Business of Tax: AI Brings Uncertainty and Opportunity to Tax Practice
Just as the computer revolution transformed tax practices a generation ago, new technologies involving artificial intelligence offer firms great opportunities and potential pitfalls.
As law firms contemplate, and accounting firms implement, new data tools such as predictive case analytics, they are also considering what effects those new tools will have not only on their everyday practices but also their business models.
Both Travis W. Thompson of Sideman & Bancroft LLP and Robert J. Kovacev of Norton Rose Fulbright US LLP compared the new AI tools available to tax practitioners and law firms to the computerized legal research revolution brought on by Westlaw and Lexis.
“This is where the world of law, the world of accounting, is going . . . the practice of law and the way we teach students law will involve these technologies in the same way that we learned Westlaw and Lexis, in one to five years. Mark my words,” Thompson told Tax Notes. The new technology will completely change the way the law is studied, presented, and researched, he said.
Tax practitioners contemplating the new tools should keep in mind the trouble that a lawyer who ignored Westlaw and Lexis 20 years ago would be in now with the changed nature of brief writing, Kovacev said.
The possibility of young tax lawyers being displaced by the efficiency of AI is a distinct one, according to Kovacev. “A lot of the young tax associate’s work is going to be in performing basic tax research, whether research for writing a brief or an opinion or providing planning advice,” he said, adding that the time freed up through efficiency gains could also be redirected toward client-oriented and analytical tasks, especially for more experienced lawyers.
Thompson agreed that the new technology can allow more of the grunt work, like document review, that was otherwise done by young associates to be done more efficiently. That said, young associates who grew up in a more technologically advanced world are also particularly enthusiastic about the new tools, he added.
Kovacev said that law firms still seem to be mostly exploring the new batch of AI tools, rather than preparing to fully implement them.
Accounting firms like PwC, on the other hand, seem to be all in on a wide variety of tools.
Marjorie Dhunjishah of PwC said the ability to use this sort of software is rapidly becoming a fundamental skill set at accounting firms. “Companies can grow from operating out of Mom and Dad’s garage to becoming a multinational company very quickly . . . and that rapid footprint expansion, the volume of data that your company has to gather, the volume of data they have access to, and then the amount of reporting to third parties that you are responsible for, just grows exponentially,” she said.
A wide variety of AI software is already on the market, and the price point for much of it has dropped enough that more firms are able to avail themselves of a broad range of tools, Dhunjishah said.
Dhunjishah separated the technology now in use by accounting firms into three buckets: tools for sifting and preparing data, analytics and data visualization tools, and robotics process automation tools.
Tax practitioners at accounting firms must handle large data sets to make the complex, interrelated, iterative calculations for the section 163(j) interest limitation, global intangible low-taxed income, the base erosion and antiabuse tax, and foreign-derived intangible income, according to Dhunjishah.
The new software doesn’t displace that grunt work but allows firms to run more models and scenarios in the same amount of time, Dhunjishah said. The increased efficiency can translate into more work done for the client in the same amount of time and at the same cost, she said. In other words, a better answer rather than a faster, cheaper one.
Thompson highlighted the potentially transformative effect that software like predictive case analytics could have on the legal industry.
The ability to drastically reduce hours of legal research and then provide a client with an objective percentage probability of success or failure in a case or tax planning scenario is a “complete game changer,” according to Thompson. The technology is essentially “moneyballing” the practice of law and reducing the human error component of it by shifting the focus to data, he said.
Skills Old and New
Tax professionals who spoke with Tax Notes agreed that learning how to use AI soon won't be an option.
“The modern lawyer . . . is going to have to get up to date and have a working knowledge of the emerging technology in a way that even I didn’t need to graduate [law school] four or five years ago,” Thompson said.
None of that comes as a surprise to University of California Irvine School of Law professor Omri Marian, who heads up the school’s graduate tax program.
His program introduced its first data science and tax practicum course this semester in cooperation with the data analytics software company Alteryx Inc. Marian said that for three years, he and other faculty at UCI had “dozens and dozens” of meetings with all different types of tax professionals to create a curriculum that would prepare students for the future and determine what skills are lacking in tax LLM graduates.
“The thing that came up over and over and over again in every single meeting was the use of technology in tax practice,” Marian told Tax Notes. “Several heads of tax from household-name multinational corporations told us in no uncertain terms that our students will be much more marketable if they know how to work with Alteryx Designer.”
All of the students now taking the inaugural data science course already have jobs lined up, according to Marian.
