The British Chambers of Commerce published its Future of Work: AI in the Workplace report on 18 March 2026, produced with Atos and the University of Essex ESRC Centre for Micro-Social Change. Around 94 percent of the firms surveyed were SMEs.
The headline is that more than half of UK firms, 54 percent, are now actively using AI. That is up from 35 percent in 2025, 25 percent in 2024 and 23 percent in 2023. Adoption has more than doubled in three years.
The second finding is the one worth sitting with. More than nine in ten SMEs using AI, 95 percent, report that it has had no impact on workforce size over the past year. Most firms, 86 percent, say job roles have remained unchanged. Meanwhile SMEs already using AI report a net productivity expectation of plus 71 percentage points.
Those three numbers together describe something quite specific, and it is not the story the market has been sold.
What is the survey actually showing?
That AI in UK small firms has been additive rather than substitutive. Adoption is broad, expectations of productivity are high, and almost nothing has changed about who works there or what they do. The tools went in around the existing team and the existing processes.
That is not a failure. It is what early-stage adoption of a general-purpose tool normally looks like. Spreadsheets did not reduce finance headcount either; they raised what a finance team was expected to produce. The mistake is not the pattern, it is buying into the pattern while measuring against a different one.
A firm that approved AI spending on a business case of cost reduction is now, on this evidence, quite likely to be twelve months in with the same payroll and a vague sense of disappointment. A firm that approved it on a business case of output per person is looking at the same data and seeing it work.
Why did the headcount savings not arrive?
Because the work small firms actually do is not shaped like the work AI most easily replaces. It is fragmented, exception-heavy and relationship-dependent, so AI absorbs pieces of many roles rather than the whole of any one. You end up with capacity released in small increments across a team, which does not translate into a removable position.
There is a second, more practical reason. In a firm of eight people, the person doing the work that AI now assists is also doing four other things. Even if AI removes 30 percent of one task, that person is not 30 percent removable. The saving is real but it is distributed, and distributed savings show up as slack rather than as a line in the accounts.
The BCC data does record one qualification worth noting honestly: 14 percent of SMEs investing in AI training anticipate headcount reductions over the next twelve months. So expectations are not uniformly flat. But the reported reality so far, at 95 percent no impact, is very different from the forecast that drove a lot of the spending.
So what is the return, if not payroll?
Capacity and consistency. The same team handling more volume, responding faster, and producing work of more even quality. That is a genuine commercial return, but it only converts into money if there is more volume to handle, which is a demand question rather than a technology one.
This is the point where a lot of UK small firms are currently stuck, and it is worth stating plainly. If AI has released 20 percent more capacity in a team and the enquiry volume has not moved, the firm has bought expensive slack. The investment was sound and the constraint was misdiagnosed. The binding limit was never how fast the work got done.
We made a version of this argument when the ONS reported the earlier rise in UK adoption, in our piece on UK AI adoption and what it means for small business automation. The BCC data a year on strengthens it rather than softening it: adoption has kept climbing and the workforce picture has stayed flat, which means the value has to be showing up somewhere else or not at all.
How should a firm decide whether AI is working?
Pick the measure before the purchase. If the case was capacity, measure work completed per person per month. If it was speed, measure time to first response. If it was quality, measure rework and complaints. Almost nobody does this, which is why so few firms can say whether the spending paid.
Three measures are usually enough for a small firm. Volume handled, time to respond, and proportion of enquiries that convert. Take a baseline for a month before anything is switched on. It is dull, it takes an hour a week, and it is the difference between knowing and guessing.
The reason this matters commercially is that it changes the next decision. A firm that can show AI released capacity has a clear case for spending on demand generation to fill it. A firm that cannot show anything has no case for spending on anything, and usually responds by buying another tool.
What should a professional firm do with released capacity?
Sell it. Capacity that is not converted into new work is a cost, and the fastest route to converting it is usually visibility rather than another automation. For most UK professional firms the enquiry volume is set by how findable and how credible they are, not by how efficiently they operate.
For accounting practices specifically, this is the moment when advisory ambitions become achievable. Firms have talked about moving up the value chain for a decade and been blocked by compliance workload. If AI genuinely relieves some of that workload, the constraint moves to whether clients and prospects know the firm offers anything beyond the annual return, which is a marketing problem with well-understood solutions.
Search engine optimisation is where that starts, because the questions a prospective client types are the cheapest and most durable source of qualified enquiries a professional firm has. Our pillar guide to digital marketing for accounting firms sets out how the pieces fit together for practices specifically.
