How many UK accountancy firms use AI? Somewhere between one in five and nearly all of them, depending on which survey you read: the main studies from 2025 and 2026 put the figure anywhere from 21% to 98%. The range isn't a contradiction. It comes from what each survey counts as "use", who it asks and who paid for it, and the gap between those numbers is the most useful thing a managing partner can know.
This guide sets the UK figures side by side, explains why they differ, and then covers what matters for a practice of 5 to 100 people: where AI is actually changing the work, what ICAEW, ACCA, the FRC and the PCRT bodies now expect, whether AI will replace accountants, and a 90-day plan for adopting it without creating a professional-liability problem.
Key Takeaway
Almost every accountant now uses some AI personally, but far fewer firms say AI is in use across the organisation: 21% of tax and accounting firms in early 2025, on Thomson Reuters' figures. Whitehat's reading of the data: that gap is unmanaged AI use, and closing it with a policy, approved tools and training matters more than which tool you buy.
How many UK accountancy firms use AI?
The honest answer is a range. The four most-cited sources give four different numbers, and each is accurate for the question it asked.
98%
Practices using AI "in some way"
Xero vendor survey, Aug 2025
46%
Professional services businesses using AI
ONS, June 2026
21%
Tax & accounting firms using GenAI firm-wide
Thomson Reuters, early 2025
23%
Business services firms using AI
DSIT, spring 2025
Sources: Xero, 3 Nov 2025; ONS, 20 Jul 2026; Thomson Reuters Institute, 2025; DSIT, Jan 2026
No official body publishes an AI adoption rate for accountancy firms alone. The closest official figure is the ONS rate for professional, scientific and technical activities, the industry group that contains accountancy alongside legal, consultancy, architecture and engineering businesses.
Why do the AI adoption figures for accountants disagree?
Because each survey measures a different thing. Before you quote any of these numbers to a partner group or a client, check five things: what counted as "use", whether the respondent was a person or a firm, how the sample was drawn, when fieldwork ran and who paid. The table puts those side by side.
| Source (sponsor) | What counted as "use" | Who was asked, and when | Result |
|---|---|---|---|
| Xero, with Cebr and Censuswide (vendor data: Xero sells accounting software with AI features) | The practice uses AI "in some way" for day-to-day tasks | 515 accountants and bookkeepers in UK practices with turnover up to £500m; online survey, 22–29 August 2025 | 98% use AI in some way; 47% at least weekly |
| ONS Business Insights and Conditions Survey (official statistics in development) | The business uses at least one AI technology from a set list | 38,637 responding UK businesses across most industries (finance and insurance excluded); 15–28 June 2026 | 46% of professional, scientific and technical businesses; 35% of all businesses with 10+ staff |
| Thomson Reuters Institute, 2025 and 2026 reports (publisher that also sells professional AI tools) | The organisation is "already using" generative AI | 2025: 1,702 professionals in 8 countries (12% UK), January–February 2025. 2026: 1,514 in 27 countries (18% UK), October–November 2025. Both screened for familiarity with AI | 21% of tax and accounting firms (2025 report). Across all professional-services respondents, organisation-wide use rose from 22% to 40% (2026 report) |
| DSIT AI Adoption Research (UK government) | The business currently uses at least one AI technology | 3,500 UK businesses with 5+ employees, telephone interviews, 12 February–2 May 2025 | 23% of business services and administration firms; 16% of all businesses |
Sources: Xero media release and methodology note, 3 November 2025; ONS, Artificial intelligence in UK businesses: 2023 to 2026, Figure 3 data; Thomson Reuters Institute 2025 and 2026 reports, methodology sections; DSIT, AI Adoption Research, published 28 January 2026. Analysis: Whitehat.
Read across the rows and four explanations account for most of the spread from 21% to 98%.
- Person versus firm. Xero asked individual accountants whether their practice uses AI "in some way", which includes one person using a free chatbot. Thomson Reuters asked whether the organisation is using generative AI as an organisation. The ONS sees the same split economy-wide: 55% of employees say they use AI for work or education, against 35% of businesses.
- How broad "AI" is. The ONS and DSIT count any technology on a list, from text generation to machine-learning data processing. "Using AI in some way" is broader still; "already using GenAI" across the organisation is narrower.
- Who gets asked. Thomson Reuters drew its sample from its own lists and screened out people unfamiliar with the technology. Xero's respondents are practitioners reachable through a research panel, not a random draw of firms. The ONS and DSIT sample from business registers, which is why their figures are lower and more representative of the whole market.
- When. Adoption is moving fast enough for fieldwork dates to matter. The ONS rate for businesses with 10 or more staff rose from 25% in June 2025 to 35% in June 2026, so a survey from spring 2025 undercounts today's position.
Only one of the four sources is UK-only and accountancy-only, and it was paid for by a software vendor, so its headline sits at the top of the range. The two government sources are the most representative but cannot isolate accountancy. The Thomson Reuters series is the only one that splits out tax and accounting firms, and it is international.
