AI for accountancy practices in the UK: where it helps, and how to keep client data safe
Manoj Tutika · Published 16 November 2026 · Last checked 16 November 2026
Working on this yourself? Talk to Mark, our co-founder. We’re here to solve it with you.
Where does AI help a UK accountancy practice today?
The honest answer from the profession is that AI helps, but less than the marketing suggests. In AccountingWEB research run with Sage in 2026, just 7% of respondents said AI had delivered a transformational impact, and 53% reported minimal or moderate benefits. Data security (62%) and a lack of skilled people (48%) were named as the biggest obstacles.
The gap is not the technology. Most tools help inside one client file at a time, while a practice's workload is the same check repeated across 80, 150 or 400 clients. AI pays off when it does that repeated groundwork on a schedule, for every client, and hands your team a short list of findings with the evidence attached.
Which tasks are ready, and which stay with the practice?
The more a task can be checked against the ledger itself, the better an agent does it. The more it depends on knowledge that sits with the client or on professional judgement, the more it stays with you.
| Work | What an agent does | What stays with the practice |
|---|---|---|
| Bank coding and VAT codes | Reviews every new transaction for every client each week | Deciding the treatment of an unusual item |
| Suspense and clearing accounts | Explains each balance and proposes the journal | Approving the journal |
| Month-end readiness | Marks each step passed, needs you or failed, with evidence | Signing off the management accounts |
| VAT return preparation | Ties the return to the ledger and lists late or odd entries | Submission under Making Tax Digital and the client conversation |
| Client queries | Drafts the list of missing receipts and open questions | Tone, timing and the relationship |
| Tax and advisory | Pulls figures, history and comparisons | Advice, judgement and PCRT responsibility |
What do ICAEW and the PCRT bodies say about AI and client data?
The professional bodies have been consistent. In January 2026 the bodies behind Professional Conduct in Relation to Taxation (PCRT), ICAEW among them, published topical guidance on AI in tax. It says that entering a client's data into a publicly available AI model without consent is likely to breach client confidentiality.
On 9 November 2026 ICAEW's regulation team went further. Confidentiality is one of the five fundamental principles in the ICAEW Code of Ethics and applies in full to AI. Firms with a ring-fenced AI system still carry risk when staff use public tools or AI features built into other software. A breach of an internal AI policy is not automatically reportable to ICAEW, but it becomes reportable where it is also a breach of the Code. ICAEW asks firms to review subsection 114 of the Code, set clear generative AI policies, train staff and review the policies regularly, and to consider UK GDPR whenever client data is involved.
ACCA, discussing the same PCRT update, suggests disclosing AI use in the engagement letter, and telling the client before work starts where AI is central to it.
What should your practice's AI policy cover?
- Which tools are approved, and a clear rule that client data never goes into a public chat tool.
- Where each tool stores and processes client data, and the written contract that covers it.
- Who reviews AI output before it reaches a client, HMRC or Companies House.
- Which clients an agent may only read, and which it may write to with approval.
- Engagement letter wording on AI use, consent and how you handle an objection.
- A record of what the agent did, who approved it and when.
- Training, and a review date, because the tools change every quarter.
How does Xero's own AI change the picture?
At Xerocon London on 8 July 2026, Xero presented JAX as its agentic AI and announced Partner Hub, available from October, showing live book health, work status and month-end readiness across a practice's client portfolio. Xero also announced document chasing for missing receipts, Bill Protection in beta and an integration with Anthropic's Claude.
For practices whose clients are all on Xero, that is a useful baseline. Two questions remain. Does the tool show you the status, or does it do the work and bring you only the exceptions? And what about the clients on other ledgers? Many UK practices run a mixed book of clients, and a separate tool per ledger means separate checks, policies and reviews.
What does the gain look like across 150 clients?
Take a practice with 150 bookkeeping clients and a monthly readiness review of 15 minutes per client. That is 37.5 hours a month of checking before anyone looks at an actual problem. When an agent runs the review for every client and your team spends ten minutes on each of the 30 clients with findings, the same review takes five hours. The other 32 hours go to client work, advice or growth without hiring.
