Real-life ways small firms use AI

Artificial intelligence (AI) can increase CPA practice effectiveness and output without additional headcount. It can be particularly helpful in small firms where staffing, budgets, and resources are limited.

Firms are already using AI to speed up tax work, support audit and assurance work, and build simple tools to improve communication with clients. According to a 2025 CPA.com report, some firms have automated more than 80% of individual tax return preparation with the help of AI-assisted tools; others have used generative AI-based research applications to cut the time for document analysis within audit and advisory in half. AI tools are also used to automate tasks, such as bank reconciliations, transaction coding, and month-end close and reporting, which frees up accountants to spend more time on higher-value work such as advisory and analytics.

“Small practitioners with little or no specialized technology background are using free or inexpensive programs to create solutions to practice problems,” said Sarah Harris, CPA, cofounder of Agate CPA, a six-person firm with offices in Atlanta and Rogers, Ark.

AI is transforming tax preparation and bookkeeping the most rapidly, according to the CPA.com report. AI tools are also making inroads in audit and risk analysis, and AI-driven forecasting, budgeting, and KPI modeling tools are emerging for advisory services.

But as AI automation accelerates, clients’ trust hinges on how firms manage data and ensure that humans review and verify what AI produces. AI can improve efficiency, but professional judgment, skepticism, and client relationships remain firmly in the CPA’s hands.

The JofA asked four CPAs from small firms how they were putting AI to work. The firms have taken necessary steps to ensure data security and confidentiality in their efforts, such as ensuring that client information isn’t provided to a public AI platform or otherwise made public and that a human reviews the process.

Cheryl Hannafin, for example, uses AI mainly in building workflows, automations, and process improvements. “In other words, I generally use AI to build the plumbing rather than to handle client data,” she said. Hannafin, CPA, is the sole proprietor of Public Trust CPA LLC in St. Petersburg, Fla.

Agate CPA incorporated anti-spam and bot protection tools, such as CAPTCHA technology, in client-facing web forms it created to reduce the risk of automated attacks and unauthorized submissions. The firm also uses rate limiting and traffic monitoring controls that restrict how frequently forms can be submitted from a single source, which can prevent abuse and automated attacks. In addition, access controls and network-level protections allow form submissions only through approved channels and help identify suspicious activity before it reaches the firm’s systems.

PERFORMING COMPLEX TAX PLANNING AND RESEARCH

One Stop CPA is a fully digital firm headed by Fort Lauderdale, Fla.-based CEO and founder Brian Davis, CPA. The practice employs four accountants and three nonaccountants. The firm specializes in tax preparation, tax planning and advisory services, and bookkeeping, primarily for doctors, real estate investors, and entrepreneurs, some of whom fall into more than one of these categories.

Many of the firm’s clients are engaged in layered transactions in real estate, multiple entities, and ownership structures. Davis has a step-by-step process for using the tax research app BlueJ to meet their needs:

Step 1: Identify the problem

Clients come to the firm with complex questions about tax implications of a deal or the best way to structure a transaction. For example, a client considering a merger and acquisition (M&A) was moving to a new state and wanted advice on the state tax treatment of the potential deal. Davis is advising the client on taxes and payout strategy as the client meets with potential buyers.

Step 2: Perform AI-assisted tax research

Davis inputs the scenario and relevant facts into BlueJ, which he said provides an up-to-date, cited, source-backed analysis.

Step 3: Add in CPA interpretation

Davis reviews the output and applies his professional judgment to assess the client’s full picture. He determines the best strategic steps and asks any additional questions. “I make the ultimate decisions,” Davis said.

Step 4: Translate into client strategy

The software provides a draft summary of the findings, which Davis uses to create a detailed research memo that is used in the client’s tax plan.

Step 5: Save time with AI-assisted structuring

Davis started using ChatGPT when it was first introduced and built a library of templates, basically a collection of Google Docs and saved prompts, around repeatable workflows. This library of templates saves him significant time while still allowing him to customize the output before it goes to the client. He has learned that, after spending time to customize a precise output with ChatGPT, he can now ask: “Can you turn this into a reusable prompt so next time I can get a similar output in one step?” Davis uses enterprise versions only, so he is not training the GPT’s large language model with any client data.

