AI for accountancy firms: what to automate first in 2026

Short answer

UK accountancy firms with 5 to 100 staff should automate chasing and collection first in 2026. That means records for Making Tax Digital quarterly updates, receipts and bank statements, and AML onboarding documents, followed by AI drafts of routine client emails. Practices already using AI report saving 7.1 hours a week, according to Xero's May 2026 survey of 520 UK accountants and bookkeepers. Keep sign-off, advice and AML decisions with people.

Yarigai is a done-for-you AI agency that builds these systems, so read this with that in mind. Every price and figure links to its source below. Read this page as Markdown.

At a glance

QuestionAnswer
Time saved by UK practices using AI7.1 hours a week on average, self-reported (Xero survey of 520, May 2026)
Mid-tier firms with AI in their technology strategy86% of 35 firms (ICAEW, February to March 2026)
Firms with an AI policy or strategy21% of nearly 600 firms worldwide (Karbon, January 2026)
Sole traders and landlords joining MTD for Income Tax in April 2026About 864,000 (ATT)
MTD quarterly update deadlines7 August, 7 November, 7 February and 7 May
MTD for Income Tax thresholdsAbove £50,000 from April 2026, £30,000 from April 2027, £20,000 from April 2028
AML record retention5 years after the client relationship ends (MLR 2017, regulation 40)
Receipt capture already in XeroHubdoc included on the Simple, Ignite, Grow, Comprehensive and Ultimate UK plans
UK entry price for an invoice extraction buildFrom £4,500 plus £299 a month, ex VAT (Softomate, October 2026)

Where AI saves time in an accountancy practice

UK practices that use AI report saving about a working day a week. The time-saving figures are self-reported and come from software vendors, so test them against your own timings.

  • Xero surveyed 520 independent senior accountants and bookkeepers across the UK between 7 and 26 May 2026. Practices reported saving 7.1 hours a week on average, and top performers, firms with net margins of 41% or more, reported up to 10.6 hours. Xero doesn't say whether that is per practice or per person.
  • In the same survey, 87% of practices using AI in daily workflows said their core processes were well documented, against 18% of practices with no plans to use AI. Only 5% expected AI to reduce headcount within a year.
  • ICAEW asked leaders of 35 mid-tier UK firms in February and March 2026. 86% have a technology strategy that explicitly includes AI, but only 17% feel able to assess its impact on their workforce, and 68% expect AI to reduce the need for some early-career roles.
  • Karbon, which sells practice management software, surveyed nearly 600 accounting professionals worldwide in its January 2026 report. 98% of firms use AI and report saving 60 minutes a day per employee, yet only 21% have an AI policy or strategy and fewer than half invest in training.

Across all three surveys, adoption is high while policy and measurement lag behind. Write each process down before you automate it, because every automation needs a written process to follow.

What to automate first

Start where staff time goes on chasing and re-keying, because those tasks follow clear rules and a person can check the output quickly.

1. Chasing clients for records

Making Tax Digital for Income Tax turns one yearly scramble into four quarterly ones for each client in scope. An AI system can remind each client before each deadline, read the reply and attachments, list what is still missing and pass anything unclear to the client manager. Have a person approve the messages before they go out.

2. Receipt and bank-statement capture

Many practices already have capture, since Xero includes Hubdoc on its UK plans. The gap sits around it, with clients who email photos, PDF statements from banks without a feed, and receipts that never arrive. AI can read those documents, suggest a category and post a draft to Xero or QuickBooks for a person to check. Check line items and poor scans by hand.

One London agency, Softomate, publishes an entry price of £4,500 plus £299 a month, ex VAT, for an invoice extraction workflow with one intake route and one document type. For how extraction works, see AI document processing.

3. Client onboarding and AML document collection

The Money Laundering Regulations 2017 decide what evidence a firm must hold, and AI can do the collecting. Under regulation 28, a firm must identify each client and verify their identity using documents or information from a reliable source independent of the client. For a company it must obtain the name, company number and registered office, and identify beneficial owners and take reasonable measures to verify them. Relying only on what is filed at Companies House does not meet the beneficial ownership duty. Electronic identification can count as verification if it is secure against fraud and misuse.

An onboarding system can send one form, collect ID and company documents, check that names and dates match, pull the Companies House record as a cross-check and flag gaps. The risk assessment and the decision to take the client on stay with a person, because the firm must be able to show its supervisor that its measures fit the risk.

