If your inbox looks anything like mine, a good chunk of it is promising that AI will transform your business by Tuesday. Most of it is noise from people who’ve never run a small firm and never had to chase a payment or answer the same enquiry forty times in a week. So before any of that gets a look-in, it’s worth sitting with what the government actually found when it bothered to measure this, because the real picture is a lot calmer, and a lot more useful, than the breathless one.
Start with the number that should take some pressure off. The Department for Science, Innovation & Technology surveyed 3,500 UK businesses for its 2026 AI Adoption Research, and found that just 16% used any AI technology at all. Fully 80% neither used it nor had any plans to. So if you’ve quietly assumed that everyone else cracked this months ago while you’re still stuck doing things the long way, you haven’t fallen behind. You’re standing with the large majority. There is no secret club, and the few who’ve joined it mostly haven’t done anything as dramatic as the marketing suggests.
What it’s actually for
Here’s the part of the same study that tells you what AI is genuinely good for today. Among the firms that have adopted it, 75% report better workforce productivity. But only 12% report any rise in revenue, and most saw their revenue not move at all. Read those two figures together and the headline writes itself: right now, AI gives you time back, not money directly.
For a small business that’s the right kind of win, even if it’s less exciting than the sales pitch. Time is the thing you’re shortest of. The hour you don’t spend re-typing a job sheet, or triaging the same five questions, is an hour you can put on work that does pay, or on going home at a sensible time for once. Anyone selling AI as a money-printing machine is running ahead of the evidence. Anyone selling it as a way to stop drowning in admin is bang on the money the research describes.
Why most firms haven’t bothered
You’d assume the hold-up is the technology being too fiddly or too unproven. The DSIT findings say otherwise. The two biggest barriers weren’t technical at all. They were “no identified need for it”, cited by 71% of businesses, and “limited AI skills”, cited by 60%. In plain terms, the blocker isn’t the AI. It’s knowing what to point it at.
That matches what I see talking to people in Suffolk. Nobody’s stuck because they can’t operate a chatbot. They’re stuck because spotting the one job in the week where a tool would genuinely earn its keep, and then actually wiring it into how they already work, is real work in itself. That gap between “AI exists” and “AI is quietly doing something useful in my business” is where almost everyone gets lost. It’s also where the value is.
It’s worth being honest about why that gap is so wide. The tools demo beautifully and disappoint quietly. You can sign up to something in ten minutes, get a slick first answer, feel briefly like the future has arrived, and then never open it again because it didn’t fit the way your day actually runs. That’s not a skills problem in the sense the survey means it. It’s that “use AI” is advice pitched at the wrong altitude. No business owner wakes up wanting to use AI. They wake up wanting the Tuesday-morning enquiry pile dealt with, or the quotes out before lunch. The tool is only ever the means, and the firms making it pay are the ones who started from the job and worked backwards, rather than starting from the technology and hunting for somewhere to put it.
The boring stuff is the good stuff
So what’s worth doing? The honest answer is the dull, repetitive, language-shaped work you already resent. Drafting and triaging the enquiries you answer fifty times a week. Turning a rambling voice note into a clean record. Answering the routine “do you cover my area” and “are you open Saturday” questions the moment they land, instead of at half nine that night. First-pass admin: summarising, sorting, tagging, so the thing reaches you already half-sorted.
That’s not me guessing at what’s possible. It’s what people actually do with these tools. In the same research, 85% of AI users put it to natural-language work, and 72% to marketing and admin. The grunt work, in other words. Nobody’s reporting that AI rewrote their business model. They’re reporting that it took a tedious, wordy task off their plate.
Which brings us to the hype. The sci-fi end, AI that “runs your company”, autonomous everything, the all-singing platform that promises thirty features and does none of them properly, is mostly that: hype, at least at the scale a small firm operates. The genuine wins are narrow and a little boring, and they tend to do one thing reliably rather than thirty things badly. That’s not a disappointment. For a business with no spare hours, narrow and reliable is exactly what you want.
There’s a tell, too, for spotting which side of that line a given pitch sits on. If someone can’t tell you which specific task they’re taking off your hands, and can’t put a rough number on the time it saves, they’re selling you the idea of AI rather than a working tool. The useful conversations sound mundane. They’re about a named bottleneck, a particular form, a recurring email, the bit of the week you dread. The hype-y ones stay vague on purpose, because vague is where the magic lives and the awkward questions can’t reach.
A worked example, so it’s not all abstract
Numbers help more than adjectives, so here’s a small, plausible one. Say you field around 30 enquiry emails a day, the usual mix of quotes, questions and time-wasters. Suppose an assistant drafts a reply and triages each one for you, and that saves roughly three minutes apiece, the time you’d otherwise spend reading, deciding and typing.
| Emails/day | Mins saved each | Hrs/week | Hrs/year |
|---|---|---|---|
| 30 | 3 | ~7.5 | ~350 |
Illustrative figures.
That’s 90 minutes a day, about seven and a half hours a week, and somewhere near 350 hours over a working year of roughly 47 weeks. Notice what that table is and isn’t. It’s nine working weeks handed back to you. It is not nine weeks of new revenue. Whether those hours turn into money depends entirely on what you choose to do with them, which is precisely the DSIT finding in miniature: the tool gives you the time, and what you do next is still your job. If you want to see where hours like these actually leak out in the first place, what admin really costs walks through it.
The thing nobody tells you upfront
AI isn’t a product you buy off a shelf and switch on. It’s a capability you aim at a specific job. The businesses getting real value out of it aren’t the ones with the longest list of subscriptions. They’re the ones who found the single repetitive, wordy task eating their week and built something small that handles it, then left it running.
That’s the exact gap those DSIT barrier numbers point to, and it’s the work we do. We don’t sell you “AI” as a thing. We find the load-bearing job in your business where automation genuinely helps: the inbox that triages itself, the enquiry answered at eleven at night before a competitor gets to it, the notes that become records on their own. Then we build the small thing that does it and keep it running. If you want the shape of that, our trainable inbound AI agent is one we’ve built more than once, and never miss the enquiry covers why that first reply matters so much.
If there’s a job in your week that’s repetitive, wordy and quietly costing you hours, that’s the one to aim this at. Drop us a line describing it, and we’ll tell you straight whether it’s worth automating, and what we’d build if it is.