The structured prompts ultimately result in client-facing deliverables that include slide presentations, road map summaries, infographics, meeting recap documents, or one-page PDF overviews for clients.

ChatGPT helps organize and structure information into a format that clients can review and understand. “The technical strategy and CPA analysis have already taken place before that point,” Davis said. “Usually at this point we close out an engagement with this deliverable and simultaneously share options for additional services, such as implementation support or compliance add-ons.”

Degree of difficulty: The tools are easy to use, but it’s important to build the skills needed to write effective prompts and interpret results.

Cost: BlueJ and similar tools are an investment, but Davis believes they are reasonable relative to their value. “The biggest ROI comes from time saved, better decisions made, and new opportunities,” he said.

Payoff: The firm can provide a higher-level service with less time spent on complicated tasks and no increase in workload. Davis used generative AI with the client considering M&A. By providing context and using the app’s follow-up questions, he was able to develop an answer within two hours — much less time than it would have taken previously. He used BlueJ for the technical tax research portions, including authority-based research and analyses of the tax implications of the transaction and residency move. ChatGPT helped organize the information, structure follow-up questions, summarize findings, and create client-facing deliverables and explanations for the client based on that research.

“This gives the CPA more time to think strategically, evaluate scenarios, and focus on applying professional judgment to the client’s specific situation, emphasizing that we do more than just compliance,” he said.

Takeaway: “AI is a starting point, not the final answer,” Davis said. “This works because it combines AI speed with CPA judgment.”

GATHERING MORE USEFUL PROSPECT INFORMATION

At Agate CPA, Harris and her cofounder, Alicyn McLeod, CPA, focus on tax planning and tax strategy, as well as monthly accounting for small businesses and wealth management. The firm has a niche in law firms but serves clients across many industries.

The partners were seeking to get more valuable information from the form on the firm’s website that prospects use to express interest. The form asked for name, email address, and the reason for a prospect’s interest in the firm, but that minimal level of detail “resulted in a lot of sales calls, but not a lot of conversions,” Harris said. “We weren’t spending our time wisely.” When the technology vendors had trouble understanding the robust and thorough solution the firm owners were seeking, they decided to build it themselves.

Harris took the lead on this project, turning to the firm’s existing tech tools to see what was possible. The firm has used ChatGPT since it was launched in 2022 and, after investigating several options, now also uses Claude and Lovable. “When we have a pain point somewhere, we like to see how we can use AI to automate our way out of it,” Harris said.

Degree of difficulty: On this project, Harris built the web form with Lovable, an AI-powered app builder, and used Claude to help embed and work it into the firm’s website. The technology was relatively easy to use and took little time. The real work involved developing the right format and questions for the new form.

To weed out companies too large, too small, or otherwise not a good fit, Harris and McLeod wanted to gather details such as whether prospects were business owners, their industry, their business’s number of employees and revenues, and whether it was a startup or long-standing company.

To highlight client needs and filter out clients that are potentially difficult or have unrealistic expectations, the new web form asks if the company has worked with a CPA before and provides multiple-choice questions on why they were seeking a new CPA.

All responses flow into ClickUp, Agate CPA’s practice management system, which enables the firm to track responses as soon as they come in, using ClickUp’s application programming interface and automations.

Cost: Minimal money spent.

Payoff: “We spend more time with prospects that are a potential good fit for us, rather than wasting time on calls that may go nowhere,” Harris said. The firm’s conversion rate rose about 25% as a result.

Takeaway: “This is something a small firm can readily do if they put in the time and thoughtfulness,” Harris said.

She advised firms to consider recurring challenges — anything that is not succeeding or projects that firm members dislike doing, do repeatedly, or believe aren’t worth the trouble. Once a pain point is selected, Harris said, open up a conversation with ChatGPT, Claude, or Lovable about opportunities to automate that area. “Just pick one thing,” she said. “AI will then guide you on what needs to be done and how to do it.”

