Have you used artificial intelligence (AI) today? Chances are that you have—even if you didn’t realize it. AI is increasingly showing up in all parts of our daily lives, from virtual assistants like Siri or Google Assistant to personalized recommendations (my Spotify DJ is playing as I work this morning). You’ve likely used AI through autocorrect and predictive text, as well as spam filters in email—and you may have even used it for your personal financial and tax services.
The Turn Towards AI
Companies are also turning to AI. Bain & Company, a global consultancy firm, recently reported that generative AI has become a business staple, with 95% of companies in the U.S. turning to it—a 12% increase in just over a year. Generative AI is typically built on large language models (LLMs) designed to understand and manipulate human language—you’ve seen it in applications like ChatGPT and Gemini—to generate new content, such as text or code (this differs from traditional AI, which primarily classifies or predicts outcomes). This isn’t just a private sector trend—it’s happening across both public and private sectors and in all industries. Danny Werfel, strategic advisory board member at alliant and former IRS Commissioner, says that for many companies, we’re at a critical inflection point, with more companies committing to spending money to support AI infrastructure and more leaders asking about how they can effectively integrate AI into their overall corporate strategy. And, Werfel says, it’s no different in the tax space. Technology is increasingly playing a significant role in tax and accounting firms—and at the IRS.
AI And The Role Of Tech At The IRS
Werfel knows a bit about the use of technology at the IRS, having served as Commissioner from 2023 to 2025 (he also previously served as the interim IRS Commissioner in 2013 after Steven T. Miller’s resignation). Werfel arrived after the 2022 tax season, which was, to say the least, a challenge. The federal agency was still recovering from the COVID-19 pandemic, which had caused delays in the tax filing season. There was a backlog of unfiled returns, and with many offices still not operating at full capacity, securing an in-person appointment was difficult—and reaching a real person on the phone was almost impossible.
There was a bright spot: Those issues gave the agency an opportunity to ramp up the use of technology as a solution. The complicating factor? Technology can be costly. The IRS didn’t yet have the infrastructure or budget to support large-scale changes, including the use of AI. So, the agency started small—it began with the call center. In the first iteration, when callers reached the IRS, their call would be routed by AI depending on the type of question they had. Sometimes, AI could answer questions—a simple question about a mailing address or due date might be answered without the need for a real person. Importantly, resolving that question quickly would get the caller out of line, freeing up space for other taxpayers. When a question required more than a simple answer, AI could route a caller to an IRS representative. The success of that model—in addition to a boost in funding from the Inflation Reduction Act—led to discussions about how AI could be used in other ways at the agency. Werfel notes that this was happening at a time when taxpayers were already using AI in other ways in the private sector—those LLMs like ChatGPT and Gemini were gaining popularity. As a result, he says, more and more people began to expect the kinds of answers that they received from those apps. That technology is also incorporated in search engines. Now, when you ask the internet a question, you expect an answer—not a list of options to research the answer on your own.
How Should AI Be Used?
That raised an interesting question: Should tax searches be done the same way? In other words, when you ask a question, should it give you reference points? Or should it give you an answer? Werfel dove into the issue by referring to what feels like an easy question: When is the tax filing date? If a taxpayer asks that question, he wonders, are they expecting a list of resources to click—perhaps directing them to the IRS website—or an easy answer, like April 15*?
