Accounting

AI for accountants: the profession your clients expect to use it

77% of tax clients want their firm using AI. 59% have no idea whether it does. Meanwhile adoption jumped from 9% to 41% in a single year, and the IRS just published its first AI rules for tax practice. The evidence, the fee math, and the new compliance lines, for firms in Vancouver, Canada, and the US.

By Adi Huric, founder of Most AI Labs·July 2026·10 min read

Accounting is the rare industry where the AI question has flipped: it is no longer whether your firm should use it, but whether your clients already assume you do. Thomson Reuters found 77% of tax clients want their firm using generative AI, and 59% have no idea whether it does. That second number is a free differentiation opportunity sitting on the table.

Written for firm owners, solo CPAs, and bookkeeping practices in Vancouver, anywhere in Canada, or the US. Every claim is cited. Vendor numbers are labeled as vendor numbers, and forecasts are labeled as forecasts.

Adoption: the fastest curve we have measured in this series

Wolters Kluwer’s survey of 2,700+ professionals found firm AI adoption jumping from 9% in 2024 to 41% in 2025, with 72% of adopters using it at least weekly and 73% of regular users reporting better-than-expected results. The 2026 data says the curve has not bent: Thomson Reuters found organization-wide AI use nearly doubled again, from 22% to 40%, and a June 2026 survey of 1,000+ US tax professionals by Blue J and CPA.com found 60% now using AI for tax research at least weekly, up from 33% one year earlier.

One adoption detail matters more than the headline: 65% of tax professionals are still using public tools like ChatGPT rather than industry-specific ones. Hold that thought until the compliance section, because the IRS has now written rules about exactly that.

What firms actually use it for

·Tax research leads everything: 69% of AI-using tax professionals put it first. The weekly-use breakdown behind it: advisory projects 44%, tax planning 40%, compliance research 39%, document analysis 36%, drafting 35%.
·Time savings are real but survey-measured: 84% of tax pros say AI saves them time, and the top reinvestments are faster client response (50%) and staff work-life balance (47%). The widely quoted “5 hours per week, worth $19,000 per professional per year” figure from Thomson Reuters is a respondent forecast, not a stopwatch measurement. We use it only as that.
·Client communication: Karbon’s survey says AI-using firms save around 18 hours per employee per month, mostly on email drafting. Karbon sells the AI in question, so treat it as directional.
·Bookkeeping categorization accuracy has no independent study. Every “99% accurate” figure circulating is a vendor claim with no outside audit. The workflow is still worth automating; just size the human-review step to reality rather than to the brochure.
·The advisory shift is measurable: 93% of firms now offer advisory services, up from 83% a year earlier, and 69% of tax professionals expect to move toward fixed-fee or value pricing as AI compresses task time.

Watch out for this

The pricing angle almost nobody mentions: the IRS Office of Professional Responsibility has warned that billing full manual-labor time for work AI shortened could violate Circular 230’s unconscionable-fee rule. AI does not just enable value pricing. In US federal tax practice, it quietly pressures you toward it.

June 2026: the IRS wrote the rules. Here is the checklist.

On June 24, 2026, the IRS Office of Professional Responsibility published its first guidelines for AI use in federal tax practice (Alert 2026-19). The obligations map onto existing Circular 230 duties: verify every AI output before it reaches a client, understand the tools you use, maintain a firm AI policy and vet third-party tools, and never let unverified AI citations into written advice. The same alert flags the sharpest edge: uploading taxpayer data to public AI tools risks unauthorized-disclosure violations under IRC §6713 and §7216, and §7216 consent must name the actual recipient. Remember that 65%-still-on-ChatGPT stat. That is the gap between how firms are using AI and how the regulator says they must.

The hallucination risk, honestly sized: the US Tax Court has seen fewer than a dozen hallucinated-citation incidents and no formal sanctions yet, against roughly 800 AI citation incidents logged across all courts worldwide, some drawing five-figure sanctions. Fabricated authority in tax work can trigger preparer penalties under §6694. CAMICO, the largest CPA-focused liability insurer, already publishes a generative AI risk FAQ and a usage-policy template for policyholders. The professional consensus in one sentence, courtesy of CPA Ontario: the signing CPA owns the result, no matter what tool produced it.