TACKLING A RECURRING PAIN POINT FOR CLIENTS

Public Trust CPA LLC focuses on government and not-for-profit clients and uses specialized contractors as needed.

To solve her clients’ particular challenges, Hannafin has found that ChatGPT is good for developing tools for communications projects and helpfully remembers details about her situation and needs. Claude has more robust data analytics, and Gemini is best for video. She uses Microsoft Power Apps to experiment with coding. Copilot is also a strong component of her tech stack.

“I’m a lifelong Microsoft user,” she said. “Copilot integrates directly with Power Automate and has helped me to develop workflows and automations, in particular the approval automation.”

Hannafin wanted to make the approval process easier for her nonprofit clients. Many nonprofit organizations maintain internal controls that include rigorous approval processes. However, obtaining those approvals can be cumbersome, particularly when approvers are required to log in to separate systems, remember passwords, or navigate unfamiliar software.

“I didn’t want clients to have to sign up for another subscription just to approve an invoice. The more systems people have to learn and log in to, the longer it takes to begin implementing stronger controls,” Hannafin said.

To solve the problem, Hannafin used Microsoft Power Automate to build an automated workflow. Vendor invoices are sent to a dedicated client email address, which automatically triggers an electronic approval request using Adobe Sign. After the client approves the invoice, the workflow creates a bill in QuickBooks and attaches the signed approval documentation, creating a complete audit trail.

“With this process, clients can get started immediately — no accounts to create, no software to learn, and no additional logins required,” she said.

Degree of difficulty: Hannafin has used Copilot to develop workflows and automations, including for this project. Advice from ChatGPT also helped her navigate Power Automate. “As AI has gotten smarter, it has become easier to use,” she noted.

Cost: There was no additional expense for this workflow because the tools she used were available on her Microsoft 365 platform and Adobe.

For feedback and advice, Hannafin subscribes to AutomationTown, a community of accountants involved with AI, for what she considers to be a reasonable fee. “You’re part of a network where everybody’s got that same mindset, where we’re all in the same industry,” she said.

Payoff: She has easier access to client information without taking up their time. “I’m very happy as long as my client is happy,” she said.

Takeaway: As Hannafin investigates AI use, this initial win is a good starting point. “I’m definitely in an experimental stage, learning how I can monetize some of these workflows,” she said.

She continues to explore a variety of technologies, which can incur added subscription costs, but she considers it worth it. “I feel like I’m investing more of my time into developing systems that I can trust and feel confident in,” she said. “And once those are developed, I feel like the payoff will be tenfold.”

GETTING A BETTER HANDLE ON CLIENT SATISFACTION

High Rock Accounting in Scottsdale, Ariz., employs eight people and specializes in client advisory services (CAS), CFO services, and tax advisory, with a vertical in tech companies.

Everybody at the firm enhances their tech knowledge through experimentation, conferences, and learning about other firms’ experiences. Also, they are inspired to understand more by following up on client questions about how to use AI in their accounting departments. “We feel it’s on us to stay ahead of them,” said Liz Mason, CPA, CEO and founder of High Rock Accounting. “We encourage our team to see how to integrate AI into our daily lives.”

Mason wanted to ensure everybody at the firm thoroughly understood the client experience. However, Ashley Rhoden, High Rock’s COO, wasn’t happy with the client satisfaction and net promoter score programs they were using. She didn’t feel she could justify the high costs of the data they were getting but did want to continue gathering client feedback.

Rhoden decided to use Claude Code to create an app that integrates into the firm’s Karbon project management system. Then, she created accounts on the platforms required to host and run the software and copied the relevant code and information from Claude over to the platforms at play.

Claude co-created the plan, the process, the flow of information, and the tools that would be used. Rhoden used Claude to advise the process and Claude Code to write the code. She then created accounts on the platforms required to host and run the software and copied the relevant code and information from Claude over to the platforms at play.

The AI app automatically updates the firm’s active client list, asking clients for opinions after big projects or once a month, depending on the scores it finds. It then feeds the results into the system, so the team has visibility into client satisfaction faster and sooner than the products the firm was paying for.

