Tax practices run on documents, deadlines and trust. AI now touches all three. In a June 2026 survey of more than 1,000 tax professionals by Blue J and CPA.com, 60% said they use AI for tax research at least weekly, up from 33% a year earlier. Eighty-four percent agreed that it saves time.
The adoption is real. So is the question every firm owner eventually asks: what happens to my clients' information when it goes into one of these tools?
Where AI is earning its place
The same survey shows where firms are applying it: advisory projects (44%), tax planning (40%), compliance research (39%), document analysis (36%) and drafting (35%). In practice, that looks like four jobs.
Document intake. Sorting a client's W-2s, 1099s and K-1s, extracting the figures, and flagging what is missing before a preparer opens the file.
Research. Summarizing guidance and pointing to the source, so a preparer can check the citation instead of searching for it.
Planning. Modeling a few scenarios ahead of year-end so the conversation with the client starts from numbers, not a blank page.
Client communication. Drafting a clear reply to a client's question, or explaining an IRS notice in plain English, for a professional to review and send.
In each case, the AI produces a draft and a credentialed person reviews it. That is also where the IRS has landed. According to CNBC, the IRS issued its first AI guidance for practitioners in June. It requires that they review and verify AI-generated work, and says billing should reflect the efficiencies gained.
Where the risk sits
Under IRC §7216, a return preparer who knowingly or recklessly discloses tax return information, unless an exception applies, can face a fine of up to $1,000, up to a year in prison, or both. Return information is broad. It includes a client's name and address, and even the fact that someone is a client.
Practitioner commentary on the rule points out that when a preparer enters this information into a third-party AI tool, that is a disclosure to the tool's provider. CNBC reports that the IRS guidance does not say whether using generative AI to prepare returns must be disclosed to clients, and that the AICPA's own ethics staff describe the profession as still working through it. A 2024 Thomson Reuters Institute survey found about a quarter of tax and accounting professionals using public generative AI tools, and 9% using tax-specific ones.
Five questions to settle before adopting any AI tool
1. Where does client data go, and who can see it?
2. Is it stored, and is it used to train the provider's models?
3. Does the firm's engagement letter or consent language cover this use? This is a question for the firm's counsel.
4. Who reviews the output, and how is that review documented?
5. Can the firm produce a record of what the tool was given and what it returned?
What a private agent changes
A custom agent built around a firm's own workflow, and run on the firm's own server, does not make the legal questions disappear. It makes them answerable. The firm decides which files the agent can reach, who can use it, what is logged, and which model provider it calls. When a client or a regulator asks where the data went, the firm has a specific answer instead of a general assurance.
That is the work we do at Apollo Claw: custom AI agents built for how your practice actually operates, with the controls a professional firm needs. If you are weighing AI for your tax practice, book a conversation at apolloclaw.ai.
This post is general information and not legal or tax advice. Consult your own counsel on disclosure and consent requirements.