AI for small accounting firms, without the hype
Where AI saves real time in a CPA practice, where it creates liability, and how to start before the next tax season hits.
What January actually looks like
January 15 arrives and a solo CPA in Centerville is still getting engagement letters signed while the first 1099 deadline is already on the calendar. A 4-person bookkeeping firm in Murray is finishing a December year-end close for one client while fielding deduction questions from three others by email. A tax-only practice in Provo has twelve new client inquiries from December referrals sitting in a shared inbox. None of this goes away with AI. The human judgment, the return review, the client relationship — those stay yours. What AI can do is clear the draft-and-prep layer that is eating the hours you need for the actual work.
Where AI actually helps
Client email drafting
Deduction questions, K-1 follow-up requests, deadline reminders for extensions. A solid first draft in under a minute. You read it, adjust the tone, send it. The conversation still sounds like you, not a form letter.
First-pass tax code research
A client asks about a Section 199A edge case or a new depreciation rule. AI can surface a readable summary of the relevant code section or a recent IRS FAQ. You verify it before it goes anywhere near advice.
Engagement letters and proposals
A new client comes in for bookkeeping plus payroll. AI drafts the scope language from your notes. You spend ten minutes tightening it instead of forty-five building from scratch. The same logic applies to service proposals.
Summarizing long client documents
Board minutes, loan agreements, multi-year financials that a client sends as background. AI can turn a 40-page memo into a one-page brief. You read the brief, pull the numbers that matter, and ask for the full document only when you need it.
Where AI is overhyped
The part most AI vendors skip.
AI giving tax advice directly to clients
Clients will rely on it. The accuracy is inconsistent enough on specific code questions that the liability risk is real. A chatbot on your website that answers "can I deduct my home office?" is not a convenience feature, it is an exposure. The answer has to come from you.
AI doing the actual return prep or reconciliation
Tools like ChatGPT do not connect to QuickBooks, Drake, or UltraTax without real development work in between. Return prep still requires a human eye on the numbers. AI can help around the edges of the workflow, not replace the core of it.
Feeding raw client data into a public AI tool
The free tier of ChatGPT is not built for files containing client PII, SSNs, or financial records. Enterprise tiers with data processing agreements exist, but require intentional setup. Using the wrong tier is a confidentiality problem, not a minor inconvenience. This is one of the first things the strategy session covers.
How Doman AI works for accountants
Most small firms don't need a custom AI build. They need three or four tools that fit how they actually work, set up in a way that respects data handling requirements, and a clear picture of how AI describes their practice when a small business owner in their city asks for an accountant. Accounting is a national market, but referrals still start local.
AI Visibility Report
How ChatGPT, Claude, Gemini, and Perplexity describe your practice when a small business owner asks for an accountant in your area. Delivered within one business day. Start here to see where you stand before tax season.
Get the Visibility Report →AI Strategy Session
90 minutes with Laird to map which AI tools belong in your workflow, how to handle data the right way, and where to focus for the most visible return on your time. Includes a 2-week follow-up. Virtual or in person.
Book a Strategy Session →Not sure where to start? The free AI Grader takes two minutes. Or read more about Doman AI.
Questions
Is it safe to use AI with client financial data?
It depends on the tool and the tier. The free public version of ChatGPT is not appropriate for raw client data. Enterprise tiers with data processing agreements exist, but need proper setup. The strategy session covers which tools handle data in a way that is actually defensible for a CPA practice.
Will AI replace my staff accountants?
No. AI can draft a client email or produce a first-pass summary of a board memo. It cannot review a return, catch a classification error, or talk a client through a notice they just received. It takes prep time off your staff's plate, not judgment.
Is this for solo CPAs or multi-partner firms?
Both. The $99 report and $999 session are sized for solo practitioners and firms up to about five people. Larger practices with specific workflow or integration questions can ask about a custom engagement.
Can AI integrate with QuickBooks or my tax software?
Not meaningfully out of the box. ChatGPT does not connect to QuickBooks. Integrations exist but require real development work. What AI handles well today is the writing and research layer around your existing tools, not replacing them.
Should I get this before tax season starts?
Yes, if you want to use it during tax season. Setting up a few good prompts and workflows in November or December means you have them ready when January hits. Starting in March while you are already in it is harder.
What does the $99 AI Visibility Report show me?
How ChatGPT, Claude, Gemini, and Perplexity describe your practice when a small business owner in your area asks for an accountant. For most small firms, the answer is nothing at all. The report tells you what to fix first. More on the report, or about Laird.
Ready to put AI to work in your practice?
Start with the $99 snapshot. Or jump straight to a 90-minute session and leave with a concrete plan before the next filing season.
Or try the free AI Grader first →
Questions? laird@domangrowth.com