Five things UK SMEs are already using AI for right now

The most useful thing anyone can tell you about AI isn't which tool to use. It's which tasks are actually worth trying AI on — and what the honest experience of using it looks like.

Here are five things UK SMEs are currently using AI for, what's working, and where it falls short.

1. Handling first-line customer enquiries

What it looks like: An AI tool — usually a chatbot or an email auto-responder — handles incoming questions that follow predictable patterns. Opening hours, order status, basic product questions, booking requests.

What actually works: Significant time saving on volume. For businesses that receive a lot of similar enquiries, AI can handle 60–80% of first-contact queries without human involvement.

Where it falls short: Anything that requires judgement, empathy, or context. A dissatisfied customer who's had a bad experience doesn't want a chatbot. Get the handoff to a human wrong and you've made the problem worse.

2. Writing first drafts

What it looks like: Using a large language model — ChatGPT, Claude, Gemini — to produce a first draft of something. A proposal, a job description, a product description, an email.

What actually works: The blank page problem disappears. Most business owners who use AI for writing report that the main value isn't the output — it's the starting point. A mediocre first draft is much faster to improve than writing from nothing.

Where it falls short: The output needs editing. AI writing is often accurate but flat — it doesn't sound like you, it doesn't reflect your specific context, and it tends to produce the most average version of whatever you asked for. The editing step is non-negotiable.

3. Scheduling and calendar management

What it looks like: AI tools that read your calendar, suggest meeting times, reschedule automatically, or integrate with booking systems to manage appointment flow.

What actually works: For businesses with high appointment volume — trades, healthcare, professional services — automating the back-and-forth of scheduling saves meaningful time. Tools like Calendly, Reclaim.ai, and others in this space are mature enough to be genuinely reliable.

Where it falls short: Integration complexity. Getting an AI scheduling tool to talk to your existing booking system, CRM, and calendar simultaneously usually requires either technical setup or a workaround. It's worth the time investment for businesses with real scheduling volume; less obviously worth it for businesses that schedule a handful of meetings per week.

4. Summarising and synthesising information

What it looks like: Feeding a long document, report, meeting transcript, or email thread into an AI tool and asking it to summarise the key points, action items, or decisions.

What actually works: Genuinely useful for businesses that deal with large volumes of text. Contract review, meeting notes, supplier terms, research reports. AI summarisation is fast and accurate enough for most business purposes.

Where it falls short: You still need to check the output. AI tools occasionally miss things, mischaracterise nuance, or produce summaries that are technically accurate but miss what actually matters. Don't send an AI-generated summary to a client or sign a contract based on one without reading the original.

5. Stock alerts and inventory signals

What it looks like: AI tools that monitor inventory levels, identify reorder triggers, flag unusual sales patterns, or predict demand based on historical data.

What actually works: For product-based businesses with consistent sales data, AI can meaningfully improve stock management. The tools that do this best are usually embedded in existing platforms — Shopify, Xero, certain ERP systems — rather than standalone AI apps.

Where it falls short: Data quality. AI inventory tools are only as good as the data they're trained on. If your stock data is inconsistent, incomplete, or not connected to your sales data in real time, the predictions will be unreliable. Getting the data right is usually more work than getting the AI to use it.

The honest pattern across all five: AI is useful when the task is repetitive, the input is consistent, and human judgement is not the critical ingredient. It's less useful — or actively risky — when the task requires context, relationships, or discretion that AI can't replicate.

Start with one task. See if it helps. Go from there.