Only about 16% of UK businesses currently use any form of AI technology. That's according to the government's own research, published in January 2026 (DSIT AI Adoption Research). Which means the vast majority of businesses haven't started.
Not because AI isn't useful. Because nobody showed them where to begin.
The conversation around AI has become unhelpfully abstract. Every conference talks about "transformation" and "strategy." Every LinkedIn post promises a revolution. But if you're running a ten-person business and you just want to stop wasting three hours a week on meeting notes, you don't need a strategy. You need a tool, twenty minutes, and someone to show you which button to press.
That's what this article is. Five specific automations. Real tools. Things you can set up this week with no coding, no consultants, and no budget beyond what you're probably already paying for.
Why most businesses haven't started
Before the practical stuff, it's worth understanding the gap. An ONS study from March 2025 found that 39% of UK firms identified "difficulty finding where AI would help" as their biggest barrier to adoption (ONS, 2025). Not cost. Not technical complexity. Just not knowing where it would actually be useful.
That makes sense. If your reference point for AI is chatbots and image generators, it's not obvious how that helps you run a business. But AI tools have moved well beyond novelty. The ones worth using now are boring in the best way — they handle the repetitive tasks you've been doing manually, and they do it reliably.
Here are five that work.
1. Meeting notes and action items — automatic
The problem: You spend five hours a week in meetings. After each one, someone has to write up the notes, distribute them, and track the action items. Usually, nobody does this properly. Important decisions get forgotten. Action items drift.
The tool: If you use Microsoft 365, Copilot can join your Teams calls and transcribe them automatically. If you're on Google Workspace or Zoom, tools like Otter.ai or Fireflies.ai do the same thing. They join the call (with a visible participant notification — no secret recording), transcribe the conversation, summarise the key points, and pull out action items.
Setup time: About 15 minutes. Connect the tool to your calendar, grant it access to your meeting platform, and let it join automatically.
What changes: Meeting notes exist for every meeting, not just the ones where someone remembered to take notes. Action items are captured in the moment, not reconstructed from memory the next morning. If you've ever had a dispute about what was agreed in a meeting, this solves it.
One caution: Make sure your transcription tool complies with UK data protection requirements. Check where the data is stored and processed. And always tell meeting participants that the call is being transcribed — it's both a legal requirement and basic courtesy.
2. Email triage and drafting
The problem: You start the day with 40 emails. Half are informational. A quarter need a routine response. Maybe five require actual thought. But you spend 45 minutes going through all of them because they all look the same in your inbox.
The tool: If you're on Gmail, Google's built-in AI features can categorise, summarise, and suggest replies. Outlook has similar capabilities through Copilot. For a more dedicated approach, SaneBox sits on top of your email and automatically sorts messages by importance, pulling newsletters and low-priority items out of your main inbox.
Setup time: About 30 minutes to configure categories and train the system on your preferences.
What changes: You open your inbox and the five emails that need attention are at the top. The routine ones have draft replies waiting for your approval — a quick review, a click, done. The newsletters are in a separate folder for when you have time. You've just reclaimed 30 minutes of your morning.
A note on draft replies: Never send AI-drafted replies without reading them. They're good at the structure and the pleasantries. They're occasionally wrong about the specifics. Treat them as a starting point, not a finished product.
3. Invoice and expense processing
The problem: You're still photographing receipts and emailing them to your accountant. Or typing invoice details into a spreadsheet. Or — honestly — shoving receipts into a drawer and dealing with it at quarter-end. None of this is a good use of your time or your accountant's.
The tool: Dext (formerly Receipt Bank) is the standard in the UK accounting world. Snap a photo of a receipt or forward an invoice by email. The AI extracts the supplier, date, amount, VAT, and category — then pushes it directly into your accounting software. Xero and QuickBooks have similar built-in features, though Dext tends to be more accurate for UK formats.
Setup time: About an hour. Connect your bank feed, link your accounting software, and process your first few receipts to train the categorisation.
What changes: Your bookkeeping happens in real time instead of in panicked batches. Your accountant gets clean, categorised data instead of a carrier bag of receipts. Your VAT returns are faster. Your cash flow visibility improves because you actually know what you've spent.
4. Content repurposing
The problem: You wrote a blog post last month. It was good. It took four hours. And then it sat on your website collecting dust while you posted nothing on LinkedIn, sent no newsletter, and didn't share it anywhere else. The content exists. The distribution doesn't.
The tool: This one is less about a single tool and more about a workflow. Claude, ChatGPT, or similar AI tools can take a blog post and turn it into a LinkedIn post, a series of tweets, an email newsletter intro, and a summary for your internal team — in about ten minutes. Repurpose.io and Opus Clip can do similar things with video and audio content.
Setup time: 1-2 hours to build your first workflow and write the prompts that match your tone.
What changes: One piece of content becomes five. Your blog post reaches people who would never have visited your website. Your newsletter has something to say every week. Your LinkedIn doesn't go quiet for three months because "nobody had time to write something."
Important: Don't just copy-paste AI output. Use it as a draft. The value is in the speed of the first version, not in publishing it unedited. Your audience will know the difference.
5. Client enquiry auto-response and routing
The problem: Someone fills out your website contact form at 3pm on a Wednesday. You see it at 9am Thursday. You reply at lunchtime. By then, they've already contacted two of your competitors who responded within the hour. Speed of response is one of the strongest predictors of conversion, and most small businesses are terrible at it.
The tool: A combination of your form builder (Tally, Typeform, or a basic web form) connected to an automation platform (Zapier or Make) that triggers an immediate acknowledgement email, asks qualifying questions, routes the enquiry to the right person, and logs it in your CRM. For website chat, tools like Crisp can handle initial responses and qualification automatically.
Setup time: About 2 hours for a complete flow: form submission → auto-acknowledgement → qualification → routing → CRM logging.
What changes: Every enquiry gets an immediate, professional response. The prospect knows their message was received. Basic information is collected before your team even sees it. When someone does pick up the enquiry, they have context instead of just a name and email address. And nothing falls through the cracks because the system tracks everything.
What not to automate (yet)
Not everything should be automated. Some things genuinely need a human.
- Client relationships. Automated follow-ups are fine. Automated relationship management is not. People can tell the difference between a thoughtful message and a triggered one.
- Financial decisions. AI can categorise expenses and flag anomalies. It should not approve invoices or make payment decisions.
- Sensitive communications. Anything involving bad news, conflict resolution, or emotional nuance should come from a person.
- Processes you don't understand. If you can't explain how a process works today, don't automate it. You'll automate the mess, not fix it. Automate what works. Fix what's broken.
What comes next
Once you've set up these five automations, something interesting happens. You start noticing patterns. You see where the remaining bottlenecks are. You realise that the weekly report you've been assembling manually could be generated automatically. That the client onboarding process has six steps that could be three. That the data entry someone does for two hours every Friday is completely unnecessary.
That's when a more systematic review makes sense — mapping your workflows, identifying the highest-impact opportunities, and building automations that are specific to how your business operates. Not before. Start with the five above. Get comfortable with the idea that software can handle the repetitive stuff. Then go further.