AI for Accountants: The Practical Guide

AI for accountants means using AI across the work of an accounting or tax practice: preparing and reviewing the books, researching and drafting tax positions, testing and documenting audit work, producing the analysis behind advisory conversations, and running the firm itself. It is a guide for the practitioner, the person whose name goes on the work, rather than for the finance team inside a company, whose questions are covered in AI accounting.

The difference matters. A company’s finance team asks which tools automate its own close. A practice asks something harder: how to use AI on many clients’ data, under professional rules, without weakening the sign-off its reputation rests on, and without pricing itself out of the time it just saved.

Why the case is stronger in accounting than almost anywhere

Most sectors adopt AI to do the same work with fewer hours. Accounting practices have a more pressing reason: there are not enough accountants to do the work that already exists.

The AICPA’s 2025 Trends report found US graduates with a bachelor’s or master’s degree in accounting fell to 55,152 in the 2023-24 academic year, down 6.6% on the year before, after steeper falls in the two years prior. The shortage is not American alone. In a 2024 poll by AICPA and CIMA with the Global Accounting Alliance and the Confederation of Asian and Pacific Accountants, 28 of 33 accountancy bodies said their jurisdictions had a shortage of professional accountants.

That changes the economics of adoption. For many firms the realistic alternative to AI is not the same work done by more people. It is work turned away, deadlines stretched and partners doing preparation they should be reviewing. Thomson Reuters’ 2025 Future of Professionals report put the expected saving at around five hours a week per professional using AI, and valued the time saved across the US CPA profession at roughly $12 billion a year.

Where it fits, service line by service line

• Bookkeeping and compliance. Reading documents into structured data, categorising transactions, reconciling accounts and flagging what does not match. This is where AI is most mature, and the principle that sorts real from oversold is the one set out for in-house finance: tasks with a checkable right answer work well, tasks without one need a person.
• Tax. Researching positions, summarising legislation and guidance, drafting client letters and first-pass computations for review. Useful, provided every cited provision is checked against the source. Tax research is where confident, wrong text is most expensive.
• Audit and assurance. Testing whole populations of transactions rather than samples, flagging anomalies for the auditor to investigate, and drafting documentation of the work done. AI widens what can be examined; it does not replace the auditor’s judgment about what the evidence means.
• Advisory. Turning a client’s numbers into analysis worth discussing: variance explanations, scenarios, cash-flow views. The saving from compliance work is what funds advisory time, which is where most firms want to grow.
• Running the practice. Engagement letters, document requests, reminders and chasing, deadline tracking and time capture. Unglamorous and often the quickest win, because it gives back hours across the whole firm.

The rules that keep it professional

Four rules separate responsible use from the kind that ends in a complaint.

A named reviewer for every output that leaves the firm. AI prepares; a qualified person reviews against the source and signs. That was always the model for junior work, and it is the model for AI work.

Client data only in tools with professional-grade terms. Business terms that exclude training on your data, sensible retention, a known list of subprocessors and a written confidentiality commitment. In Germany, tax advisers are bound by professional secrecy under Section 203 of the Criminal Code, and may involve service providers only under the conditions in Section 62a of the Tax Advisers Act, including a confidentiality obligation agreed in text form. Other jurisdictions have their own equivalents, and the principle is the same everywhere: the tool vendor is part of your confidentiality chain. The wider questions are covered in data sovereignty.

The working must be traceable. Every automated entry, computation or conclusion needs a visible basis: which document, which rule, which figure. A file that cannot show its working is not a file you can defend.

Clients should know. Tell clients how AI is used on their work and how it is checked. Most will be reassured; some will have their own requirements, and it is better to learn them before the work is done.

What happens to the junior pipeline

The routine work AI absorbs is the work juniors used to learn on. Remove it without a replacement and a firm ends up with capacity but no future reviewers.

There is a more encouraging reading. In Thomson Reuters’ 2026 survey, tax professionals expected the time it takes a junior to reach trusted judgment to shorten by about a year, while legal professionals expected it to lengthen. The difference probably lies in how the work is redesigned. Juniors who check AI output against the source documents, and have to explain why a figure is right, see more cases in a month than they once prepared in a season. Juniors who simply stop doing the work learn nothing.

The design choice is the firm’s. It should be made deliberately rather than discovered in five years.

What it does to fees

Under hourly billing, time saved is revenue lost. Under fixed or value-based fees, time saved is margin. That asymmetry is pushing firms towards fixed fees for recurring compliance work, with the saving reinvested in advisory time that clients value and are willing to pay for.

The practical question for each service line is simple: if this work took half the time next year, would the firm earn more or less? If the answer is less, the pricing needs to change before the tools do. The same pressure is reshaping consulting, and it is the central commercial question in AI for professional services firms generally.

Where to start

Pick one recurring deliverable that every client receives, such as the monthly management pack or the year-end document request, and redesign it end to end with AI inside and a named reviewer at the end. Put the client data rule in writing before you begin. Measure turnaround, error rates found at review and the hours recovered, and decide in advance where those hours go.

Then repeat it for the next deliverable. The firms getting value are not the ones with the most tools. They are the ones that have wired one workflow properly and then another.

If you want AI in one recurring deliverable first, tell Daniel which one, what your practice already uses, and what you need the practice to do for you. Book 30 minutes with Daniel.

Frequently Asked Questions

How can accountants use AI in their practice?

Across five service lines: bookkeeping and compliance, tax research and drafting, audit testing and documentation, advisory analysis, and running the practice through engagement letters, document requests, deadlines and time capture. In every case AI prepares the work and a qualified person reviews and signs it.

Will AI replace accountants?

The evidence points the other way for now. The profession has a shortage: the AICPA found US accounting graduates fell 6.6% to 55,152 in 2023-24, and 28 of 33 accountancy bodies in a 2024 global poll reported shortages. For most firms AI is a way to handle work they would otherwise turn away, with accountants moving towards review and advisory.

Is it safe to put client data into AI tools?

Only into tools with professional-grade terms: no training on your data, controlled retention, known subprocessors and a written confidentiality commitment. In Germany, tax advisers are bound by professional secrecy under Section 203 of the Criminal Code and may involve service providers only on the conditions in Section 62a of the Tax Advisers Act. Consumer AI tools are not suitable for client data.

How does AI affect how accounting firms charge?

Under hourly billing, time saved by AI becomes lost revenue; under fixed or value-based fees it becomes margin. Many firms are moving recurring compliance work to fixed fees and reinvesting the saved time in advisory work clients value more.

How should accounting firms train juniors when AI does the routine work?

Redesign the routine work rather than removing it. Juniors who check AI output against source documents and explain why figures are right see more cases faster. Thomson Reuters’ 2026 survey found tax professionals expect the time to trusted judgment to shorten by about a year, which depends on that redesign being done well.

What is the best first AI project for an accounting practice?

One recurring deliverable every client receives, such as the monthly management pack or year-end document request, redesigned end to end with AI inside and a named reviewer at the end. Set the client data rule first, then measure turnaround, errors caught at review and hours recovered.