership of the trial. Document what works before you go further.
The first use case doesn't need to be exciting. It needs to be real, repeatable, and worth the time it saves.
Keep reading
Not sure where to start in your specific business?
I run a free thirty-minute call to help business owners identify their first practical AI use case — no sales pitch, just a straight conversation.