Ensuring proper AI governance in tax and audit

The tax and audit world is in the middle of an artificial intelligence gold rush. But just like the race to tame the West, some outlaw behavior is emerging along the way. 

According to Thomson Reuters' latest Future of Professionals report, 81% of tax and audit professionals are now utilizing AI tools at least several times a week. But surprisingly, 35% say they use tools that their firm hasn't authorized for work. That creates a big governance risk for both practitioners and clients alike. Any of these unauthorized solutions — informally called "shadow AI" — that touch client engagements leave a hole in that record: no log of what the tool was given, what it produced, or whether anyone checked it against the standard the firm would otherwise apply.

The IRS has been watching this phenomenon closely, and this summer, the agency took its first formal steps to establishing some AI guardrails. It's a move that all tax and audit firms should note: Compliance requirements are coming for AI. It's time to get a handle on how your people are using it.

The risk is real

While it may not seem like unauthorized solutions cause that much harm, there's a large swath of evidence to the contrary. The use of AI has led to filings containing fabricated information, commonly called "hallucinations," which range from common errors to fake case citations. Penalties in these cases have included significant financial sanctions, public censure and mandatory ethics courses.

Still, according to the Thomson Reuters report, 28% of tax and audit professionals simply do not have access to professional-grade AI tools provided by their employers. In that vacuum, professionals who are likely using widely available consumer-grade chatbots and AI tools in their personal lives are experimenting on their own. The problem is, these consumer tools do not safeguard confidential data, ground outputs in authoritative content, or produce reasoning that can be explained and defended.

Anyone who has used a publicly available AI tool knows that errors are common. But it's far easier to catch an edited photo that's not quite right than comb through every tax case citation or line-item deduction and spot an automated error. And that's why the IRS is updating its guidance.

IRS reinforces governance

In June, the IRS Office of Professional Responsibility released introductory guidelines clarifying how existing professional standards — namely Circular 230 — apply to the use of artificial intelligence by tax practitioners. The guidance covers that while AI offers powerful tools, practitioners remain fully responsible for the accuracy of their work, the confidentiality of client data and the reasonableness of their fees. This means practitioners must thoroughly review all AI-generated documents for accuracy before submitting them to the IRS or providing them to clients.

Firm leadership also has a responsibility that requires the implementation of policies and procedures for the use of AI. The IRS's guidance says firms must provide comprehensive training to all staff on the risks and requirements of using AI, establish and document protocols for secure data handling and monitoring AI accuracy and thoroughly vet any third-party AI tools before use.

The best path forward

The pressure on tax professionals to win the AI race is palpable. Nearly half (48%) of tax and audit senior leaders report being already under "some" or "significant" financial pressure to act faster on AI. But that haste can create a vulnerability for tax professionals and firms, and they need to resist blindly running into the fray.

AI should be making its way into every tax professional's workflow and should be front of mind for executives looking to ensure those implementations are effective and smooth. But to truly thrive and avoid the punitive and reputational harm that can come with the use of shadow AI, every move needs to be measured. With the IRS now embedding AI governance into its standards of professional conduct, the stakes have gotten higher for firms that are not sweating the details on how their teams are using AI. Over the long term, those firms that take a measured, thoughtful approach to AI adoption will not only win more clients, they will achieve better outcomes and experience significant reputational gains along the way.