How CPAs Use AI Day-to-Day in 2026: Workflows for Tax Season, Audit Defense, and Advisory Work

AI adoption among CPAs jumped from 9% to 41% in a single year. What CPAs actually do with it day-to-day has been less examined. This guide walks through the real workflows that public accountants, tax preparers, and advisory practices are using AI for in 2026 — what works, what doesn't, and how to do each one without triggering an IRC §7216 disclosure.

The Three Categories of AI-Assisted CPA Work

Most CPA AI usage falls into three buckets, each with different risk profiles.

Research and learning. Looking up an unfamiliar code section, understanding a recent court ruling, learning a new accounting standard. The prompts contain no client information. Low risk; AI is largely a faster Google for these tasks.

Drafting and explanation. Writing client communications, drafting memos, explaining tax positions in plain language for clients. These prompts often contain client-specific context. Medium-to-high risk depending on the AI tool.

Analysis and review. Analyzing client returns, summarizing client documentation, reviewing complex transactions. These prompts contain substantial client information. Highest risk under IRC §7216 and AICPA standards.

Workflow 1: Tax Research Acceleration

The use case: you encounter an unfamiliar situation — an obscure provision, a recent ruling, a client question about a topic you haven't worked with recently. Traditional research takes 30 minutes to several hours.

The AI workflow: query an AI tool with the abstract problem (no client identification), get a fast plain-English explanation and pointer to the relevant primary sources, then verify against your authoritative tax research subscription (Thomson Reuters Checkpoint, Bloomberg Tax, CCH AnswerConnect).

Time saved: typically 50–80%, particularly for moderately complex topics. Bloomberg Tax's 2025 AI Assistant data showed 84% of users finding relevant insights faster.¹

Compliance check: No client information in the prompt. This is the lowest-risk AI use case for CPAs. Any tier of AI tool works.

Common mistake: Trusting AI output without verifying against primary sources. AI models can confidently produce wrong citations and outdated information. Always verify.

Workflow 2: Client Communication Drafting

The use case: explaining a tax position, a return result, or a planning recommendation to a client in writing. CPAs spend significant time on client-facing communications that require clarity, accuracy, and a friendly professional tone.

The AI workflow: provide the AI with the technical content and ask it to rewrite for the client's audience. The technical content often includes client-specific facts and numbers.

Compliance check: Whether this is acceptable depends on the AI tool. With consumer ChatGPT or Claude.ai, providing client-specific facts is a disclosure under IRC §7216. With enterprise contracts that include zero-retention provisions, it's acceptable with documented vendor diligence. With on-device AI, no disclosure occurs and compliance is structural.

Practical pattern: Many CPAs draft generically first (“explain a Roth conversion in plain English for a high-income taxpayer”) then add client specifics in a follow-up step using compliance-appropriate tooling.

Workflow 3: Audit Defense and IRS Correspondence

The use case: drafting responses to IRS notices, preparing supporting documentation for an audit, organizing the legal and factual arguments for a position your client took.

The AI workflow: feed the IRS notice (after redacting taxpayer identification if appropriate), ask for an analysis of the issues raised and a structured outline for the response.

Compliance check: IRS notices typically identify the taxpayer prominently. The full notice contains return information. This is high-risk for IRC §7216 unless the AI tool keeps the data on your device or operates under appropriate enterprise contracts.

Practical pattern: On-device AI shines here because the full document context can stay local. The CPA can ask the AI to help structure the response argument while the underlying notice never leaves the office.

Workflow 4: Year-Round Advisory Work

The use case: client advisory work — entity choice decisions, succession planning, retirement plan design, multi-year tax projections. The advisory work that increasingly defines high-value CPA practice.

The AI workflow: think through scenarios in conversation with AI, get plain-English explanations to share with clients, generate variations of recommendations for different fact patterns.

Compliance check: Advisory conversations typically involve substantial client-specific context. Same tier rules apply: consumer AI for general scenario thinking with no client identifiers; on-device or enterprise AI for anything client-specific.

Workflow 5: Document Review and Summarization

The use case: reviewing partnership agreements, trust documents, operating agreements, foreign account statements — documents where the relevant tax information is buried in dense legal or financial text.

The AI workflow: paste relevant sections into AI, ask for a tax-focused summary highlighting elections made, basis adjustments, distribution provisions, allocation rules.

Compliance check: Client documents contain extensive client information. Highest risk tier under §7216. On-device AI is the cleanest approach because the source document never leaves your computer.

Workflow 6: Continuing Education and Practice Updates

The use case: staying current on tax law changes, AICPA standards updates, and accounting pronouncements. CPAs face an ever-growing volume of required reading and CE requirements.

The AI workflow: feed CE course materials, journal articles, or regulatory announcements into AI for plain-language summaries highlighting the practice-relevant takeaways.

Compliance check: Public materials, no client information. Lowest risk tier. Any AI tool works.

The Tier System Every CPA Practice Should Adopt

Map workflows to tools systematically. A workable three-tier approach:

Tier 1 — Public AI (ChatGPT free, Gemini, Claude.ai):Approved for research with zero client identifiers. Prohibited for anything involving a client's name, return data, financial details, or other identifiable information.

Tier 2 — Enterprise AI (ChatGPT Enterprise, Claude for Enterprise, Microsoft Copilot for M365, dedicated tax tools like Thomson Reuters CoCounsel or Bloomberg Tax AI Assistant): Approved for client work if vendor contracts include zero-retention provisions, the contract addresses IRC §7216 specifically, engagement letters disclose AI use, and you have audit-trail documentation.

Tier 3 — On-Device AI (Hey Eduardo, Ollama-based tools):Approved for any work without restriction. No third-party disclosure occurs. Cleanest defensible position under every applicable standard.

Engagement Letter Updates Every CPA Should Make

Add an AI disclosure paragraph to your engagement letters. Suggested language to adapt:

“In providing services under this engagement, [Firm] may use artificial intelligence (AI) tools to assist with research, drafting, analysis, or related tasks. [Firm] uses AI tools that either operate entirely on local hardware or are subject to written data protection agreements that prohibit model training and require deletion of inputs after processing. [Firm] will not disclose your tax return information to any AI service provider in a manner inconsistent with IRC §7216. [Firm] reviews all AI-generated content before providing it to you and remains fully responsible for the work product delivered.”

The Quarterly Practice Audit

Every quarter, ask three questions of your team:

1. Which AI tools did you use this quarter?
2. For which categories of work?
3. Did you ever input client-specific information into a Tier 1 tool?

Document the answers. Address any Tier 1 client-data usage with retraining. Update your tier policy if new tools emerged that should be approved or prohibited. The quarterly cadence catches drift before it becomes a §7216 exposure or an AICPA standards violation.

Part of our AI by Profession cluster: See the pillar guide for the cross-profession overview. For tax-specific AI tools, see our tax AI tools roundup. For broader compliance framework, see the step-by-step compliance guide. For the architectural argument behind on-device AI, see the On-Device AI pillar.