The IRS and tax practitioners have ways to leverage artificial intelligence and other digital solutions when assisting certain low-income taxpayers who do not have access to or are not comfortable using modern technologies, according to two tax professors.
San Jose State University Professor Annette Nellen and UC Law San Francisco Professor Amy Spivey spoke February 20 at the American Bar Association’s Midyear Tax Meeting in Los Angeles, California, on the role tech can play in evolving tax administration based on their experience with Low-Income Taxpayer Clinics (LITCs).
Digital divide.
Tax professionals should take data indicating that the vast majority of individuals use the internet or have a smartphone with a grain of salt, said Nellen. Such data does not take into account how someone may not be “using the internet the same” way as another, so do not “assume” a client is “an expert,” she cautioned.
Nellen advised practitioners to consider the “digital divide,” a term used in a February 2023 report issued by the Government Accountability Office (GAO) that refers to the “gap between those with and without access to broadband.” According to the GAO, “millions of Americans either don’t have broadband available to them or can’t afford to pay for it.”
Many taxpayers do not have a computer and use their smartphone for all their internet needs, the panel discussed. But as moderator Don Crawford of Baker McKenzie remarked, many government websites are not optimized for mobile devices. Spivey commented that the California Franchise Tax Board is an example of a state-level tax agency that is mindful of this and updated its site to be mobile-friendly.
Nellen reflected on how often former IRS Commissioner Danny Werfel would say, “if someone wants to interact with the IRS digitally, they should be able to do so.” While that is an understandable goal, she said, it does not factor in the realities of the IRS’ capacities and parts of the Tax Code itself. For instance, by law certain forms must be administered by mail, technically prohibiting the development of an electronic version.
In an ideal world, a taxpayer would be able to easily make real-time updates to their tax account, Nellen said. Someone looking to make a Code Sec. 83(b) election, as an example, should not need to do anything beyond logging into their online account on the IRS website and checking a box to update their records.
Taxpayer accounts should be a true one-stop shop and allow users to see the full picture of their tax situation, since relevant data exists “out there” in the digital space already, Nellen continued.
AI.
The IRS could develop their AI capabilities to better tailor information to taxpayers, the panel suggested. By integrating an AI model to a taxpayer’s account, taxpayers could circumvent navigating the IRS website, which can be overwhelming or intimidating to users who wouldn’t know where exactly to look for specific answers, said Nellen.
As for what practitioners can use AI for when working with low-income clients, Spivey also advised delivering to the taxpayer customized and carefully curated resource materials. She demonstrated how AI tools can now generate documents like spreadsheets and slide presentations.
A practitioner can enter a prompt for a client learning the basics of available tax breaks they may be eligible for, like the Earned Income Tax Credit and the Child Tax Credit. This quickly-generated content supplements tax consultation, Spivey explained. And the language learning models can be fed writing samples so the materials are written how the tax professional would write them.
AI can also assist practitioners themselves. In another demonstration, Spivey generated a guide for a hypothetical professional interested in opening their own LITC. This would be helpful “if you don’t know where to start — maybe you’re new to the area, or you aren’t familiar with the types of organizations that… help low income individuals in your coverage area,” said Spivey.
Spivey advocated for the practical, every-day application of AI. Practitioners should consider what tasks they wish they could take off their plate or expedite. “[Y]ou can tell it to draft you a template for a retainer agreement that you can then edit and rewrite,” she said.
But she warned practitioners to not use “open source” AI models like ChatGPT. Instead, a firm should invest in a paid subscription to a tool designed for business use. And guardrails need to be in place to ensure the firm isn’t giving clients incorrect or misleading information, as AI is not “always right” and can put a business at legal risk.
Even for personal research purposes, there needs to be that human element providing governance, Spivey recommended. The best models are not immune to hallucinating answers, such as inventing fictional cases and citations.
“Hallucinations are a really, really important thing to look out for,” said Crawford. “Not just from an accuracy standpoint, but certainly from a professional and integrity standpoint.”
Take your tax and accounting research to the next level with Checkpoint Edge and CoCounsel. Get instant access to AI-assisted research, expert-approved answers, and cutting-edge tools like Advisory Maps and State Charts. Try it today and transform the way you work! Subscribe now and discover a smarter way to find answers.
