AI for accountants: less chasing, more time for advice

Where AI for accountants earns its keep: five workflows

These examples cover jobs you repeat every week or quarter. They are candidates for a scoped pilot, not claims about completed accountancy projects. The final workflow depends on your systems, permissions and the checks agreed with your team.

1. Client onboarding chasers

A new client says yes. Then the chase starts: the signed engagement letter, ID documents, proof of address, last year’s accounts and tax return.

A tracker can show what each new client still owes you, prepare reminders on your timetable and record replies. Approved documents can be filed in the right place. A question or a new identity document should go to a named person; the pilot should test those hand-offs before automated reminders are enabled.

The AI should not decide whether identity checks pass. Your team retains responsibility for client due diligence, unresolved identity questions and the records required by the Money Laundering Regulations.

2. Records collection for Making Tax Digital

For clients in scope, Making Tax Digital for Income Tax adds four quarterly updates to the year. HMRC says sole traders and landlords with qualifying income over £50,000 in the 2024 to 2025 tax year should have started from 6 April 2026. The threshold falls to £30,000 from 6 April 2027 and £20,000 from 6 April 2028. The standard quarterly deadlines are 7 August, 7 November, 7 February and 7 May.

A records workflow can prepare reminders against the applicable dates, sort incoming statements and receipts, and flag gaps. Your team confirms which clients are in scope and resolves uncertainty. Updates still go through compatible MTD software. Check HMRC’s eligibility guidance and quarterly update instructions when setting the timetable.

3. Bookkeeping queries

Most practices keep a list of unexplained transactions. Someone turns that list into emails, one client at a time.

An AI assistant can draft those questions in plain English from the transaction export, collect the answers and suggest a category for each line. A bookkeeper approves or changes every suggestion before anything is posted.

Check what your bookkeeping software already does first. ICAEW notes that machine learning is increasingly built into accounting software, and that this is how smaller organisations are most likely to adopt AI. The best first step may be a feature you already pay for.

4. Management report drafting

Once the numbers are reconciled, AI can draft the commentary from the approved figures: the main movements against last month or budget, plus questions for the client.

The reviewer checks every number against the source. ICAEW warns that outputs may be inaccurate or misleading. A useful instruction is to write “not explained by the data” rather than guess at the reason for a change. No draft should go out unreviewed.

5. Copilot in Excel

Copilot can help analyse and edit workbooks, generate formulas and prepare charts. Available features depend on the subscription and organisation settings. In an initial workshop, work on a copy of a trial balance export or reporting template and check the formulas, source ranges and results before relying on them. Microsoft’s Excel guidance explains the current features.

File access needs attention too. Microsoft says its business Copilot can surface organisational information a user already has permission to view. An overshared SharePoint folder therefore remains a problem. Review access with whoever runs your IT before rollout. Our Microsoft Copilot training can cover spreadsheets, commentary and reporting with source checks.

Using AI for accounting firms without breaking confidentiality

A new tool does not change your professional duties. ICAEW’s guidance calls for human oversight and keeping confidential client and internal information out of public AI tools.

Assess the personal data involved before choosing a tool. A data protection impact assessment is required where processing is likely to create a high risk to people. Record how you assess that risk, including when you decide a DPIA is not required. The ICO’s AI guidance is under review following legislative changes, so your compliance lead should check the current position.

Before a practice pilot, agree five controls with your compliance lead and IT provider:

• Approved accounts. Use accounts and settings the firm has reviewed; a business subscription alone does not establish that a particular use is appropriate.
• Only the data the job needs. Begin with fictional or approved anonymised files and limit access to the agreed purpose.
• A named reviewer. Decide who checks drafts and authorises anything client-facing.
• Agreed records and retention. Decide what evidence of each run and approval is necessary, who can access it and how long it is retained.
• Documented data flows. Record the information each tool can reach, its suppliers, storage and access arrangements.