The staffing reality behind the urgency

US accounting degrees fell 6.6% to 55,152 in the 2023-24 academic year, and new CPA exam candidates dropped from 42,626 to 28,082 (a fall exaggerated by candidates rushing to beat the January 2024 exam changeover, so read those two years together). There are green shoots: accounting enrollment grew about 12% year over year for two straight semesters in 2024-25. But three in four firms are still hiring, and billable capacity keeps shrinking, with the share of firms averaging under 1,500 billable hours per person rising. The honest Canadian note: CPA Canada publishes no equivalent pipeline report, and CFE pass counts are roughly flat year over year, so we will not import the US crisis narrative wholesale. The capacity squeeze, though, is universal.

The profession’s profit math is the quiet why-now: firm revenue grew 7.9% while partner income grew 3.2%. Revenue is growing faster than profit. Capacity is the constraint, and recovered hours are the only input every firm can still buy.

The money math for a small firm

The Rosenberg Survey’s 2025 edition (FY2024 data) shows average income per equity partner at $615K, and the bracket most readers of this article occupy, $2M to $5M firms, grew income per partner about 25% to $464K while $10M to $20M firms declined 7.2%. Rosenberg attributes that to leverage and billing-rate discipline, not AI, and so do we. What AI changes is the input: a base Form 1040 averages $280 from a CPA (NATP 2025 fee study), so a solo preparing 300 returns grosses about $84K from 1040 work. Our arithmetic, with the assumption labeled: shave even 20 minutes per return and you free roughly 100 hours per season, worth $15,000 to $25,000 redeployed into advisory at typical rates. No verified time-per-return benchmark exists yet, which is why we show the assumption instead of hiding it.

The Canadian and BC picture

CPA Canada and the AICPA publish a joint AI guidance series, and CPA Canada is lobbying Ottawa for a formal AI assurance framework. The CRA itself runs 19 AI systems but states it does not use AI to decide individual personal tax returns. And there is a trust story hiding in plain sight: the CRA’s own AI chatbot warns users its answers “may not be fully accurate” and advises consulting a professional; the Auditor General found its roughly $18M chatbot answered just 2 of 6 trial questions correctly before its late-2025 upgrade. When the tax authority’s own bot tells Canadians to see a professional, “human-reviewed AI” is not a compliance burden for your firm. It is the sales pitch.

For BC firms, the privacy anchor is the joint guidance from Canada’s privacy commissioners, including BC’s OIPC, on generative AI: consent, necessity, openness, accountability. The regulators are enforcing it too; the OPC and three provincial commissioners, BC included, completed a joint investigation of OpenAI confirming Canadian privacy law applies to AI providers. CPABC has run a practitioner series on AI ethics and risk in its own magazine, so the local professional bar is set.

The honest bottom line

·Write the firm AI policy first. 65% of tax pros are on public ChatGPT while the IRS now expects vetted tools, verified output, and named-recipient consent for client data. Close that gap before adding anything new.
·Start where the evidence is: tax research and drafting. It is the most-adopted use case (60% weekly and climbing) with the clearest professional-rules fit: AI drafts, the CPA verifies, the CPA signs.
·Tell your clients. 77% want their firm using AI and 59% cannot tell if it does. A paragraph in your engagement letter and a line on your website is free differentiation.
·Rethink pricing before the hours compress. 69% of the profession sees fixed-fee coming, and Circular 230 pressure points the same direction. Value pricing is the AI dividend; hourly billing gives it away.
·Treat bookkeeping-AI accuracy claims as unaudited. Automate the workflow, keep the review, and measure error rates on your own books before trusting any brochure number.

If you want this mapped to your firm with your own return volumes, realization rates, and staffing plan, that is what our AI consulting work does. It starts with the free 7-day audit: where your hours, fees, and client communication actually leak, with the math shown before you spend anything. Fixed price. You own what gets built.

Wondering what an ad budget buys here? Our free lead calculator runs the verified benchmark math for this industry: professional-services leads run about $94 on Google, and why no accounting close-rate benchmark exists.

Sources

If any of this is your week

Start with the 7-day audit.

7 business days. A real document. Yours to keep — whether you hire us or not.

Start with the 7-day audit Or talk to the founder first

AI for law firms: what the evidence actually supports

The same evidence-first treatment for the other document-heavy profession.

Read→

Hospitality

AI for restaurants: where thin margins meet real automation

What survived contact with real kitchens, including the famous failures.

Read→