By Paul Mattox, CPA
Founder: W. Paul Mattox CPA
Raleigh, NC.
Artificial intelligence (AI) is quickly integrating into accounting, tax and financial services, and the Internal Revenue Service (IRS) is taking note. The agency recently announced plans to deploy AI to enhance audit supervision of large partnerships, aiming to increase efficiency. As the agency leverages AI, similar tools influence tax preparation.
Clients increasingly seek efficiency that AI models promise to deliver. However, this drive for convenience should come with caution, as current technology will not catch every complex tax nuance. A key concern is “hallucinations,” when AI generates inaccurate outputs rather than admitting it is uncertain. According to one of the leading platforms, OpenAI, hallucinations are inevitable.
Recommended Articles
Firm Management October 8, 2025
OPINION: Saving Hours with AI-Powered Accounting
For tax professionals, exercising caution is paramount. Questions on how to best use AI persist, alongside concerns over privacy and accuracy, especially when AI is marketed as a replacement for human expertise. While AI is useful to draft a report summary, it’s completely different to rely on the same technology to interpret complex tax laws.
Currently, there are no specific federal laws defining appropriate AI use in tax. So, accounting firms operate in a gray area while following general IRS guidance. There must be a responsible level accepted as standard since overdependence could lead to negative consequences.
Former IRS Commissioner Danny Werfel has cautioned that “AI is an opportunity and a risk,” highlighting both its potential and its capacity to create new challenges. While AI can streamline repetitive processes, the technology is fast-moving, and without clear guardrails, the chance for fraud greatly increases.
Select companies take a careful approach. PwC has internally rolled out “ChatPwC,” to assist with routine tasks, while ensuring experienced professionals retain control over critical decisions. PwC recognizes that while AI is revolutionizing finance, human oversight is essential.
In contrast, emerging AI-powered accounting firms are engaging in complex areas without adequate protections. This raises a red flag. If the AI gets it wrong, clients can be exposed to audits or fines.
As the tax extension deadline approaches, many clients look for quick-turnaround services. Certain firms advertise themselves as a fast and easy alternative to traditional CPAs. Some take an unscrupulous approach by undervaluing experienced, human CPAs in favor of AI-powered filings. Even worse, some claim their models automate all, or nearly all, the filing process. For select tasks, AI lacks the proper experience required to apply the law and overuse could result in legal violations or noncompliance.
It is essential to not misunderstand this perspective. Innovation is valuable and should be supported, however, human oversight and established guidelines are essential.
To illustrate this issue, take the research and development (R&D) tax credit for example, an incentive encouraging companies to invest in R&D. Accountants are key to identify which expenses qualify and often gather the necessary documentation. Since it can be a substantial financial write-off, attention to detail is crucial, not only for accuracy, but to maximize the credit.
The R&D credit requires a nuanced understanding of IRS rules, including detailed documentation of in-depth employee interviews to identify time spent on R&D activities. When records are not detailed, interviews help shore up outstanding questions on eligibility. AI-powered solutions can skip the interview process entirely, which is needed to substantiate qualified activities. In fact, some firms use this as a selling point.
Determining eligibility cannot be easily automated: each business is unique, and many have nuanced interpretations of what qualifies. Unfortunately, current AI models cannot reliably handle all nuances, so professional human judgment is imperative. If AI completes most of this process, it heightens the risk of audits and legal issues.
Another concern is the compensation structures of certain firms: which take a percentage of their client’s tax credit. While this appears to be standard, in practice, it could incentivize firms to claim that clients are eligible. AI models can “hallucinate” and generate false information to err on the side of qualifying for a credit even if the activity doesn’t truly qualify, with no way to completely avoid these inaccuracies.
Additionally, AI operates without standards that professional CPAs must adhere to. When paired with the lack of guidelines, there are questions to address: How are the tools trained? How often do they make errors? Even the best models have error rates as high as 30 percent, so how can clients and accountants be protected from AI-related scams?
Tax fraud, even if unintentional, is a criminal offense. Firms need to ensure AI is not excessively relied upon. Many AI-driven firms don’t mention that clients themselves are held legally liable for any mistakes. If and how these firms would represent clients in the case of an audit and court proceedings remains unclear.
Accounting firms should never rely on AI alone. While AI is a powerful tool, it cannot replace professional judgment. Blind trust in AI algorithms is not viable.
The need for regulatory guidance for the tax industry is clear. Without oversight and guardrails, compliance issues can continue to grow and outpace current practice.
Thanks for reading CPA Practice Advisor!
Subscribe Already registered? Log In
Need more information? Read the FAQs