The site has to be able to receive that demand. Website design and development work here is mostly about making the firm verifiable and the next step obvious, and about capturing every enquiry into one place so that the automation you have already bought has something to act on.
And the automation itself belongs after both, not before. AI automation delivers its clearest return when it is applied to a process that already carries volume, because that is when small per-transaction savings add up to something visible.
Should a firm tell clients it uses AI?
Yes, but describe the outcome rather than the tool. Clients do not buy AI, they buy faster answers and fewer errors. A page saying the firm uses AI reads as fashion; a page saying enquiries get a substantive reply the same working day reads as a commitment, and it is checkable.
There is a live risk in the other direction for professional firms, and it is worth handling openly. Some clients hear AI and think their confidential information is being fed into a public tool. Saying nothing does not avoid that question, it just means the client answers it themselves. A short, plain statement of what is automated, what a qualified person reviews, and how client data is handled removes the objection before it forms.
The firms getting this right treat it as a trust asset rather than a technology announcement. That is the same principle behind every other credibility page on a professional services site: state the specific thing a sceptical reader can verify, and let the adjectives go.
Is it too late to be an early adopter?
For the tools, yes, and it does not matter. At 54 percent adoption the competitive advantage is no longer in having AI, it is in having measured it. The firms that can point to what changed will make better decisions for the next three years than the ones that cannot.
There is also a quieter advantage still available. Because 86 percent of firms report roles unchanged, very few have redesigned how work actually flows. The gain from rethinking a process is generally larger than the gain from adding a tool to an unchanged one, and almost nobody has done it yet.
A four-week plan
Week one, baseline. Volume handled, time to first response, conversion rate. One month of honest numbers, no tooling required beyond a spreadsheet.
Week two, name the constraint. If capacity is tight, automation is the answer. If capacity is loose and enquiries are scarce, visibility is the answer. Firms that skip this step reliably buy the wrong one.
Week three, act on the constraint rather than the trend. This is the discipline the BCC numbers argue for, because the survey shows plenty of firms have bought the trend and reported no change.
Week four, publish something. Whatever capacity has been released, the way it turns into revenue is that more of the right people find you. Our note on the website mistakes costing firms clients is a reasonable place to check the basics first.
Triomatic Marketing is an AI-powered, founder-led agency, and we keep 82% of our clients largely by telling firms when the problem is not the one they came to us with. Our UK digital marketing agency page covers how we run this work for British firms. To talk it through, message Aria on WhatsApp via triomaticmarketing.com, or book a free 15-minute discovery call at https://calendly.com/hello-triomaticmarketing/15min.
FAQs
How many UK firms use AI in 2026?
The British Chambers of Commerce Future of Work: AI in the Workplace report, published on 18 March 2026 with Atos and the University of Essex, found 54 percent of UK firms actively using AI. That compares with 35 percent in 2025, 25 percent in 2024 and 23 percent in 2023. Around 94 percent of firms surveyed were SMEs.
Has AI reduced headcount in UK small businesses?
Not so far. More than nine in ten SMEs using AI, 95 percent, report it has had no impact on workforce size over the past year, and 86 percent say job roles have remained unchanged. A smaller group, 14 percent of SMEs investing in AI training, anticipates headcount reductions over the next twelve months.
What return are UK firms getting from AI?
Mainly capacity and consistency rather than cost reduction. SMEs already using AI report a net productivity expectation of plus 71 percentage points. That converts into revenue only if there is additional demand to absorb the released capacity, which is a marketing question rather than a technology one.
How should a small firm measure whether AI is working?
Choose the measure before buying. Track volume of work handled per person, time to first response, and the proportion of enquiries that convert, with a one-month baseline taken before anything is switched on. Firms that skip the baseline generally cannot say whether the spending paid for itself.
Why did AI not replace roles in small firms?
Because work in small firms is fragmented, exception-heavy and relationship-dependent, so AI absorbs parts of many roles rather than the whole of any one. The saving is distributed across a team and shows up as slack capacity rather than as a removable position.
Is it too late for a UK firm to gain an advantage from AI?
The advantage has moved. At 54 percent adoption, having AI is no longer a differentiator, but having measured its effect is. Because 86 percent of firms report roles unchanged, very few have redesigned how work flows, and redesigning a process usually returns more than adding a tool to an unchanged one.