AI in Tax Practice (2026 Edition)

#The frame

Most “AI in accounting” content falls into two camps:

Camp 1: AI will replace accountants. Usually written by tech publications. This camp gets the specifics wrong, because rules and liability stop end-to-end AI tax prep. It gets the timeline wrong too (10 years, when 2-3 years fits better). And it gets the impact wrong: boutique firms gain more than they lose.

Camp 2: AI is dangerous and unreliable, so we’re not touching it. Usually written by traditional firms protecting the status quo. This camp gets the technology wrong (today’s models beat 2023 baselines by a lot). It gets competitive dynamics wrong too (firms that skip AI will lose clients to firms that use it). And it gets client expectations wrong, since more and more clients expect AI-enabled response times.

Here is the honest middle: AI is genuinely useful for specific tax-firm workflows, useless for others, and dangerous for a third group. The trick is knowing which is which.

This article covers what we’ve actually deployed at ETS. What works. What doesn’t. What we’re testing. What we won’t touch.

#What we use AI for (and how)

#Use case 1: Document parsing + extraction

When a client uploads a W-2, 1099, K-1, or tax return PDF, AI pulls out the structured data. That means employer name, wages, federal withholding, state withholding, and box codes, into a clean format. Manual data entry used to take 5 to 10 minutes per document. Now it takes about 30 seconds, with a human review step.

For an engagement with 20 to 50 source documents, this saves hours of time per client. Multiply that across 30 active clients per partner, and it is a real jump in capacity.

Tool stack: GPT-5, Claude, and specialized tax-data extraction services. A human reviews the output before it enters our books or returns. Before you put client data into any AI tool, check where the data is stored and who can see it. Get written client consent where the tax rules require it. Keep Social Security numbers out of any tool that processes data outside the U.S.

#Use case 2: Bank-feed categorization (in Kick)

Our bookkeeping platform Kick has built-in AI categorization. It learns each client’s own patterns over time. For a client with 500 transactions a month, AI gets about 85% of categorizations right out of the box. After a few months of learning, that rises to about 95%.

That last 5 to 15% is where the human bookkeeper adds value. That is also exactly where AI was always going to struggle: edge cases, unclear transactions, anything that needs business context AI cannot have.

#Use case 3: Tax research

For tax-law research (recent provisions, edge cases, IRS rulings, state-conformity questions), AI research tools speed up the work. Think of tools like Blue J and general AI models. A research question that used to take 1 to 2 hours of CCH or RIA Checkpoint navigation now takes 5 to 15 minutes.

Critical caveat: a human validates AI citations before they appear in client work. Hallucinated case law and made-up code sections are a real risk. The pattern works like this: AI finds the relevant authority faster, then a human checks that it actually applies.

#Use case 4: First-draft client communications

For routine client emails (engagement letters, status updates, document requests, meeting recaps), AI writes a clean first draft in seconds. A human editor then adjusts it for voice, accuracy, and context.

This saves about 70% of the time spent drafting emails. The voice still sounds like ours, because we edit for tone before sending. No client-facing communication goes out without human review.

#Use case 5: Operational documentation

This is the unsexy use case nobody talks about. Internal SOPs, runbooks, training docs, process documentation. AI is remarkably good at drafting operational content from a conversational description of how something works.

Across a small firm, this adds up. Good operational documentation that did not take partners 40 hours to write is real time saved.

#Use case 6: Code + automation

Our internal systems (Basecamp automations, Kick integrations, custom dashboards, this website) take real software development. AI-assisted coding (Claude Code, Cursor, GitHub Copilot) lets us build internal tools about 3 to 5 times faster than before.

The ETS website you are reading right now was built mostly through AI-assisted development, against tight specs. A human reviewed every step. AI sped up the execution.

#What we use AI for vs what we don’t

The clearest way to see the line we draw: the same capabilities that are safe on the left become dangerous on the right.

A note on each “don’t”:

Final tax positions. Both partners review every tax position on every client return. One checks the tax substance. The other checks the client-facing communication. AI can suggest research. AI does not decide positions.

Form 8867 due-diligence affirmation questions. Form 8867 due diligence comes from IRC 6695(g) and Treas. Reg. 1.6695-2. The preparer who signs the return answers those questions. Circular 230 section 10.34 also asks us to make fair inquiries when client facts look wrong. At our firm, the licensed EA who signs the return answers these questions. AI is not allowed near them. This is the strictest part of our tax-prep workflow.