Of course, when it comes to taxes, nothing is easy. That asterisk on April 15? That’s because that wouldn’t be an absolute answer. As Werfel notes, that could change depending on where you live—Outside of the country? In a state that observes Patriots Day? In a disaster area? What about whether you were on extension? More context is clearly needed. When it comes to LLMs, quality control is always a concern, especially in what we’d consider the early days, but that’s particularly true in complex areas like tax. There is, Werfel says, not yet a readiness for tax authorities to completely endorse an LLM functionality. That doesn’t mean that tax authorities aren’t looking for more opportunities to use that sort of functionality. Currently, when you call the IRS, an IRS agent will research your question in real-time, using internal resources such as the Internal Revenue Manual (IRM). That agent is trained on how to use the technology and conduct the necessary research to find the best answers for taxpayers. It would follow that you could build more efficiencies by deploying that same methodology inside the agency using LLMs. It wouldn’t be for public consumption—the IRS has its own closed system. And, since an LLM’s output is only as good as its input, that would require the agency to keep building out its resources and increase training. IRS employees who use the system would ask better questions than taxpayers because they would know not only what to ask, but how to ask. Werfel, a fantasy football enthusiast, explains it this way. When he searches the web for details on how to improve his game, “I know what to ask about fantasy football better than my mom does… I’ve done it for 15 years and she’s never done it.” That is custom quality control. You must learn how to perfect the input so that you can produce better output. That’s why, he explains, an LLM inside the IRS might appear similar to ChatGPT, but it will behave differently. That same strategy would also apply to the private sector. Each organization will need to develop its own approach to the use of AI and LLMs. Some companies, for example, may want to restrict employees to specific apps or search results (many companies already do this). Others, like Johns Hopkins University, where Werfel serves as Executive in Residence at the School of Government and Public Policy, may not want those limits because they are conducting large-scale research and require a wider net. AI can also be used in other ways, such as drafting agendas, summarizing meetings, and reducing overall administrative burdens for employees. Organizations, including the IRS and those in the private sector, are also incorporating AI in public-facing ways, such as chatbots for customer service and fraud detection.
The People Factor
This is not to say that full steam ahead on the tech front means eliminating people—far from it. Werfel believes that AI can bring higher efficiencies when used in concert with human interaction. That means training. That might be easy inside a closed IRS system, but what about in the private sector? It’s well established that lawyers, accountants, and tax professionals aren’t always the first to embrace technology (insert jokes about those still using AOL.com as their professional email addresses here).
In addition to training, some of the skills will come naturally, says Werfel. He recalls that when he first started his career, he didn’t receive the now-standard laptop—he got a dictaphone. He would dictate letters and memos and leave them with the secretary pool. (Those of us who remember those days think back fondly on the magic of returning to your desk the next morning with your typing neatly arranged desk.)
Over time, he said, that changed. He got used to word processing and email. The secretary pools disappeared. And the workforce shifted accordingly.
Werfel believes that the same kinds of shifts are happening today. The use of AI in the workplace, including the IRS, means that you can unlock new productivities. “The workplace,” he says, “will reshape.” However, he emphasizes that he does not think it means people will go away.
The Future Of AI In The Tax Space
“Companies that invest in the right spot,” Werfel says, will be successful. That includes understanding the readiness for change—and a little bit of inertia.
At the IRS, he says that will result in greater efficiency and less paper (National Taxpayer Advocate Erin Collins has referred to paper as IRS’s “kryptonite.”).
When used properly, it can also help the tax agency protect taxpayers who may be at risk of exploitation. The IRS has used its systems to detect scams targeting taxpayers, including those that employ AI voice cloning to impersonate IRS agents and those that create fake websites to steal personal data such as Social Security numbers and bank account information.
“AI,” Werfel says, “is an opportunity and a risk.”
While AI certainly has a role in tax—especially repeatable tasks with predictable outcomes—it cannot replace the need for informed and knowledgeable tax experts. Last year, the Taxpayer Advocate highlighted an informal review by the Washington Post that found “that two of the leading tax preparation companies’ chatbots provide inaccurate or irrelevant responses up to 50 percent of the time when initially asked 16 complex tax questions.” (Our Forbes informal review last year found some hits and misses—and that has continued this year following the rush by taxpayers to understand the provisions One Big Beautiful Bill Act). The Taxpayer Advocate has since warned that “taxpayers should not solely rely on AI-generated tax advice.”
Of course, technology isn’t perfect—that means there will be mistakes. The IRS has attempted to stay ahead of some of this by launching awareness campaigns when the agencies see issues emerging on social media. Those consumer alerts are key—Werfel compares them to the posted “Call Before You Dig” warnings. Before you hit “submit” on your keyboard, he says, think about whether you have enough information—or whether you need to ask questions.
When asking questions—especially tax questions—on the internet, Werfel says it’s important to think about the level of complexity. If you’re asking a simple question, it’s more than likely that tech can handle it. But if you’re asking something more complicated, like, he jokes, “a three-part question,” the tech may be more error-prone. That’s when you want to rely on proven methods, like working with a tax professional or relying on trusted websites like IRS.gov.