Confirm the parties’ data-protection roles and any required contractual terms before connecting live client information.

Microsoft says prompts, responses and data accessed through Microsoft Graph in its business Copilot are not used to train its foundation models. That helps, but it does not replace your own access controls. Our AI policy template gives your team written rules to start from. For more on sign-off, read our article on keeping professional judgement visible when your practice uses AI.

AI tools for accountants in the UK: what we use

We are independent, and naming a tool does not mean its maker endorses or partners with us. These are possible tools to assess against a practice’s requirements:

• Microsoft 365 Copilot and Copilot Chat for drafting, Excel analysis and meeting notes, where your licences include them.
• Power Automate for chasers, approvals, reminders and moving files between approved Microsoft services.
• ChatGPT or Claude on business accounts for drafting and summarising, with shared instructions the whole team uses.
• The AI already inside your practice and bookkeeping software, which we check before suggesting anything new.

What we would build first in your practice

We would usually start with the records chaser. It runs every week, the numbers are easy to count, and the AI never makes a professional decision.

1. Free 15-minute call. We agree which chase to look at first.
2. AI audit, £750 fixed price. We look at how work moves through the practice and give you ranked opportunities, the risks and a written plan. If you are not sure your data and tools are ready, start with our AI readiness assessment.
3. Build and test. We quote after we agree a written scope. The proposed first stage uses fictional or approved anonymised files. Connecting live information and enabling messages need the practice’s agreement and a named reviewer.
4. Train and hand over, or we run it. Your team gets the documentation and training to use, check, pause and maintain the workflow. Or we look after it through managed AI, from £250 a month plus a one-off setup fee from £1,750.

On a records chaser we would track three numbers before and after: records outstanding at the start of each quarter, chase emails your staff send by hand, and days from first request to a complete file.

What it costs

Prices are in pounds sterling. Your written quote confirms VAT and payment details, including for practices in the Republic of Ireland. Software licences are extra and paid separately. The audit plan is yours to use with us, your own team or another supplier. Full detail is on our prices page.

The people behind Blue Canvas

Blue Canvas AI Ltd is registered in Northern Ireland (company number NI737502). Phil Patterson, founder and director, has helped teams adopt AI since 2022, after years in business development and venture capital. Davy Anthony, our AI Implementation Consultant, brings process improvement and data analysis experience. Meet the team.

Bring us the job your team chases most

Book a free 15-minute call. Tell us which client records you chase, which software holds them and who signs the work off. If AI is not the right fix, we will say so.

Book a free 15-minute call

FAQ

Can accountants put client data into ChatGPT?

Not into public AI tools. ICAEW’s guidance is to keep client and confidential internal data off them. Use business accounts whose data controls you have reviewed, share only what the task needs and check whether the workflow needs a data protection impact assessment.

Will AI replace our bookkeepers?

The proposed workflows support drafting, sorting and chasing. People retain responsibility for the work. ICAEW’s guidance calls for human oversight; automation should be assessed against accuracy and usefulness, not assumed staff reductions.

Do we need Microsoft 365 Copilot licences?

No. Several workflows on this page, such as chasers and document sorting, do not need Copilot. If you want Copilot in Excel, Microsoft says it might not be included with your Microsoft 365 subscription. We check what your licences include before we scope anything.

Can AI submit Making Tax Digital updates?

Updates must be submitted through compatible Making Tax Digital software. The workflows described here help collect and organise records; your practice remains responsible for checking and authorising submissions.

How much does it cost to get started?

The first call is free and takes 15 minutes. Our AI audit is a fixed £750 and does not commit you to anything else. Builds are quoted after we agree a written scope.

Do you work with practices outside Northern Ireland?

Yes. We are based in Derry and work with businesses across Northern Ireland, Great Britain and the Republic of Ireland. Most work is done remotely, and in-person sessions can be agreed before work starts. Our prices are in sterling, and your quote confirms currency and VAT for a practice in the Republic of Ireland.