Justin McKelvey
Ran snowboard shops. Taught myself to code. Now I build AI that does real work.
AI in Auditing (2026): What a Small CPA Firm's Audit Team Can Hand AI, and What the Standards Say About the Evidence
How should a small CPA firm use AI in its audit practice?
A small CPA firm should use AI in auditing to draft and to read, never to conclude. Let it draft PBC request lists, planning memos, risk assessment narratives and workpaper documentation from the engagement team's own notes, and let it do first reads of leases, contracts and minutes. Then treat everything it produces as information the team still has to test. The judgments (risk, materiality, whether the evidence is enough) and the opinion stay with the engagement team. As of 2026, that's also roughly where the auditing standards draw the line.
This page is for firms with an audit practice of a few partners and a handful of seniors and staff, mostly private-company audits, maybe an employee benefit plan season. The Big Four AI platforms are real. They're also not what a 15-person firm is buying this year.
What the standards say about AI-produced evidence
There are two rulebooks, depending on who you audit, and they land in the same place.
| Private-company audits (AICPA) | Public-company audits (PCAOB) | |
|---|---|---|
| The standard | SAS 142, Audit Evidence (amends AU-C 500) | Amendments to AS 1105 (Audit Evidence) and AS 2301 (responses to assessed risks) on technology-assisted analysis |
| Effective | Audits of periods ending on or after December 15, 2022 | Audits of fiscal years beginning on or after December 15, 2025 |
| What it means for AI | Recognizes automated tools and techniques, and has the auditor evaluate information regardless of the source or the procedure used to get it | Spells out what the auditor does with items a technology-assisted test flags, and stresses controls over the information being analyzed |
| The practical rule | Tool output is information, not evidence on its own. The team evaluates its relevance and reliability, follows up on what it flags, and documents that it did. | |
That last row is the whole policy for a small firm. If a senior can't explain how an AI-drafted conclusion was checked, it isn't in the file yet.
The audit-file map: what AI drafts, what it assists, what stays with the auditor
| Audit file item | AI's role | What that means in practice |
|---|---|---|
| PBC request list | Drafts | From last year's list and this year's changes. The in-charge edits it before it goes to the client. |
| Engagement letter and planning memo | Drafts | From your firm templates and the partner's notes. The partner owns every judgment in it. |
| Risk assessment narrative | Drafts the writing | The team decides the risks; AI turns the team's bullets into documentation. |
| Leases, contracts, board minutes | Assists, auditor tests | A first read for terms that hit the financials (commitments, related parties, debt covenants). Every flagged item is traced to the source page by a person. |
| Analytics and sample selection | Assists, auditor tests | Use your audit software's tools; a general chat assistant isn't a sampling tool. Document how the population was tested for completeness. |
| Workpaper documentation | Drafts | Write-ups of work already performed. The preparer signs off on what was done, not on the prose. |
| Materiality, sufficiency of evidence, the opinion | Auditor only | Not drafted, not suggested. These are the calls the standards put on the engagement team. |
What AI in auditing is worth to a small firm: the math
Run this on your own numbers. The example uses placeholders, labeled as such, so the arithmetic is visible.
- Assumption: a firm does 30 private-company audits a year, and each one carries about 12 hours of writing (planning memo, narratives, documentation of work done).
- Draft-first: if AI drafts and the team edits, call it half: 6 hours back per audit, 180 hours a year.
- What that's worth: the median accountant or auditor earns $40.23 an hour (O*NET, 2025 wages). 180 hours is about $7,241 a year of staff time.
- Against the tool: five seats on an annual Claude Team plan are $20 a seat a month, $1,200 a year. One seat billed monthly ($25) pays for itself at about 37 minutes saved a month.
The honest catch is busy season. Those hours come back in February and March, when you need them most, but only if the templates and prompts were built in the fall. A firm that starts in January gets most of the benefit next year. If you're choosing between assistants, the Claude AI pricing breakdown covers what each plan includes for a team.
Client data: keep it narrower than you think you need to
Audit files hold some of the most sensitive data a business has: payroll, bank statements, customer lists, related-party details. Three working rules for a small firm:
- Firm business account only, with data terms that exclude training on your inputs. Never a personal account.
- Draft from your notes, not their ledger. Most of the time savings come from writing up work the team already did, which doesn't need the client's raw data at all.
- Check the engagement letter and your confidentiality obligations before any identifiable client data goes into an AI tool, and write the firm's answer down so staff don't each decide it alone.
Your tax practice has its own, stricter rules for return information; those are covered in ChatGPT for accountants, alongside the rest of the firm's non-audit work.
AI tools for auditors: start with what you already pay for
Before buying an audit-AI product, check what your audit software and practice-management system already include on your plan, because your data is already there under a contract you've signed. Then add a firm-wide business assistant for the drafting work. A specialist tool earns its place only when it does a named job those two can't, and its output still goes through the same evaluation as everything else. I haven't quoted audit-software prices here; check the vendor's own page the day you buy.
"AI in auditing" vs auditing AI
The two get mixed up in search. AI in auditing, this page, is auditors using AI to do audit work. Auditing AI is reviewing how a company uses AI: which tools, which data, what could go wrong. That second one is what I mean by an AI audit, and for some firms it's also a new advisory service to offer clients.
Who can help a CPA firm roll out AI in its audit practice?
I work with firms of about 10 to 200 people on the adoption side: picking the assistant, writing the firm's data rules, and building the templates and prompts the audit team actually uses, before busy season instead of during it. I don't sell audit software and I don't do attest work. A free 30-minute strategy call is the place to start. If you want the full picture first, the AI Readiness Assessment is $2,500 flat for a written roadmap in two weeks, credited in full if we do an install within 90 days, and installs start at $4,500 and take about two weeks. The whole ladder, written for firms like yours, is on my AI consulting for accounting firms page, and the firm-wide view (tax, bookkeeping, advisory) is in AI for accounting firms.