What this means for your firm: assume most of your people are already using AI, whatever the firm has decided. The question for a partner group isn't whether to adopt it, but whether that use is visible, approved and supervised.
Where is AI used in UK accounting practice now?
AI is changing the preparatory and processing layers of accountancy work first, and leaving judgement with the accountant. Xero's respondents said AI let them complete routine tasks in 31% less time on average, and 46% reported productivity gains. ACCA's 2026 guidance to practices focuses on drafting, research and data work, with human review at the points that carry professional responsibility.
| Service line | Largely automated | Augmented (AI drafts, accountant decides) | Still human-led |
|---|---|---|---|
| Bookkeeping | Receipt capture, transaction coding, bank reconciliation matches | Exception queues, month-end commentary | Client-specific judgements on treatment |
| Tax | Data extraction from records into returns | Research summaries, first-draft client letters | Advice, positions taken, sign-off |
| Audit | Full-population data analytics | Risk identification, anomaly flags, working-paper drafts | Professional scepticism, conclusions, opinion |
| Advisory | Management-pack assembly | Forecast scenarios, board-paper drafts | The conversation with the client |
Source: Whitehat assessment, informed by ACCA In Practice (March 2026), the FRC's AI in audit guidance (2025 and 2026) and Xero (November 2025).
The time saved is real but modest per task, and it compounds. Xero put the saving at almost 19 hours a week for the average practice. What firms do with that time is where they differ: Xero reports that 76% of practices said AI had influenced their hiring strategy, towards advisory, technical and communication skills.
What we'd tell a managing partner: three questions, answered from the evidence
What are UK firms actually doing with AI? Mostly embedded automation, not bespoke generative AI. The everyday uses are extracting data from invoices and receipts, coding transactions, reconciliations, summarising documents, tax research and first drafts (Thomson Reuters; ACCA, December 2025). That is why the adoption figures above range so widely: "using AI" usually means AI inside software the practice already runs, not staff putting client work through a chatbot. When they do, the risk is that it is a general-purpose tool the firm never approved.
What is the most common mistake? Allowing tools before setting the rules. Four failures recur in the UK guidance:
- Uncontrolled use. ACCA names unchecked AI use as one of the major risks to confidential client data, and says every firm should have firmwide policies setting out when an AI tool can and cannot be used (ACCA, March 2026). Removing a client's name is not always enough: an unusual or high-profile client can still be identified (ICAEW).
- Trusting fluent output. AI can oversimplify and produce plausible-looking output that gets relied on when it should not be (FRC, March 2026), and the PCRT guidance warns of hallucinated content such as non-existent case law. In Sage's survey of 1,000 UK accountants and bookkeepers, 91% of regular AI users said they regularly need to review or adjust AI output before using it (Sage, June 2026).
- No audit trail. Nobody can show which tool was used, what went in, or who checked the result.
- Clients not told. PCRT suggests disclosing AI use in the engagement letter, and telling the client before the work where AI is fundamental to it (ACCA, September 2026).
What should a managing partner do first? Not buy a tool. In this order:
- Name an owner and write a one-page AI policy: permitted uses, prohibited data and mandatory human review.
- Approve specific tools for specific tasks. A transcription tool is not automatically fit for tax research.
- Make verification evidenced. The January 2026 PCRT guidance advises confirming that any case law or legislation an AI tool cites actually exists, and keeping an audit trail, such as copies of sources and screenshots of prompts, to evidence reasonable care (PCRT AI guidance (PDF)).
- Train everyone before granting access.
- Update engagement letters, and ask your professional indemnity broker in writing whether your planned use changes your cover.
- Then pilot one high-volume workflow, such as bank-feed coding, measure the hours it saves, and only then widen it.
That last step is where most of the value is proven or disproven, and where an outside view helps.
Planning where AI fits in your practice? Our AI consultancy and implementation team helps professional firms choose, govern and roll out the right tools.
See AI Consultancy & ImplementationWhat do ICAEW, ACCA, the FRC and PCRT expect from firms using AI?
The UK has not written an AI rulebook for accountants. Instead, the regulators and professional bodies have said how existing duties of competence, confidentiality and due care apply when AI is involved. The common thread is simple: you can use AI, but you cannot hand it your responsibility.
- ICAEW: its updated Code of Ethics, enforced from 1 July 2025, builds the impact of technology into members' ethical duties; a 2026 edition, in force from 1 July 2026, adds new sections on tax planning (ICAEW, April 2026). ICAEW announcement.
- ACCA: its March 2026 guidance for practices recommends firmwide AI policies, care over what client data goes into which tool, vendor due diligence, telling clients in the engagement letter which AI tools will be used, and AI training within CPD. ACCA In Practice.
- FRC: the audit regulator published guidance on AI in audit in June 2025 and a second piece, on generative and agentic AI tools, in March 2026. Both centre on documenting the tools and on how auditors gain appropriate confidence in AI outputs. FRC, March 2026.
- PCRT bodies: the seven bodies behind Professional Conduct in Relation to Taxation issued topical guidance on the ethical use of AI tools in January 2026. Members remain responsible for tax work produced with AI. PCRT AI guidance (PDF).
