AI for accountants and bookkeepers: what it does well and the rules that still apply

By Ranjit Bhinge, founder of Blur Studio · Updated October 24, 2026 · 7 min read

Short answer: Accountants use AI to pull data from receipts and bank feeds, sort transactions, research tax questions and draft client emails and memos. Tools include the AI agents in QuickBooks, Dext for document capture and Blue J for tax research. The data duties do not change. Tax preparers fall under the FTC Safeguards Rule, IRS Publication 4557 asks them to keep a written security plan, and tax return data may need client consent under Section 7216 before it goes to an outside tool.

You run or work in an accounting or bookkeeping firm and want to know what AI can actually do safely with client financial data.

I'm Ranjit, and I run Blur Studio: 15+ years building software, 120 apps taken live, and clients who have gone on to raise $10M+. I build AI systems and processes that help businesses do a lot more with a lot less time. I also work differently from most of what else is out there, more on that further down this page. Say hello.

How are accountants using AI?

The best uses are tasks with a clear pattern and a person who checks the result.

Capturing receipts and bills

Clients send receipts, bills and statements by email, phone photo or upload. AI reads each document, pulls out the date, supplier, amount and tax, and posts it to the books. Staff then review the entries they flag.

Sorting transactions

AI can suggest a category for each bank line from past entries. It learns the client's usual suppliers. Staff spend their time on the odd items.

Researching tax questions

AI research tools answer a tax question in plain words and point to the sources behind the answer. A staff member still reads those sources before advising a client.

Drafting client emails and memos

AI writes a first draft of a document request, a year-end letter or a short tax memo. The accountant edits it and sends it. This saves time on the writing and keeps the judgement with the accountant.

Chasing missing documents

An AI workflow can check which clients still owe documents and send polite reminders. It can also answer simple status questions, such as whether a return has been filed. My page on client portals covers how firms give clients one place to upload files and see progress.

What AI tools do accountants use?

Here are some of the main tools. Check each one against your own security rules before you use client data.

QuickBooks AI agents: QuickBooks lists AI agents for accounting, sales tax, payments, payroll and month-end close. The accounting agent automates routine transaction work. Which agents you get depends on the plan.
Dext: Dext captures receipts and bills from email, a mobile app and other sources. Its site says it extracts the data, categorises it and syncs it to QuickBooks Online, Xero, Sage, Zoho and others. Firm plans are priced per client per month, with a 10-client minimum.
Blue J: Blue J is an AI tax research tool. Its site says answers come with sources you can check, drawn from primary authority, Tax Notes and IBFD. It offers a 7-day free trial.
General AI assistants: Many firms use a business AI assistant for drafts and summaries. Only use a business plan that keeps your data out of model training, and keep client data out of personal accounts.

For more on the bookkeeping side, see my guides to AI bookkeeping and AI document processing.

Which tool should a firm try first?

The three tools named above solve different problems, and firms often buy the wrong one first. If your biggest time cost is typing in receipts and bills, start with a capture tool like Dext, not a research tool. If it is sorting and categorizing transactions, check what your accounting software's own AI agents already do before paying for anything extra, since QuickBooks and similar platforms increasingly bundle this in. If it is research time on unusual tax questions, a tool like Blue J only pays for itself once your staff hit enough edge cases to need it regularly, a small firm with routine returns may not. Match the tool to the actual bottleneck, measured in real hours, not to whichever one has the best demo.

What rules apply to client data when accountants use AI?

Using AI does not change your duties to clients. It adds one more place where client data can go. These are the main US rules to know.

The FTC Safeguards Rule

The FTC Safeguards Rule covers financial institutions under the FTC's authority, and the FTC lists tax preparation firms as an example. A covered firm must have an information security program. Key parts include:

A Qualified Individual: One person must run the program. They do not need a special title or degree.
A written risk assessment: You must write down your risks and review them from time to time.
Encryption: Customer information must be encrypted on your systems and when it is sent.
Multi-factor authentication: Anyone who accesses customer information must use at least two ways to prove who they are.
Oversight of service providers: You must check that vendors protect your data, and set that out in contracts.

An AI vendor that handles client data is a service provider under this rule. Since May 2024, firms must also tell the FTC about a security breach that affects the unencrypted data of at least 500 people. That notice is due as soon as possible and no later than 30 days after you find the breach. Firms with data on fewer than 5,000 consumers are exempt from some parts of the rule.

IRS Publication 4557

IRS Publication 4557, Safeguarding Taxpayer Data, is the IRS guide for tax professionals. It says tax preparers are covered by the Safeguards Rule. It says they must keep a written information security plan, and it points to IRS Publication 5708 for help writing one. It also asks firms to use encryption, multi-factor authentication, strong passwords, backups and secure disposal of old devices. If data is stolen, it tells firms to report it to the IRS through their local Stakeholder Liaison and to the states where they file returns.

