AI for Tax Preparers: Research, Drafting, and Review

AI for tax preparers fits best in the research and drafting stages: reading code sections, pulling together checklist items, and shaping client communications before a professional reviews and signs them. It does not replace the judgment a preparer applies to a specific client's situation, and it cannot file a return or take legal responsibility for the result. Tools like fenbs let preparers keep approved procedures, source references, and review rules where every AI they connect reads them before starting work.

The practical benefit is time. Searching the Internal Revenue Code, cross-referencing a revenue ruling, and drafting a plain-language explanation of a deduction all take concentration. An AI assistant can compress the search phase and produce a first draft that a preparer then checks, corrects, and approves. That pattern keeps professional judgment where it belongs while reducing repetitive work.

What AI can handle in tax preparation work

AI works well on tasks where the output needs a professional to verify it before it goes anywhere. That check is not optional: a misread code section or an out-of-date ruling in client advice is a professional liability problem, not just an inconvenience. Used with that constraint in mind, AI saves real time.

• Searching the Internal Revenue Code, IRS publications, and revenue rulings for relevant sections and summarizing what they say.
• Drafting first versions of client letters, engagement letters, and explanatory memos that a preparer then edits and approves.
• Building checklists for specific return types, such as schedules, required attachments, and common omissions to review.
• Generating plain-language explanations of deductions, credits, or elections for clients who want to understand their return.
• Comparing how a provision applies to a hypothetical situation and flagging where the answer depends on facts you have not yet confirmed.

What AI should not do without a qualified preparer's review: interpret the facts of a specific client's situation as final, advise a client on whether to take a position on a return, or produce any document that goes to the IRS or a client without a professional's check and signature.

Research workflows: reading code and rulings

Tax code research is well-suited to AI assistance because the primary sources are text. The IRS website↗ (opens in a new tab) publishes the code, regulations, revenue rulings, and publications. An AI that can read those sources can save a preparer the time of reading through every section. The discipline is to treat its summary as a starting point, not a conclusion. Verify the cited section number before relying on it.

A good research workflow looks like this: give the AI the specific question, ask it to identify the relevant code sections and cite them, then open those sections yourself and confirm the reading. Where the answer depends on recent guidance, check the IRS's own publications rather than relying on an AI summary that may not reflect the latest update. The AICPA's tax resources↗ (opens in a new tab) are also worth checking for authoritative guidance on practice questions.

Keep a record of which sources you verified and when. If you use a research result in client advice, you need to be able to demonstrate that you checked a primary source, not just that an AI produced a plausible summary. That record also protects you if a client later questions the basis for a recommendation.

Drafting checklists and client communications

Checklists are one of the most useful AI outputs for a tax practice. A well-structured checklist for a schedule or return type captures the items a preparer needs to review, the documents to request, and the common errors to check. An AI can draft one from a description of the return type. The preparer then reviews it for completeness, adjusts for their practice's specific standards, and approves it as an official procedure.

Client communications benefit from a similar approach. Draft the letter, then read it carefully: check that the deduction amounts and code citations are accurate, that the advice matches what the return actually shows, and that nothing in the letter commits the practice to a position the preparer has not reviewed. Plain-language explanations are especially worth checking because an AI will sometimes simplify a nuance out of existence.

• Start with a specific return type or client situation, not a general request.
• Ask for a draft and a list of the key assumptions it made.
• Review every number, code citation, and factual claim against the source.
• Edit the draft before it goes anywhere near a client.

Where professional judgment is non-negotiable

A preparer's judgment matters most when the client's facts are unusual, when a position on a return requires supporting analysis, or when the law is genuinely unclear. AI does not know the client. It cannot weigh the risks of a tax position, decide whether a transaction has economic substance, or tell a client whether to amend a prior year return. Those decisions belong to the licensed professional who has reviewed the complete picture.

Ethical obligations under Circular 230 and the AICPA's Statements on Standards for Tax Services also belong to the preparer, not to an AI. An AI does not have a Preparer Tax Identification Number and is not subject to professional discipline. The preparer who relies on AI output without reviewing it is still responsible for everything that output says.

Keeping approved procedures current

Tax law changes every year. A checklist or procedure that was accurate in one filing season may miss a new provision in the next. The practical answer is to tie procedure reviews to the legislative calendar: after a major tax act, review the affected checklists and update the approved versions. Record the review date and who approved the update.

When a client situation reveals a gap in an existing procedure, record what was missing and how the checklist should change. That lesson is worth saving: next season's return prep will benefit from it, and so will a new team member who uses the same procedure. The goal is a set of procedures that reflect current law and current practice knowledge, not just a set that was right when they were first written.

Source references and verified links

Tax work requires citing sources. An AI research output that says "see IRC Section 162" is useful only if the preparer has confirmed that Section 162 actually says what the AI claims. Keeping a record of verified sources in one place prevents re-checking the same section multiple times and makes it easier to demonstrate due diligence.

A links record for a tax practice might include the specific IRS publications that apply to each service area, the revenue rulings relevant to frequently encountered client situations, and the authoritative guidance from professional bodies like the AICPA and the American Bar Association's tax section↗ (opens in a new tab). Those references save time when the same question comes up across multiple clients. See also our guide to using AI at work for general workflows that translate well to a professional setting.

How to do this with fenbs

1. Create a free account. Start a project for your tax practice with a description of the service areas you want to improve first.
2. Add your approved procedures as Skills and prompts. Write the research prompt you use for a common return type, and record the checklist template you have verified as current for this filing season.
3. Add your key source references as Links. Include the IRS publication number, the URL, and a note on when you last confirmed it was current.
4. Record professional review rules in Decisions and rules: who must review any client-facing output, which code sections require partner-level sign-off, and which steps are non-negotiable before a document leaves the practice.
5. Connect your AI assistant from Integrations and ask it to read your project context before starting any research session. Before ending the session, ask it to save what it found and any sources it cited so you have a record for the next session.

fenbs does not connect to tax software platforms like UltraTax or Drake. It holds the procedures and references your AI reads before working, not the return data itself. Keep client information in your secure tax software and record only non-sensitive procedural information in fenbs.