So far, however, UCI’s embrace of new data science tools as a core part of its curriculum looks to be more the exception than the rule among law schools.
Kovacev observed that while the use of robotic process automation by accounting firms hasn’t led to mass layoffs, it has shifted the profile of the type of professional those firms are looking for.
Accounting firms are starting to seek candidates with a data analytics or computer engineering background rather than a pure accounting or finance background, and it’s possible that a similar shift could be coming to tax law firms, Kovacev said.
“I don’t think law schools are generally ready to deal with that transition,” Kovacev said. He blamed institutional inertia, remarking that the law school faculty in charge of setting the curriculum are often “about as far removed from this kind of practice-oriented issue as one can imagine — with exceptions.”
Despite the importance of learning new skills, the rise of new technology in tax practice is also placing a premium on low-tech interpersonal skills, according to Thompson.
With tools like predictive case analytics, tax attorneys will be forced to engage more with opposing counsel, even as the software streamlines negotiations, especially if the opposition is using the same technology. During settlement negotiations, opposing counsels “can just get right to the point, really, potentially avoiding trial, keeping costs down for the client, and getting on and helping the next taxpayer,” Thompson said.
And while tax professionals have to understand how to use the new technology themselves, they also need to be able to turn around and communicate the outcomes to clients in a way that they can understand and act on, Thompson said.
Dhunjishah also emphasized that understanding and using new technology doesn’t displace the core need for subject matter competence; rather, she said she was optimistic that it would actually accelerate skills development. Instead of spending time on tasks like cutting and pasting or reconciling data, staff will be able to spend more time critically analyzing the law, formulas, and the results of calculations, she said.
It Takes a Village
For Kovacev and others, the burden of training the modern tax professional to use new technology in their practice shouldn’t fall just on schools.
“It’s everyone’s responsibility, including the person who wants to be a lawyer,” Kovacev said. Law students need to be aware of the shift in the industry and seek out opportunities to educate themselves, and law firms should provide training opportunities to their associates, he said.
That doesn’t mean the modern tax lawyer needs to learn how to code so they can write their own programs; rather, they just need to be able to use the tools, Kovacev said.
Dhunjishah described how PwC has undergone a companywide “digital upskilling” effort aimed at introducing employees to a wide variety of advanced software and providing baseline skills so that they can perform data analysis, automation, and visualization. The company sets aside time throughout the year for employees to train with their teams on various tools.
“Just like tax laws change, tools in the marketplace change over time,” Dhunjishah said.
PwC’s effort has extended to offering “digital academies” to its clients, as well as “working downstream into our pipeline” by collaborating with the academic community and advising professors on how to incorporate new technology into their curriculum, Dhunjishah said.
One promising sign, according to Thompson, is that older, more experienced practitioners are starting to embrace the use of new tools. Professional organizations like the American Bar Association and state bar associations are holding panels or continuing legal education events to discuss emerging tech issues, which they weren’t doing even two years ago, he said.
Kovacev said that changes to firms’ business models aren't the only potential disruption raised by use of the new AI tools. For one thing, AI may have implications for practitioners’ opinion-writing standards, including potential malpractice, he said.
For example, a practitioner might issue an opinion stating that a particular tax position is more likely than not to withstand scrutiny, but then predictive analytics software might say that the position has only a 40 percent chance. It’s not clear if a practitioner can go ahead and issue that opinion and be insulated from potential malpractice claims for taking a stance at odds with the software, according to Kovacev.
“From an ethical perspective and a Circular 230 perspective, if this becomes prevalent and becomes a recognized standard of tax practice, someone is going to have to grapple with those issues,” Kovacev said.
Practitioners should also be concerned about how IRS personnel will use the agency's version of the tools, Kovacev said. “Is this going to be a black-box algorithm, or are the inputs going to be transparent?” he asked.
Kovacev noted that there have already been reports of AI tools — for example, one focused on potential recidivism as a sentencing concern — inadvertently incorporating biases such as racial disparities into their underlying data sets. He gave the hypothetical of a taxpayer on audit declining to extend the assessment statute of limitations. That decision is absolutely within the taxpayer’s rights, but the algorithm could find that a denial is correlated with a tax understatement and advise issuance of a notice of deficiency, he said.
“The question is, will the humans in charge have A, the discretion or B, the desire to do anything to counteract that bias?” Kovacev asked. The national taxpayer advocate found that revenue agents with an AI reasonable-cause-assistant tool didn’t augment their decisions with insight from the program but merely followed whatever the program said without considering facts and circumstances outside the program’s nonexclusive list, he added.