What I'd do this week
Pick one finished audit from last season. Count the hours that went into writing (not testing), and have one senior redraft that file's planning memo with an AI assistant from the original notes, timing it. That single number tells you whether this is worth setting up before busy season, and it costs you an afternoon.
Sources
- PCAOB amendments to AS 1105 and AS 2301 (technology-assisted analysis), PCAOB Release No. 2024-007 and SEC Release No. 34-100774: effective for audits of fiscal years beginning on or after December 15, 2025 (PCAOB implementation page, read October 15, 2026).
- AICPA SAS No. 142, Audit Evidence (amends AU-C 500): effective for audits of periods ending on or after December 15, 2022; recognizes automated tools and techniques (AICPA statements as reported by CPA Practice Advisor, read October 15, 2026).
- Wages: O*NET OnLine, Accountants and Auditors (13-2011.00), 2025 median $40.23/hr, read October 15, 2026.
- Claude Team pricing: claude.com/pricing, read October 15, 2026 ($20 per Standard seat per month billed annually, $25 monthly).
- Search demand: DataForSEO, United States, English, read October 15, 2026 ("ai in auditing" 1,300 Google searches a month; "ai for auditors" 412 AI-assistant searches in September).
- Assessment and install prices: justinmckelvey.com/ai-consulting/accounting-firms, as published October 15, 2026. This page is not professional standards guidance; read the standards themselves for your engagements.
Get the Free AI Content Toolkit
The exact system I use to turn one idea into a month of content: atomization framework, voice template, prompt library, weekly system.
Frequently Asked Questions
- How is AI used in auditing?
- At a small firm, mostly for drafting and reading: PBC request lists, planning memos, risk assessment narratives, workpaper documentation, and first reads of leases, contracts and board minutes for items that affect the financial statements. Larger firms also run analytics over full transaction populations. In every case, the output is information the engagement team still evaluates; the judgments and the opinion stay human.
- Will AI replace auditors?
- No. The standards put the judgments (risk, materiality, the sufficiency of evidence, the opinion) on the auditor, and AI doesn't sign anything. What changes is the drafting load: the hours a senior spends writing up work already done. That's where a small firm gets time back.
- Can auditors use ChatGPT or Claude on client files?
- Not in a personal account. A firm business plan with data terms that exclude training on your data is the minimum, and many firms keep identifiable client financial data out of general chat tools entirely, using them for drafting from the auditor's own notes instead. Check your engagement letters and your professional confidentiality obligations before any client data goes in.
- What do the auditing standards say about AI?
- For private-company audits, the AICPA's SAS 142 (effective for periods ending on or after December 15, 2022) recognizes automated tools and techniques and has the auditor evaluate information regardless of how it was obtained. For public-company audits, the PCAOB's amendments to AS 1105 and AS 2301 on technology-assisted analysis apply to audits of fiscal years beginning on or after December 15, 2025. Neither standard treats tool output as evidence on its own.
- Is "AI in auditing" the same as auditing AI?
- No. AI in auditing is auditors using AI tools to do audit work. Auditing AI means reviewing how a company itself uses AI: its tools, data rules and risks. If that's what you're after, it's a different engagement.
More on AI for Business
AI for Banks (2026): What a Community Bank or Credit Union Can Hand AI, What Stays Human, and How to Vet the Vendor
How a community bank or credit union should start with AI: draft procedures, member letters and credit memo narratives, keep credit decisions and BSA work human, sort every bank record before it goes near a tool, and run the AI vendor through third-party risk management like any other vendor.
ChatGPT Business vs Plus (2026): Which One a Small Business Should Actually Pay For
ChatGPT Plus vs Business, as of September 2026: Plus is $20 a month for one person and OpenAI may train on your chats unless you opt out; Business is $20 a seat billed annually or $25 monthly, two seats minimum, with no training by default, admin controls, shared projects for groups, and the Company Knowledge plugin. The side-by-side from OpenAI's pages, and three owner scenarios with a verdict each.
ChatGPT for Customer Service (2026): What Works, What Breaks, and What It Costs a Small Business
ChatGPT for customer service, as of September 2026: it drafts replies, answers policy questions from your own documents, and triages an inbox for $20 to $25 a seat a month on ChatGPT Business. It can't look up an order, issue a refund, or remember a customer unless you connect the system that holds them. The honest capability line, the plan you need, a setup in six steps, the failure modes, and the point where a real agent takes over.
ChatGPT Pulse (2026): What It Was, Why OpenAI Retired It, and How to Get the Morning Brief for Your Business
ChatGPT Pulse, the daily research cards OpenAI previewed for Pro users in September 2025, was sunset on June 17, 2026 with a 14-day wind-down. What it did, which plans ever had it, why scheduled tasks replaced it, and how an owner rebuilds the morning brief on a Plus or Business seat without handing ChatGPT the wrong things.
Written by
Justin McKelvey
I spent 12 years running snowboard and mountain-bike shops in Colorado, then taught myself to code. Since then: 50+ products shipped, Head of Product at a national sports platform, CTO at an AI startup. Now I put AI to work in real businesses in Austin, TX and remote, and write down what actually works.
Work with meWant more help? Get the free toolkit · Book a strategy call
Before you go
Score your AI readiness in 3 minutes.
Free, no call. 30 yes/no questions show where AI actually pays off in your business (and where it doesn't yet).
Get my readiness score →