Quality improves too. An agent does not skip the check in a busy January, and it looks at every transaction rather than a sample. Start with one check, read-only, and compare its findings with your team's for a month, as our guide to AI in accountancy describes.
What do Thred's agents do for a UK practice?
Thred is a team of AI agents that takes the recurring finance work off your desk, on the accounting software your clients already use. For practices it runs in one portal across every client. You connect Xero, Exact Online or Moneybird once, ask a question in plain language, and the agent picks the right client from the conversation.
Every write is shown as an approval card with the client's name, the accounts and the amounts, and goes into the ledger only after you approve it. Per client you choose read-only or approval for each write, and every action is recorded. The scheduled Bookkeeping Quality and Close agents run on Exact Online today, checking each client's books on a schedule and reporting findings with evidence. Data is hosted in the EU, with a data processing agreement in every contract.
Behind them stands the full team. The Bookkeeper keeps the record up to date, The Accountant handles filings, The Analyst monitors performance and variance, The Collector and The Payer manage receivables and payables, The Treasurer manages cash, The Connector finds context in connected tools, and The CFO is the finance lead for clients who have none. Read how this works in AI for accounting firms or see Thred for accounting firms.
Frequently asked questions
How is AI used in accounting in the UK?
Most UK practices use AI for groundwork: reading documents, coding bank transactions, reviewing VAT codes, drafting client queries and checking month-end readiness. In AccountingWEB research published with Sage in 2026, 7% of respondents reported a transformational impact and 53% minimal or moderate benefits, so the gains so far are uneven.
Is it safe to put client data into ChatGPT?
Not into a public AI tool without the client's consent. The PCRT guidance on AI in tax, published in January 2026, says that entering a client's data into a publicly available AI model without consent is likely to breach confidentiality. Use tools under a written contract that tells you where data is stored and who can see it.
What does ICAEW say about AI?
ICAEW says the confidentiality principle in its Code of Ethics applies in full when accountants use AI. In November 2026 it advised firms to set clear generative AI policies, train staff, review the policies regularly and think carefully before entering any client information into an AI tool.
Do I need to tell clients I use AI?
ACCA, discussing the January 2026 PCRT update, suggests disclosing AI use in the engagement letter and telling the client before work starts where AI is central to it. Agree any restriction a client asks for in writing, and decide in advance how you handle an objection.
Will AI replace accountants in the UK?
No. AI takes over checks and preparation, while judgement, advice and professional responsibility stay with the accountant. The practical effect is capacity: the same team covers more clients, and reviews findings instead of doing every check by hand.
Which AI tools work with Xero for practices?
Xero has its own AI, JAX, and in July 2026 announced Partner Hub, a view of book health and month-end readiness across a practice's clients. Third-party tools add data capture, review and agents that work across clients. Thred connects to Xero, Exact Online and Moneybird from one practice portal.
Sources
- Sage, How to adopt AI in your accountancy practice in 2026 (AccountingWEB research figures)
- AccountingWEB, AI in Accounting 2026: State of the Nation report
- ICAEW, AI and client data: complying with the principle of confidentiality (9 November 2026)
- ICAEW, PCRT bodies publish guidance on using AI in tax (January 2026)
- ACCA In Practice, Engagement letters, AI and PCRT (September 2026)
- Xero blog, Xerocon London 2026: the agentic era of small business finance (8 July 2026)
Read next
- AI and Making Tax Digital for Income Tax: what a practice can automate, and what it must sign off
- AI in accountancy: what works today in a compilation practice
- Thred vs Sage Intacct: a Sage Intacct alternative that keeps you on Xero and adds AI agents
- Thred vs Silverfin: a Silverfin alternative for UK practices that gets client books right before year end
- How to grow your accounting firm without hiring more staff
- What does Xero JAX do? Features, pricing and limits, checked 16 November 2026
Where Thred comes in
Make your firm more efficient with AI, so you have more time for relationships, new customers and advisory work.