Client-facing real-time communication. Humans, not AI, meet our 1-3 business day response SLA on Basecamp. Clients book us to talk to us. If we needed AI to respond to them, they would be buying a different product.

IRS representation and audit defense. Anything with direct IRS interaction (POA, audits, appeals, Tax Court) is human-only. This is not because AI lacks the skill. It is because the rules and the relationship both require a human presence.

Anything where AI hallucination is a tax-position risk. The most dangerous AI failure is confidently wrong output that gets cited as authority. Code section numbers that do not exist. Cases with the wrong holdings. Regulations that say the opposite of what AI claims. We verify every AI-sourced citation before it touches client work.

#The hybrid workflow

Here is what it actually looks like for a typical Tax Analysis engagement. Total partner-hours per Tax Analysis: about 16 hours per engagement, down from about 24-28 pre-AI.

That cut of about a third of our time is what lets us take on 3 new clients a month without adding staff. The cap stays the same. The throughput per partner goes up a lot. Clients still get the partners on every call. They just get them faster than before.

#What we’re testing (not yet deployed)

A few capabilities we’re piloting, but have not put in production:

1. Client portal voice assistant. A Basecamp portal Q&A bot that can answer client questions about their own documents and their engagement status. Currently testing. Not deployed, because the trust and liability questions are not settled yet.

2. Predictive estimated-tax modeling. An ML-based quarterly estimated tax calculator that learns from each client’s actual income patterns. Currently in alpha testing.

3. Automated bookkeeping reconciliation. End-of-month reconciliation that flags gaps, suggests adjustments, and produces audit-ready output. Promising, but not production-ready.

4. Automated state-tax conformity tracking. A system that tracks which state has matched which federal tax rule. OBBBA alone creates 50-state conformity confusion. Tracking this by hand is a hassle. Automating it would save weekly research hours.

#What’s clear by 2026

A few patterns we now consider settled:

1. AI-assisted tax research is a real productivity win. But humans still verify citations. Hallucinated authorities are the #1 failure mode.

2. AI-assisted bookkeeping categorization works, and gets better with client-specific training data. This was the 2023-2024 promise. Now it is reality.

3. AI does not replace partner judgment. It shifts what partners spend time on. Less time on data entry. More time on strategy and the client relationship.

4. Boutique firms gain more than mid-size firms. Larger firms have more associates whose work AI partially replicates. Smaller firms have less associate work to start, so they use AI to expand partner throughput instead.

5. Rules and liability lag the technology. Circular 230, state-bar rules, and malpractice concerns all assume a human judges every step. AI deployments have to work inside these limits, not around them.

#What we tell clients

Three things, when AI comes up in conversation:

1. We use it. Aggressively. Where it adds value. We don’t hide that.

2. It doesn’t replace the human work you’re paying for. Every position is human-reviewed. Every form is human-signed. Every conversation is human-led.

3. It lets us deliver more substance per dollar than firms that aren’t using it. Bigger Tax Analyses. Deeper research. Faster turnaround. Same partner-client model.

Comparing ETS against another firm? Ask both firms what they use AI for and how. The answers will tell you a lot about how each firm thinks about the next decade.

#What I’d tell other firm builders

If you’re running or starting a tax or accounting firm, and thinking about AI adoption, here is what we’d tell you:

Don’t try to replace human work end-to-end. The rules and the liability make this dangerous. Don’t be the test case.

Do automate the parts of the workflow that don’t need judgment. Document extraction. First-draft generation. Research speed. Bookkeeping categorization. These are where the wins are.

Verify every AI output before it touches client work. Especially citations and numerical claims. AI confidence is not the same as AI correctness.

Invest in operational documentation. AI is shockingly good at producing it from conversational input. Every SOP you write down becomes leverage for future hires and future automation.

Stay close to the regulatory developments. AICPA, state CPA societies, NAEA. They are all forming positions on AI in practice. Don’t get caught with a deployment that creates compliance issues.

The firms that figure out the human-AI hybrid well over the next 3-5 years will be the boutique firms of the next decade. The firms that either reject AI entirely, or try to fully automate around it, will both struggle.

---

This article is part of “From the Operator Seat”, Ramon’s lane on firm philosophy + operations. These articles are for prospects who want to see how the firm actually works. They are also for other tax-firm builders thinking about the same questions.

Comparing tax firms and wondering how AI affects what you’d get from ETS? The Discovery call is the right next step. Honest answer: we use it heavily where it adds value; every partner-led conversation is still partner-led.