Arguably, artificial intelligence (AI) is the most talked about subject impacting the CPA profession today. Experts declare this is just the tip of the iceberg with endless potential forthcoming.

Despite our diverse and varied experiences, and our age differences, we both agree that AI has been launched, landed and is growing exponentially. AI is, and will continue to be, a powerful tool for CPA firms of all sizes. However, there are some potential minefields that can explode — namely divulging client personal information, failure to corroborate tax and accounting research and failure to disclose to clients that AI models will be used in providing professional services.

Client Personal Information

We have seen too many tax preparers send unencrypted/redacted tax returns to sites that will prepare a return seeking to optimize the return prepared by the firm. In doing so, significant personal information is exposed causing professional liability claims and regulatory sanctions under Circular 230/Internal Revenue Code section 7216 and state boards of accountancy under their confidentiality rules.

Failure to Corroborate AI Search Findings

We believe AI research findings should be the beginning point, not the end point. Clear documentation of the accountant’s research is essential in defending a professional liability claim. In certain scenarios, it’s best to have multiple firm staff run independent findings. Corroborating AI findings will cement knowledge and undoubtably aid in defending professional liability claims.

Disclosure of AI

We highly recommend disclosure of the use of AI and, in certain cases, an opt-out provision for clients to check. Many clients aren’t in favor of AI models or haven’t fully transitioned to AI. Client concerns and misunderstandings can be managed through full disclosure. We recommend the following be included in firm engagement letters:

While providing the services set forth in this Agreement, in order to enhance the quality and efficiency of services provided, FIRM may use tools incorporating artificial intelligence, including but not limited to, algorithms, machine learning and automated process (“AI”). The use of such technologies will comply with applicable laws and regulations. FIRM will use the appropriate due diligence and best practices to ensure that any AI tools and methods are secure. Client data shall not be used, in whole or in part, to train, refine or improve AI models. All outputs generated by AI tools shall be subject to review and validation by FIRM prior to inclusion in any deliverable. 

Turning AI Into a Practical Tool

Concerns about privacy, corroborating research and disclosure are important because AI is only useful if it is used responsibly. However, in considering the other side of the conversation — using AI inside a working accounting firm and figuring out where it actually helps — the focus is on how AI can make a firm better. It’s not about AI replacing accountants or making professional judgment less important. AI is a tool that can help firms work more efficiently and spend less time starting from a blank page.

Practical Uses Inside a Firm

Firms can use AI in practical ways throughout the day. It helps to draft and clean up emails, summarize meetings and long conversations, organize follow-up items and turn rough notes or recordings into the beginning of a procedure. It can also be used to brainstorm ideas, organize research, create training material and repurpose presentations or webinars into content. AI should not finish the work, but it can help get to a useful first draft much faster.

Creating Capacity

Where AI is a potential superpower for accounting firms is in creating capacity. A lot of a firm’s time disappears between the actual accounting and tax work. Staff are writing emails, documenting conversations, creating instructions, looking through information and following up on open items. Saving a few minutes on each of those tasks may not sound dramatic, but across an entire firm it adds up. That creates more time for review, client conversations, team training and eventually more advisory work.

Human Review, Privacy and Guardrails

Those benefits do not remove the concerns raised about AI. No one can assume that everything AI produces is correct. Sometimes it is wrong. Sometimes it sounds too polished or generic. Sometimes it completely misses the point. Someone still has to review the facts, adjust the tone and make sure the final result reflects the authors’ views. When AI is used for research, the answer is only a starting point and still needs to be verified against authoritative sources. The professional responsibility does not get handed over to the software.

The same is true for privacy and security. Firms need to understand which tools their teams are using, how those tools handle information and what client data should never be entered. Ignoring AI completely is not the answer. A basic policy, approved tools, training and a clear review process are more realistic than pretending it is not happening. Disclosure also matters because clients should understand how technology may be used in delivering services.

Start Small and Build from There

A firm does not need a massive AI strategy before getting started. Pick one low-risk, repetitive task and test it. Maybe it is summarizing an internal meeting, drafting a routine email or documenting a process. Decide what information can be used and who will review the result. Then see whether it actually saves time and improves consistency. If it works, document the process and build from there. If it creates more cleanup than value, move on.

AI by itself is neither a superpower nor a minefield. The difference is how a firm uses it. AI will not fix a bad workflow; in some cases, it may just help a bad workflow move faster. It works best when it is used by a firm that already cares about clear processes, consistent client experiences, security and human review. AI is not about automating everything or pretending the profession has it all figured out. It’s about continuing to test practical ways to remove unnecessary work while protecting the judgment and trust clients expect.