The more useful detail is where that AI is landing. Among businesses using or planning to use AI, 81% apply it to administration, ahead of marketing (71%) and IT (66%). The ONS adds that, among AI-using firms with 10 or more staff, administrative and clerical roles are the ones most affected (38%, and rising).
So the first wave of AI isn't transforming business models. It's doing the paperwork. That's a great place to start, as long as the paperwork runs on processes and systems that can take it.
The gains are real, but mostly modest
DSIT is refreshingly honest about results. Adopters most often report better workforce productivity (71%) and more efficient or flexible processes (60%). Half report an overall productivity increase, but describe the gains as modest. Most (64%) have seen no immediate effect on operating costs, and only 14% rate AI as highly or critically important to their operations.
The Social Market Foundation's survey of 1,015 UK SME leaders, published on 19 October, explains why. Advanced users report significantly better outcomes than basic users, and the highest-value uses sit in process-heavy areas such as HR, sales, supply chain and IT. The SMF estimates that wider and deeper adoption could add £38 to £47 billion to SME revenue across Great Britain.
In other words, the difference between a nice time-saver and a real return is depth: AI wired into how the work actually flows, not just a chat window beside it.
Why admin AI stalls
Admin is where AI meets the business's plumbing: invoices, orders, customer records, HR files, supplier emails. If that plumbing is patchy, AI simply speeds up the patchiness. Three patterns stand out in the evidence:
- The systems underneath are thin. The OECD's 2026 review of UK SME technology finds that SMEs have largely caught up on process tools such as websites, e-sales and accounting software, but adoption of management systems like CRM and ERP remains modest, especially in smaller firms. AI that drafts a customer reply is only as good as the customer record it can see.
- Most AI is free or bundled, and nobody owns it. DSIT found 47% of businesses able to estimate their AI spend spent nothing in 2025, relying on free tools and AI features built into existing software. That keeps costs down, but it also means AI arrives tool by tool with no one deciding how the pieces fit.
- The real cost is implementation time. Businesses told DSIT that cost was less about tool prices and more about the time, effort and resource needed to implement and embed AI. Skills gaps affect 67% of adopters, and 65% of non-adopters lack a clear business case.
Five steps to make the back office AI-ready
- Pick one admin process end to end. Order to invoice, lead to quote, or new starter to first payslip. Map every step and handoff before choosing a tool.
- Fix the record before you add the assistant. Decide which system holds the truth for customers, products, stock and invoices, and clean that data first.
- Use the AI you already pay for. DSIT's respondents plan to put budget into off-the-shelf applications (61%), task automation (57%) and embedding AI into existing systems (54%). Switch on and configure the features inside your accounting, CRM and helpdesk tools before buying anything new.
- Connect, don't copy and paste. The biggest admin wins come when AI can read from one system and update another through proper integrations, with a log of what it did.
- Keep people in the loop, on purpose. 81% of businesses apply human oversight to AI outputs, and trust is lower for autonomous agents. Decide which steps AI can complete alone and which it only drafts for approval, then measure time and error rates against a baseline after six to eight weeks.
The optimistic bit
Doubling in a year is fast, and 65% of businesses expect their AI budgets to grow. UK firms aren't short of appetite. What separates the modest gains from the meaningful ones is unglamorous: a mapped process, a reliable system of record, and connections that let AI act on real data. Get those right in the back office, and the same foundations carry AI into sales, service and operations next.
Why this matters for Clarity Growth
Clarity Growth helps UK SMEs and scale-ups build connected systems, where websites, ecommerce, CRM, finance and operations share clean data and automation and AI are added where they measurably help. Back-office AI is often where we start: mapping one admin process, choosing the system of record, connecting the tools a business already owns, and configuring built-in AI with sensible approvals and logging. It's exactly the kind of depth the SMF research links to better outcomes, delivered as a small, measurable project rather than another subscription.
Sources
- AI adoption survey report 2026: executive summary — GOV.UK (DSIT) (published 29 October 2026; 3,503 business interviews, January to April 2026: 32% AI use vs 16% in 2025; admin 81%, marketing 71%, IT 66%; impacts, barriers, spending and oversight figures)
- Business insights and impact on the UK economy: 8 October 2026 — Office for National Statistics (BICS Wave 165: 30% of businesses using AI in September 2026; admin and clerical roles most affected for 38% of AI-using firms with 10+ staff)
- Byte-sized growth: Supporting SMEs to get more from AI — Social Market Foundation (19 October 2026; survey of 1,015 UK SME leaders: advanced users report better outcomes; £38 to £47 billion potential SME revenue)
- The state of SME technology adoption in the United Kingdom — OECD (22 April 2026; SMEs strong on process tools, modest on CRM/ERP and advanced technologies)