Regulation 40 requires a firm to keep due diligence records for five years after the relationship ends, then delete the personal data unless another law requires it to be kept. A system can track that date for every former client and queue deletions for a person to approve.

4. Drafting client emails

AI can draft routine replies from the client file, such as deadline reminders, requests for missing documents and plain-English explanations of a tax bill. A person reads, edits and sends each one, so nothing reaches the client without review. Karbon found 82% of respondents say AI has a positive effect on collaboration, communication and client relationships.

How Making Tax Digital changes the workload

Making Tax Digital for Income Tax started in April 2026 for sole traders and landlords with qualifying income above £50,000. The Association of Taxation Technicians reported that around 864,000 people were expected to join in that first phase.

  • The threshold falls to £30,000 in April 2027 and £20,000 in April 2028.
  • Clients in scope must keep digital records in compatible software and send quarterly updates to HMRC.
  • Standard update deadlines are 7 August, 7 November, 7 February and 7 May, and each update runs from 6 April. For the 2026 to 2027 tax year, the second update, covering 6 April to 5 October, is due on 7 November 2026.
  • Clients with a 1 April to 31 March accounting period can use calendar quarters, but they must choose before their first update and can't switch within the tax year.

For a practice, 200 clients in MTD means 800 quarterly updates a year (200 × 4), on top of each client's tax return. Time one chasing cycle in your own practice. At 15 minutes a cycle, for example, 800 cycles take 200 hours a year (800 × 15 ÷ 60). The thresholds fall twice by April 2028, so the client count rises.

Hours saved by workflow

Only whole-practice figures exist, and both come from self-reported vendor surveys. No published source gives an hours-saved figure for an individual accountancy workflow, so the table gives a method instead of a number.

WorkflowWhat AI doesPublished hours-saved figureHow to estimate yours
Chasing clients for MTD recordsReminds before each deadline, reads replies, lists what is missingNone foundClients in MTD × 4 updates a year × minutes per chasing cycle
Receipt and bank-statement captureReads documents, suggests categories, posts drafts for reviewNone foundDocuments a month × (manual minutes minus review minutes)
Onboarding and AML documentsCollects and cross-checks ID and company documents, flags gapsNone foundNew clients a year × hours per onboarding × share of steps automated
Drafting client emailsDrafts replies from the client file for a person to sendNone foundRoutine emails a week × minutes saved per email
All AI use in a practiceMixed tasks7.1 hours a week (Xero, UK, self-reported); 60 minutes a day per employee (Karbon, worldwide, self-reported)Compare a month of timesheets before and after

To measure one workflow:

  1. Count how often it happens in a month.
  2. Time five examples done by hand.
  3. Time the same five with the automated draft plus a person's check.
  4. Multiply the difference by the monthly count.

Data protection under UK GDPR and the ICO

Client records are personal data, so an AI provider that processes them acts as your processor under UK GDPR. Article 28 says a controller may use only processors that give sufficient guarantees of appropriate technical and organisational measures. The processing must be governed by a contract, and the contract must be in writing.

The ICO's guidance on AI and data protection covers data protection impact assessments (DPIAs) and transparency for AI systems. It was last updated on 15 March 2023, and the ICO says it is under review because of the Data (Use and Access) Act. Karbon found 83% of respondents report growing awareness of data security.

Practical steps for a practice:

  1. Use paid business tiers with a signed data processing agreement, not free consumer chatbots.
  2. Check whether the provider trains on your data and how long it keeps prompts.
  3. Run a DPIA before AI touches client records.
  4. Keep data in the practice's own accounts, with deletion dates that match regulation 40.
  5. Say in your privacy notice that you use AI tools to process client records.

What to keep human

Keep anything that needs professional judgement or creates liability with a qualified person. AI prepares the work, and people decide.

  • Signing off accounts, tax returns and quarterly updates.
  • Tax advice and judgement calls.
  • AML risk assessments, decisions to take on or drop a client, and any suspicion of money laundering.
  • Anything that sends money, emails a client or deletes data.
  • Line items, poor scans and anything the system flags as unclear.

ICAEW found 71% of mid-tier firms believe AI will let them move up the value chain in what they offer.

How Yarigai builds AI for accountancy firms

Yarigai builds AI agents, inbox triage, document reading and filing, invoice chasers, receipts into Xero or QuickBooks, finance dashboards and internal tools that replace spreadsheets. The Xero and QuickBooks work sits in its AI finance automation service.

