AI for accountants: five workflows worth automating first

For most accountancy practices, the five workflows worth automating first are chasing clients for records, keying bank statements and receipts, new-client onboarding and AML document collection, the shared inbox, and preparing monthly management packs. In each, the automation does the collecting, sorting and keying, and your team keeps the judgement and the sign-off. Making Tax Digital for Income Tax makes the first one more pressing, because qualifying clients now report every quarter.

For all five, we build with Anthropic’s Claude through its business service, which doesn’t train on customer data and is covered by a data processing agreement. We never use consumer chatbots with client data, and in-house AI on our own hardware in Cardiff is coming soon for work where data shouldn’t leave the building.

1. Chasing client records, including quarterly updates

Today: the chasing runs all year, for year-end, VAT and now Making Tax Digital for Income Tax. Since 6 April 2026, sole traders and landlords with qualifying income over £50,000 in 2024 to 2025 have had to use it. The threshold falls to £30,000 from 6 April 2027 and £20,000 from 6 April 2028, each based on the previous tax year’s turnover from self-employment and property, before expenses. Quarterly updates are due on 7 August, 7 November, 7 February and 7 May. That’s four rounds of requests and reminders a year for each client, usually tracked on a spreadsheet.

1. The automation works from the deadline list in your practice management system and sends each client a request for what you need, in your wording.
2. It chases on a schedule, stops when records arrive, and passes persistent non-responders to a manager.
3. It logs what’s in, so the team has a live view of who’s outstanding.

Stays with a person: the conversation with the client who never sends anything, deciding whether what arrived is complete, and the submission itself, which goes through your compatible software.

Watch: HMRC has said it won’t apply penalty points for late quarterly updates during the 2026 to 2027 tax year, but they still apply to late tax returns. Clients in Making Tax Digital must keep digital records of each item’s amount, date and category, so ask for records in a form that fits.

How it’s delivered: a single project, or built from plan hours on Monthly AI Support.

2. Keying bank statements and receipts

Today: clients with bank feeds are fine. The rest send PDF statements, photos of receipts and forwarded invoices, and someone types them in. If a client’s bank already feeds their accounting software, don’t automate what’s already automated. This is for the gaps.

1. The automation reads statements, receipts and invoices from your upload portal or inbox.
2. It pulls out the date, amount, supplier and VAT, and suggests a category based on how that client’s transactions have been coded before.
3. It prepares them for import, flagging anything unclear: an unreadable receipt, VAT that doesn’t add up, a possible duplicate.
4. A bookkeeper reviews and posts.

Stays with a person: coding unusual items, VAT treatment and posting.

Watch: HMRC says you can’t reclaim VAT using an invalid invoice, a pro-forma, a statement or a delivery note, so the automation flags items where only a statement is held. VAT records must be kept for at least six years.

How it’s delivered: a single project using our document processing, tested on a batch of real client documents before it goes live.

3. New-client onboarding and AML document collection

Today: engagement letter, ID, proof of address, company details, beneficial owners and a letter to the previous accountant. Each item is requested, chased and saved somewhere, and work can’t start until the file is complete.

1. When a new client signs up, the automation sends your engagement letter and a checklist of what’s needed.
2. It chases each item, files it to the client record, and flags obvious problems such as an expired passport or an address that doesn’t match.
3. For companies, it pre-fills officers and people with significant control from the public Companies House register.
4. It assembles the file for whoever signs off customer due diligence.

Stays with a person: the risk assessment, the due diligence sign-off and the decision to go ahead. The Money Laundering Regulations 2017 require you to identify and verify the client and, for companies, to identify the beneficial owners and take reasonable steps to verify them. You can’t meet the beneficial ownership duty by relying solely on the Companies House register. Step 3 is a starting point, not a check.

Watch: the regulations also require written policies and procedures of your own, and your AML supervisor may be a professional body or HMRC. The automation follows your procedures; it doesn’t replace them. Due diligence records must normally be kept for five years after the business relationship ends.

How it’s delivered: a single project.

4. The shared inbox

Today: info@ gets records, questions, forwarded HMRC letters, new enquiries and spam, and someone sorts it all before any of it gets answered.

1. The inbox assistant reads each email and works out the client and the type: records, a query, HMRC correspondence or a new enquiry.
2. It routes each one to the right person and files attachments to the client folder.
3. It drafts routine replies, such as acknowledgements or upload instructions, for checking.
4. It flags anything time-sensitive, such as an HMRC letter with a deadline.

Stays with a person: sending, any tax answer, and anything that looks like a complaint.

Watch: an email sent to the wrong client is likely to be a personal data breach. It must be recorded, and reported to the ICO within 72 hours where it’s likely to pose a risk to people. That’s why replies are drafted for checking, not sent automatically.

How it’s delivered: The inbox assistant can be built on Monthly AI Support.

5. Month-end management pack preparation

Today: export from the client’s accounting software, paste into the Excel template, update the charts, then write the commentary from scratch.

1. Once a person has reconciled and closed the month, the automation pulls the figures into your pack template.
2. It calculates variances against budget and last year and highlights the big movements.
3. It drafts commentary on those movements for the accountant to edit.

Stays with a person: the close, what a movement actually means, and sign-off. AI can write a plausible reason that’s wrong; only your accountant knows the spike was a one-off.

Watch: UK GDPR requires personal data to be limited to what’s necessary. Payroll lines can reveal individuals’ pay, so pass only the figures the pack needs.

How it’s delivered: plan hours. Once the template is built, each extra client is a small set-up job. A single project suits a practice that wants many clients done at once.

How to pick the first one

• Lots of sole traders and landlords? Start with chasing. Take a practice with 200 clients in Making Tax Digital for Income Tax: that’s 800 rounds of requests and reminders a year before year-end work begins.
• Bookkeeping the bottleneck? Start with keying, but only for clients without bank feeds.
• Small team, noisy inbox? Start with the inbox assistant.
• Leave judgement alone. Tax advice and the decision on a client’s ID checks stay with people.

Our free automation audit is up to two hours with Ross going through how your practice works, then a written report of what to automate first, roughly what it would save, and the data-protection points. Our sample report is for a fictional 22-person Cardiff accountancy practice, so it’s a close look at what you’d receive. There’s more on our AI for accountants page.