Does the EU AI Act apply to UK accountancy firms? Only where a firm places AI systems on the EU market or its AI outputs are used in the EU. For most UK practices serving UK clients, the professional-body guidance above is what applies. Firms with EU clients should note that the EU's Digital Omnibus on AI, Regulation (EU) 2026/1744, pushed most high-risk obligations back to December 2027.
What are the risks of AI for an accountancy firm?
The risks are professional before they are technical. ACCA names four areas: governance, client confidentiality, client knowledge of AI use, and training. Three failure modes cause most of the trouble.
The Cost of Getting It Wrong
Confidential data in public tools: pasting client records into a free chatbot can breach confidentiality duties and data-protection law. Know where each tool stores data and whether it trains on it.
Unverified output: generative AI produces fluent errors, including invented rules and figures. Advice built on a hallucination is still your advice, and ACCA warns it can expose a firm to negligence claims.
No audit trail: if you cannot show which tool produced what and who reviewed it, you cannot evidence due care. The FRC's guidance makes documentation central for auditors.
Skills are the other constraint. In Xero's survey, the top barriers to using more AI were a lack of training (36%) and a lack of experience (32%), not cost.
Will AI replace accountants?
Not on current UK evidence, but it is changing what firms hire for and how they price. The ONS found that among businesses using AI to improve operations, 63% reported no change in headcount, 1% an increase and 6% a decrease. British Progress, analysing ONS survey data across hundreds of occupations, found no evidence that AI had replaced jobs at scale in the UK.
For a firm, the planning question is capacity rather than headcount. If routine compliance work takes a third less time, the firm either wins more work with the same team, moves people into advisory, or finds clients expecting lower fees for the same output. Decide which of those you want before the market decides for you. The same pattern is visible in how AI is reshaping consulting and AI in investment banking: tasks change first, roles follow more slowly.
A 90-day AI adoption plan for a UK accountancy firm
This plan is designed for a practice of 5 to 100 people. It puts governance first because that is what the professional bodies ask for, and because it makes everything after it faster.
Days 1–30: find out what's already happening
Ask every team which AI tools they use, for what, and with which client data. Write a one-page policy: approved tools, prohibited data, review rules. Add an AI clause to your engagement letter template, as ACCA suggests.
Days 31–60: pilot two workflows
Pick one processing task (say, bank reconciliation exceptions) and one drafting task (say, first-draft client letters). Measure time per job before and after, and log every error the reviewer catches. Check vendor answers on data storage and model training before you sign.
Days 61–90: train, decide, and reset the plan
Run training on prompting, spotting bad output and the confidentiality rules, and record it as CPD. Keep the pilots that saved time without adding review risk. Then decide what the freed capacity is for: more clients, more advisory work, or both.
Once the internal work is running, the growth side follows. Our guide to digital marketing for accountants covers how firms turn advisory capacity into new clients, and content marketing for accountants covers what to publish.
Frequently asked questions about AI in accounting
How many UK accountancy firms use AI?
Between about 21% and 98%, depending on the survey. Xero's vendor survey found 98% of UK practices use AI in some way; Thomson Reuters found 21% of tax and accounting firms used generative AI across the organisation in early 2025. The ONS puts AI use at 46% for professional, scientific and technical businesses.
Will AI replace accountants?
Not on current UK evidence. ONS data show most businesses using AI report no change in headcount, and British Progress found no evidence of job replacement at scale. AI is changing which tasks take time, and so what firms hire for.
Can accountants put client data into ChatGPT?
Only with care. ACCA warns that unchecked AI use can compromise confidential client data. Check where a tool stores data and whether it trains on inputs, use business versions with the right terms, and set a firm policy on what may never be entered.
Is there UK regulation of AI in accounting?
There is no AI-specific statute for accountants. ICAEW's Code of Ethics (enforced from July 2025), ACCA's practice guidance, the FRC's AI in audit guidance and the PCRT bodies' January 2026 AI guidance set how existing professional duties apply.
Where should a small accountancy firm start with AI?
Start by finding out what staff already use, then write a short policy and pilot two workflows with measured before-and-after times. Training and an engagement-letter clause come next. Tools matter less than governance at this stage.
Planning AI adoption in your firm?
Book a short call with Clwyd to talk through where AI fits in your practice, what to govern first, and how to turn the time it frees into growth.
Sources: Xero (3 Nov 2025); ONS, Artificial intelligence in UK businesses: 2023 to 2026 (20 Jul 2026); Thomson Reuters Institute 2026 AI in Professional Services Report; Thomson Reuters 2025 report (PDF); DSIT AI Adoption Research (28 Jan 2026); ACCA (Mar 2026); British Progress (2025); Regulation (EU) 2026/1744.
Clwyd Probert
Founder, Whitehat
Clwyd Probert is the founder of Whitehat, a London-based SEO, AI and inbound marketing agency and HubSpot Diamond Partner. He works with B2B and professional-services firms on AI adoption and growth.