Add your AI tools to that written plan. List each tool, what data it sees and who can use it.

Section 7216 and tax return data

Section 7216 of the tax code limits how preparers use and disclose tax return information. A Journal of Accountancy article on AI advises firms to check the Section 7216 rules before tax return data goes to an AI tool. Some uses need the client's consent in the specific format the rules require. Ask your tax counsel which of your AI uses need consent.

The AICPA Code and tax standards

The same article notes that the AICPA Code's confidential client information rule applies to AI use. It also says tax work must follow the data protection section of the Statements on Standards for Tax Services. No federal law or CPA standard requires firms to tell clients they use AI. Some states, such as California and Utah, require some disclosure. The article suggests firms lean toward telling clients and getting consent.

Is there AICPA guidance on AI?

AICPA guidance on AI does exist. In April 2026, AICPA & CIMA published a Small Firm Generative AI Policy Template. It helps small CPA firms write a policy for using public AI tools such as Copilot, with a focus on managing risk, privacy and compliance. The template is for AICPA and CIMA members.

Even if you are not a member, write a policy. Keep it to a page or two. Cover which tools staff may use, what data may never go into them, who reviews AI work and how you tell clients.

What are the risks of AI in an accounting firm?

Wrong answers: AI can state a tax rule that is wrong or out of date. Always check the source before you advise a client.
Data leaks: Client data pasted into a personal AI account may be stored outside your control.
Weak vendor terms: Some tools keep or reuse the data you send. Read the terms and data settings before you sign up.
Over-trust: Staff may stop checking work that AI usually gets right. Keep a review step in every workflow.

How should an accounting firm start with AI?

List your repeat work: Find the tasks that take the most staff hours, such as receipt entry or document chasing.
Update your security plan: Add a section on AI tools to your written information security plan.
Pick one tool and one task: Start with a tool you already pay for, such as the AI in your accounting software.
Test on low-risk data: Use sample files first. Then move to real client work with a review step.
Measure and decide: After a month, compare the hours spent before and after. Keep what works.

When does a custom AI tool make sense for a firm?

Most small firms should use the AI built into the products they already have. A custom tool is worth it in these cases:

Many systems: Client data sits across a practice tool, a portal, email and tax software, and staff copy it between them.
Your own workflow: You have a review process or client service model no product supports.
Data control: You want client files processed in systems you control, under your own security plan.
Per-seat costs: You pay for several AI tools for every staff member, and the total keeps rising.

How would Blur Studio approach AI for an accounting firm?

I would start with your written security plan and the tasks that use the most staff time. I then build AI workflows that connect to the software you already use. That could be an intake assistant that checks client uploads, or a reminder system that chases missing documents. Every workflow keeps a person in the review step. Your firm owns the code and every login, and I keep the system running after launch.

Questions people ask

Can accountants use ChatGPT with client data?

Accountants should only put client data into a business AI plan that keeps data out of model training. The tool should also be listed in the firm's written security plan. Tax return information may need client consent under Section 7216 before it goes to an outside tool.

Does the FTC Safeguards Rule apply to tax preparers?

Yes, the FTC lists tax preparation firms as covered. IRS Publication 4557 also says professional tax preparers fall under the rule and must keep a written information security plan.

Does the AICPA have guidance on generative AI?

AICPA & CIMA published a Small Firm Generative AI Policy Template in April 2026 to help small CPA firms write an AI use policy. It is available to members.

Will AI replace accountants?

AI takes over data entry, sorting and first drafts. Clients still need an accountant to check the work, apply judgement and give advice they can rely on.

Do accountants have to tell clients they use AI?

No federal law or CPA standard requires it, according to the Journal of Accountancy. Some states, such as California and Utah, require some disclosure, so check your state and consider telling clients anyway.

What is the single biggest AI mistake accounting firms make?

Treating every AI tool as equally safe with client data. A capture tool that only reads receipts is a different risk than a general chatbot a staff member might paste a client's full financial statement into. Write the policy before the second kind happens by accident, not after.

Roughly what does this cost?

As a broad reference, fixes and finishes start at around $1,500, and new builds start higher, always a fixed price agreed in writing before I start, never an hourly rate. My AI app cost guide explains what actually drives a quote up or down.

Is my firm too small for this?

No. Most of the owners I work with have no tech team at all, that is the whole point. If the smarter move is a cheap software subscription instead of a custom build, I will tell you that on the first call, even though it means I do not get paid.