will AI replace tax preparers?

Eight of the 12 tasks O*NET identifies for tax preparers show zero measurable AI penetration right now. That's not a rounding error. It means the core of what you do every day, preparing returns, reviewing financial records, interviewing clients, applying deductions strategically, is still human work.

The client interview is the clearest example. When you sit down with someone and ask about their side income, their home office, their elderly parent they've been supporting, you're pulling information that isn't in any document. People don't know what's deductible. They don't know what questions to answer unless you ask the right ones. An AI tool can't do that interview, and a client won't open up to a chatbot the same way they do to someone they've trusted with their finances for a decade.

Tax planning is another area that's genuinely yours. Advising a small business owner on whether to take a Section 179 deduction this year or spread it out requires understanding their cash flow, their plans for next year, and their risk tolerance. The IRS tax code has over 70,000 pages. Applying it well to one specific person's situation, especially an atypical one, takes judgment that goes well beyond pattern-matching. That's what keeps you in the room.

The biggest failure point for AI in tax work is hallucination on edge cases. General-purpose models like ChatGPT can confidently state a wrong deduction limit or cite a tax provision that was amended two years ago. In a field where a $500 error can trigger an audit or a penalty, that's not a minor inconvenience. It's a liability.

Privacy is the other real problem. Tax returns contain Social Security numbers, bank account details, income figures, and dependent information. Feeding that data into a cloud-based AI tool raises serious questions under IRS data security requirements and state-level privacy laws. The IRS Publication 4557 sets specific safeguards for preparer data security, and many AI tools haven't been built with those requirements in mind. The risk sits with you, not the software vendor.

AI also can't represent a client before the IRS. Only enrolled agents, CPAs, and attorneys can do that. If a return gets questioned and a client needs help in an examination, the AI that helped prepare the return is irrelevant. You're the one who has to stand behind the work, which means you're the one who has to understand it fully.

The one task where AI has genuinely broken through is error-checking. Tools built into professional tax software, like Intuit ProConnect and Drake Tax, run automated validation checks that catch arithmetic errors, missing entries, and mismatched figures before you file. That used to require a second human review of every form. Now it happens in seconds.

On the scheduling and client communication side, tools like Calendly and basic AI-assisted email drafting in Gmail or Outlook handle appointment booking and follow-up reminders without you touching them. It's not dramatic, but it removes a real friction point during the February-April crunch when your inbox is unmanageable.

For general tax law lookups and explaining provisions to clients, tools like Bloomberg Tax and Thomson Reuters Checkpoint have added AI-assisted research features that pull relevant code sections and recent rulings faster than manual search. These don't replace your judgment about what applies, but they cut the time you spend hunting for the right IRS bulletin from 20 minutes to 2. TurboTax and H&R Block's consumer-facing AI tools are worth understanding too, not because they replace you, but because your clients are using them for simple returns, which means the clients who come to you are increasingly the ones with complex situations.

The part of your day that's genuinely shorter is the mechanical error-checking pass. Validation that used to mean flipping between screens and cross-referencing totals by hand now runs automatically before you review. You're spending less time catching data entry mistakes and more time on the return itself.

What hasn't changed is the interview, the review of source documents, and the actual filing decision-making. You're still going through bank statements, W-2s, 1099s, and receipts. You're still asking the follow-up question that surfaces a deduction the client didn't know to mention. That part of the day is the same as it was five years ago.

The shift in balance is subtle but real. Admin overhead, particularly scheduling and routine client follow-up, takes less of your attention during peak season. That time goes back to client work. The ratio of time spent on judgment calls versus clerical tasks has moved in your favour, which is the right direction if you want to be doing more of the work that's actually hard to replace.

The BLS projects 4.5% growth for tax preparers between 2024 and 2034, which is roughly in line with the average for all occupations. With 90,600 people currently employed and 10,400 annual job openings, the field isn't shrinking. But the growth story has some nuance worth understanding.

The clients doing simple returns, a single W-2, standard deduction, no dependents, are increasingly handling their own taxes with consumer tools. That's been happening for 15 years and it's not new. What's new is that those tools are getting faster and more accurate, so that segment of the market probably won't grow. The growth in professional tax preparation work is concentrated in the complex end: small business owners, self-employed individuals, people with investment income, estate planning needs, and multi-state filings. Those clients need you specifically because the AI tools make mistakes on complex returns.

The Anthropic Economic Index rates tax preparers at a 17% AI exposure score, which is low for a knowledge-based profession. For context, that score reflects how much of your actual task list AI can handle today, and the answer is: not much. The exposure is real but narrow. It's concentrated in one task. That's a stable position, not a precarious one.

The 17% exposure score is likely to edge up over the next five years, not because AI will take over the complex work, but because more of the data-gathering and document-sorting steps will get automated. Optical character recognition tools that pull figures directly from uploaded W-2s and 1099s into return software are already available and improving. Expect that to become standard across all major professional tax platforms by 2027 or 2028.

For this role to face genuine pressure, AI would need to reliably handle ambiguous situations: the client whose business income is hard to categorise, the return that involves a recently changed state tax law, the situation where two valid interpretations of a deduction exist and you have to choose the defensible one. That kind of judgment requires accountability that AI tools can't carry. It also requires a professional who can be audited, questioned, and held responsible. Those barriers don't disappear in five years. They might not disappear in ten.

Double down on the eight irreplaceable tasks. Specifically, get better at tax planning conversations, not just return preparation. There's a real difference between a preparer who files accurately and one who proactively calls a client in October to suggest a strategy before the tax year closes. The second person is much harder to replace and can charge more for it.

The small business side of this work is worth prioritising. Small business returns are more complex, more relationship-dependent, and more likely to involve ongoing advisory work across the year. According to BLS data, self-employment is growing in nearly every sector, which means more potential clients who need quarterly estimated tax help, business expense tracking advice, and entity structure guidance. That's not software work.

Get comfortable with the research tools in your existing professional software. Bloomberg Tax, Checkpoint, and similar platforms are adding AI-assisted search features, and being fast with them makes you more efficient without exposing client data to outside systems. The preparers who struggle in the next decade won't be replaced by AI directly. They'll be out-competed by other preparers who use the documentation and research tools to handle more clients in less time, while doing better work on the complex cases. Stay on the complex cases. That's where the work is going.