A California CPA or tax-prep firm that wants a tax team fluent in AI before the next busy season needs three things in place before January: a workshop built around the firm’s own document intake and client correspondence, a clear answer on what rules govern putting client Social Security numbers and financial records into an AI tool, and a way to pay for it that doesn’t come out of partner draws. This guide covers all three, plus what a workshop costs and what CalCPA already offers on its own.
What training looks like before busy season
The window that matters is September through December, before the first W-2s and 1099s land in the client portal. A workshop run in this window is hands-on and built around the firm’s real intake process, not a generic “intro to ChatGPT” session: staff bring an actual client organizer, a redacted engagement letter, and last year’s correspondence, and the session works through how each moves faster with AI without changing who is liable for the return.
Sessions typically run half a day to two days depending on staff count and how many workflows the firm wants covered, with a small group working in the tools they’ll use in January. The goal is a document intake checklist, a drafted client-letter template, and a written rule for what never goes into a general-purpose AI tool, not just a list of product names.
Which firms this is built for
This fits sole practitioners through firms of roughly 5 to 50 people: solo preparers, small partnerships, and regional firms large enough for a dedicated tax season crew but too small for an in-house IT function. Practice type matters more than headcount. A firm doing individual 1040 volume work gets the most from the organizer-processing and correspondence pieces below; a firm doing more advisory work gets more from tax-law summarization. Most firms this size do some of both, so the workshop covers all three.
Firms above 50 people, or with an existing IT function, are usually better served by a custom rollout scoped to their document management system; this guide is written for the firm that doesn’t have that function yet.
What the workshop covers
Document intake and organizer processing
The highest-volume busy season task is turning a client’s uploaded documents (W-2s, 1099s, K-1s, mortgage interest statements, brokerage 1099-Bs, charitable contribution receipts) into an organized set checked against last year’s return. AI tools are strong at classifying document types, extracting the numbers on them, and flagging what’s missing; a workshop session walks staff through doing this inside the firm’s own portal and file-naming convention, not a demo folder, so the habit sticks past the training day.
Client correspondence drafting
Every firm sends the same handful of letters every season: missing-document follow-ups, engagement letter cover notes, extension explanations, and “here’s what changed” notices. AI tools draft strong first passes of all four once they’ve seen the firm’s own voice and a few redacted past letters, so more of a preparer’s day goes to review work a machine can’t do.
Summarizing tax-law changes for clients
Every filing season carries federal and California changes a firm needs to explain to clients in plain language, not statute text. AI tools turn a dense revenue procedure or a California Franchise Tax Board notice into a client-ready paragraph well, provided a preparer checks every number and date against the source before it goes out. A Thomson Reuters Institute survey of tax professionals, published March 2, 2026, found document summarization cited by 57% of respondents as a current AI use case, and tax research cited by 69%. Training covers the tools by product name (ChatGPT, Claude, Gemini, Microsoft Copilot), not version number, since firms pick based on what their existing software already integrates with.
How it’s delivered
Delivery follows the same rule as every engagement in this program: on-site is standard for firms in the Bay Area, and remote by default elsewhere in California, with in-person delivery available for an added fee and included on larger engagements. A Los Angeles firm can see exact terms on the Los Angeles training guide; firms elsewhere in the state can check the California training playbook for the same rule stated once, in full.
How to pay for it: California’s ETP Small Business Program
California’s Employment Training Panel reimburses part of the cost of training like this through its Small Business Program, for California-domiciled firms with 100 or fewer California employees and no more than 250 worldwide. The reimbursement rate for classroom, lab, or instructor-led training is $28 per trainee hour, per ETP’s own Fixed Fee Reimbursement Rate table, current as of this writing. A ten-person tax team running a two-day workshop can offset a real share of the direct cost this way; minimum training hours and a retention period apply, so confirm current terms on etp.ca.gov before applying. The step-by-step ETP funding guide covers eligibility, application paths, and reimbursement paperwork in full.
The confidentiality and AI rules a tax practice needs to know
A tax practice’s real exposure is federal before it’s ever state, because a preparer’s confidentiality duty is a criminal statute, not a general privacy guideline. IRC Section 7216 makes it a crime for a tax return preparer to disclose or use a client’s tax return information outside narrow exceptions. Per the IRS’s own Section 7216 Information Center, pasting a client’s tax data into a general-purpose AI tool is a disclosure under that statute unless an exception applies or the client has given knowing, written consent in the format Treasury Regulation 301.7216-3 and Revenue Procedure 2013-14 require. That consent has to name who receives the data; a blanket engagement-letter clause doesn’t satisfy it.
The IRS Office of Professional Responsibility addressed AI directly for the first time on June 24, 2026, in Alert 2026-19, “Introductory Guidelines for Responsible AI Use in Federal Tax Practice.” The alert doesn’t create new rules; it maps existing Circular 230 duties (due diligence, competence, firm procedures, and the standard for written advice) onto AI-assisted work, and it’s explicit that a preparer, not the AI vendor, owns every number and citation an AI tool produces.