  • Yarigai builds inside the practice's own accounts and tools. The practice owns the code, the data and the accounts from day one, and there is nothing to sign up for.
  • A person approves anything that sends money, emails a customer or deletes data.
  • AI and hosting bills are paid by the practice at cost, and the monthly running cost is on the quote.
  • A free 30-minute scope call on Google Meet comes first. Then comes a blueprint of about a week, credited to the build, with a written plan, a clickable prototype, the monthly running cost, a fixed price and the payback maths.
  • Builds take 2 to 6 weeks with a demo of working software every week, and a single workflow can go live in about two weeks. Monthly care is optional, and the practice can cancel any month and keep everything.

The guarantee is "If the build isn't profitable for you, we won't build it." If the blueprint's maths doesn't come out ahead within the first year, Yarigai doesn't take the job. For when a fixed workflow beats an agent, see AI agent or automation.

Common questions

Will AI replace junior staff at accountancy firms?

Not quickly, on current survey evidence. Only 5% of UK practices in Xero's May 2026 survey expected AI to reduce headcount within a year. ICAEW found 68% of mid-tier firms expect AI to reduce the need for some early-career roles.

Is it safe to put client data into ChatGPT?

Not into a free consumer account without checking its terms. Under UK GDPR Article 28, an AI provider that processes client records is your processor and needs a written contract with sufficient guarantees. Use a paid business tier with a data processing agreement, check its training and retention terms, and run a DPIA first.

Can AI carry out AML checks for an accountancy firm?

It can collect, read and cross-check the documents, but the firm stays responsible. Regulation 28 of the Money Laundering Regulations 2017 requires verification against a reliable independent source, and Companies House filings alone don't meet the beneficial ownership duty. A person should make the risk assessment and approve each new client.

Does Xero already capture receipts?

Yes. Hubdoc is included on Xero's Ignite, Grow, Comprehensive and Ultimate plans in the UK. A custom build adds value around it, for example by chasing missing receipts. For a client's own invoicing, see custom invoicing versus Xero.

How much does AI automation cost for an accountancy firm?

No independent survey covers it. One London agency's published entry prices, updated October 2026, are £1,500 plus £99 a month for a simple n8n or Make automation and £4,500 plus £299 a month for an invoice extraction workflow, ex VAT. Platform fees and AI usage are billed on top.

How does Making Tax Digital for Income Tax affect accountancy firms?

It moves clients above £50,000 of qualifying income from one return a year to four quarterly updates plus the return, starting April 2026. The Association of Taxation Technicians reported that around 864,000 sole traders and landlords were expected to join in the first phase. The threshold falls to £30,000 in April 2027 and £20,000 in April 2028.

Sources

Checked on or before 7 November 2026. Prices change, so check the live page before you decide.

  1. Xero research reveals UK's top accounting and bookkeeping firms are twice as profitable, with AI adoption at the core, Xero, published 4 August 2026.
  2. Mid-tier firms focused on AI and consolidation, ICAEW Insights, published 22 May 2026.
  3. Karbon State of AI in Accounting 2026 Report shows strategy, training, and governance unlock AI's full potential, Karbon, published 20 January 2026.
  4. Use Making Tax Digital for Income Tax: Send quarterly updates, HMRC, updated 16 October 2026.
  5. Making Tax Digital: 864,000 to move to new digital system next week, Association of Taxation Technicians, published 30 March 2026.
  6. Making Tax Digital for Income Tax: Quarterly update direction, HMRC, updated 27 March 2026.
  7. The Money Laundering, Terrorist Financing and Transfer of Funds (Information on the Payer) Regulations 2017, regulation 28, legislation.gov.uk, checked 7 November 2026.
  8. The Money Laundering, Terrorist Financing and Transfer of Funds (Information on the Payer) Regulations 2017, regulation 40, legislation.gov.uk, checked 7 November 2026.
  9. Guidance on AI and data protection, Information Commissioner's Office, last updated 15 March 2023, checked 7 November 2026.
  10. UK General Data Protection Regulation, Article 28: Processor, legislation.gov.uk, checked 7 November 2026.
  11. Pricing Plans | Xero UK, Xero, checked 7 November 2026.
  12. AI Automation Pricing for UK Businesses: What Every Service Costs and What ROI to Expect in 2026, Softomate Solutions, updated 1 October 2026.

Want the maths for your practice?

Book a free scope call. If the build isn't profitable for you, we won't build it.

Work with us