Degree of difficulty: It took four or five hours to create a customized app. Rhoden gave Claude well-thought-out prompts describing what she wanted, and it came back with step-by-step instructions and a project plan.

Along the way, Rhoden was able to get advice on problems encountered.

“She asked Claude, ‘Tell me why I’m running into this roadblock. What do we need to do?'” Mason said.

Cost: The price of a Claude team account with very low ongoing maintenance expenses.

Payoff: Lower costs and the chance to monitor client satisfaction in real time and better understand which approaches were underperforming and which were working well.

“That information and feedback loop is just so valuable to the team,” Mason said.

Takeaway: Based on this success, the team has developed a list of apps that, once built, promise to make a big difference.

Also, Mason said, “We started hosting an ‘AI Happy Hour.'” She added that, “We use AI to build automations and integrate AI at the steps required to maintain the automations.”

The team spends the first 10 minutes discussing issues it would like to solve using automation and AI to maintain automations, such as time spent on repetitive tasks, roadblocks in workflow, or money spent on a program that the team could build itself. For example, in a few hours, Susan Wozena, CPA, the firm’s head of tax — with no tech background — created a working prototype of a new onboarding and tax organizer using Claude and Claude Code.

Mason believes that anyone who has what it takes to be an accountant can create their own tech tools. “I think we all need to be thinking about how these projects can change our lives,” Mason said. “Start asking what’s possible.”

---

About the author

Anita Dennis is a New Jersey-based freelance writer. To comment on this article or to suggest an idea for another article, contact Jeff Drew at Jeff.Drew@aicpa-cima.com.

---

LEARNING RESOURCES

Small Firm Success Series

his quarterly webcast, which is free to AICPA members, offers insights on staffing, technology, compliance, and practice management, equipping firms with strategies to thrive.

2pm–3pm ET, Aug. 20, Oct. 22, and Dec. 17

WEBCAST

PCPS membership

The Private Companies Practice Section (PCPS) is an add-on firm membership within the AICPA that supports CPA firms of all sizes by providing practical and customizable practice management resources.

SECTION

Digital CPA 2026

Join CPA.com in San Diego and online from Dec. 6—9 to discover emerging technologies, unlock actionable strategies, and collaborate with peers who are designing a new era of accounting.

CONFERENCE

AICPA ENGAGE

Firm practice management will be on the agenda at the biggest event in the accounting profession.

June 7–10 at the ARIA, Las Vegas

CONFERENCE

For more information or to make a purchase, go to aicpa-cima.com/cpe-learning or call 888-777-7077.

---

MEMBER RESOURCES

Engage365 communities

Engage365 communities bring together AICPA members to support one another in their careers, participate in discussions, and share resources. Use your aicpa-cima.com login to get started.

AICPA Private Companies Practice Section (PCPS) community

PCPS members are invited to join the PCPS Engage365 community. Whether you’re navigating firm management, exploring new technologies, or supporting your team’s development, this is your place to connect and grow together.

Engage365 tax communities

Engage365 is an AICPA members-only community where tax practitioners connect, share resources, and exchange real-world insights. Tax professionals can join two communities tailored to their practice area — Corporate and International Tax and Individual and Small Business Tax — making it easy to connect with peers and exchange perspectives and practical guidance.

Articles

“How to Build Reusable Skills in Anthropic’s Claude AI,” JofA, June 8, 2026

“9 Tips to Write More Effective AI Prompts,” JofA, May 5, 2026

“5 Human Competencies CPAs Need in the AI Age,” JofA, May 1, 2026

“Defining Commonly Used AI Terms,” JofA, May 1, 2026

“AICPA Small Firm Generative Artificial Intelligence (AI) Policy Template,” AICPA, April 29, 2026

“How are Finance Teams Really Using AI and Automation?” JofA, April 1, 2026

Podcast episodes

“Aligning With AI: Lisa Simpson on How to Overcome Sense of Overwhelm,” JofA, June 4, 2026

“What CPA.com’s CEO Sees Next for AI, Tax, and the Profession” JofA, May 28, 2026