On the state side, most tax practices this size never trigger the California Consumer Privacy Act: it applies only to a business clearing one of three thresholds (gross revenue over $26,625,000, sharing the data of 100,000 or more California residents a year, or getting half its revenue from selling personal data), and a firm with a few hundred or a few thousand clients is nowhere near the volume prong. The full CCPA applicability guide walks through the test in detail; for most CPA firms, IRC 7216 is the rule that governs day-to-day AI use, and CCPA coverage is worth checking once, not managing every season.
On professional guidance: AICPA published a Small Firm Generative AI Policy Template on April 29, 2026, and CalCPA ran “The CPA’s AI Checklist: What to Ask Before You Deploy” in its May 3, 2026 issue of California CPA magazine, both stressing vendor due diligence (confirming a tool won’t train on client data without authorization) alongside the 7216 consent question. The California Board of Accountancy has not issued CPA-specific AI-use guidance as of this writing; its board heard an AI trends and risk-management presentation at its November 2025 meeting, and its one enacted AI-specific law, AB 1405, signed September 9, 2026, governs enrollment for autonomous AI systems performing audits starting January 1, 2027 — a narrow rule for AI conducting the audit itself, not a firm’s internal use of ChatGPT or Copilot.
What it costs
Market pricing for a hands-on workshop this size runs roughly $2,000 to $15,000 depending on group size, session length, and workflows covered, before any ETP reimbursement. A follow-on project, such as a standing document-intake pipeline built into the firm’s existing portal, runs roughly $25,000 to $150,000 depending on scope. These are market ranges, not one firm’s list prices.
Alternatives: CalCPA’s own CPE, and what a workshop adds
CalCPA runs its own AI-focused continuing education, and a firm should know what’s already available before paying for more. Its Education Foundation catalog includes self-paced and webcast sessions such as “August 2026 AI Update: Current Issues, Future Possibilities” (a quarterly briefing on LLMs, chatbots, and AI rules, held August 10, 2026) and shorter single-credit sessions on AI agents for advisory work and AI in the financial close. These are strong for staying current and earning CPE credit, at far less cost than a private workshop.
What they don’t do is work through one firm’s own letters, intake process, or confidentiality gaps with a facilitator in the room. A CPE session teaches the topic to strangers; a private workshop builds the firm’s own checklist, letter templates, and AI-use rule from its real files. A firm that has taken CalCPA’s AI Update and still has no written intake process or consent template is the clearest fit for the hands-on version.
Frequently asked questions
Can a CPA firm put client tax data into ChatGPT or Claude at all?
Only with either an applicable exception or the client’s knowing, written consent in the format IRC Section 7216 and Treasury Regulation 301.7216-3 require, per the IRS’s Section 7216 Information Center. A blanket engagement-letter clause generally doesn’t meet the format requirement; the consent has to identify who receives the information.
Does the California Consumer Privacy Act apply to my small tax practice?
Usually not. It applies only if a business clears one of three thresholds — gross revenue over $26,625,000, sharing the data of 100,000 or more California residents a year, or getting half its revenue from selling personal data — and most firms in the 5-to-50-person range clear none of them. The full applicability test walks through all three.
Has the California Board of Accountancy issued rules on CPAs using AI?
Not specific-to-CPAs guidance as of this writing. The Board discussed AI trends and risk management at its November 2025 meeting, and its one enacted AI-specific law, AB 1405, signed September 9, 2026, covers enrollment for autonomous AI audit systems starting January 1, 2027, not a firm’s internal use of AI tools.
How much of an AI workshop can ETP reimburse?
$28 per trainee hour for classroom or instructor-led training under the Small Business Program, for California-domiciled firms with 100 or fewer California employees and no more than 250 worldwide, per ETP’s reimbursement rate table. Minimum hours and a retention period apply; the step-by-step guide covers the application process.
Is a live workshop worth it if my firm already takes CalCPA’s AI CPE courses?
CPE courses are strong for staying current and earning credit; they don’t build a firm-specific document-intake checklist, letter templates, or written consent process. A firm that has taken CalCPA’s AI sessions and still lacks a written AI-use rule for client data is the clearest fit for a private workshop.
Key takeaways
- A busy-season-ready AI workshop for a California tax practice covers three things: document intake and organizer processing, client correspondence drafting, and summarizing tax-law changes for clients — by product name, not model version.
- IRC Section 7216 is a criminal disclosure statute, and it’s the rule most California tax practices need to manage day-to-day, ahead of the CCPA, which few firms this size ever trigger.
- The IRS Office of Professional Responsibility addressed AI in federal tax practice for the first time on June 24, 2026 (Alert 2026-19); it applies existing Circular 230 duties to AI-assisted work rather than creating new ones.
- California’s ETP Small Business Program can reimburse $28 per trainee hour of the workshop cost for eligible firms with 100 or fewer California employees.
- CalCPA’s own AI-focused CPE is a real, lower-cost option for staying current; a private workshop is the option for a firm that needs its own intake process, letter templates, and consent rule built, not just explained.
Arthur Wandzel
