AI In Tax: Possibilities and Challenges

Introduction

Welcome and Introduction

Welcome everyone, I'm Kara Griffith, the president and CEO of Tax Analysts. Thank you for joining us today for our discussion on artificial intelligence (AI) and the future of tax. In every industry in our economy, people want to learn what AI and generative tools like ChatGPT can do. Closer to home, we want to know the impact these new technologies will have on the legal, tax, and accounting industries.

We're delighted that you're with us today as we explore this significant issue. Today's event is another in Tax Analysts' series of public discussions called "Taxing Issues." We've been bringing the tax community together with leading policymakers and experts for bipartisan discussions on the future of tax policy and administration.

As always, we welcome your feedback on how we can make our webinars more useful and your suggestions on future webinar topics. You can send your feedback and suggestions to events@taxanalyst.org. For those of you who would like CPE credit today, please participate in the program for at least 50 minutes and answer at least three of the poll questions, which will appear throughout the webinar on the right side of the viewing window above the comment section.

To receive credit for your answers, you need to register with Pigeonhole, our interactive Q&A system, by providing your name and email. Also, please remain in the default viewing window, because you won't be able to see the questions if you expand the program to full screen. The course materials, which were sent out in advance, are available for download under the viewing window.

Importance and Impact of AI

Needless to say, AI is the new big thing. It's having a major impact on businesses, nonprofits, and other sectors of our economy. In the legal, tax, and accounting space, its ability to provide automated services is raising serious concerns if not some trepidation. As we all know, tax preparation is one thing, but real-time insights and personalized advice is quite another.

On the upside, AI can provide many benefits. It can strengthen data analysis and forecasting. Practitioners can use it to stay abreast of new regulations, dig through financial data, and spot outlier transactions. It can perform what were once manual tasks, reducing human error and freeing up practitioners to do more sophisticated work. At the same time, AI can enable practitioners to provide better, more targeted advice to their clients.

Many tax accounting law firms are jumping on board. Recently, PwC announced a global partnership with Harvey, an AI startup, to provide PwC's legal business solutions professionals access to Harvey's AI platform built on OpenAI and ChatGPT technology. ChatGPT has attracted much attention because it can understand human language prompts and produce human-like dialogue and content. Generative AI doesn’t just spit out information found in a database; it generates new content.

In a legal setting, Harvey uses natural language processing, machine learning, and data analytics to automate and enhance certain aspects of legal work. PwC hopes Harvey will generate insights and recommendations from large volumes of data, giving professionals better information and enabling them to find better solutions faster. The firm has assured everyone that an actual human being will oversee the process and review all the outputs.

PwC is far from alone. Allen and Overy, a London-based law firm, is also partnering with Harvey. KPMG has announced it will incorporate AI technology into its global audit platform through a collaboration with MindBridge Analytics. Auditors will thus be able to use embedded audit intelligence tools, visualize analytics, and other tools to perform risk analysis and assessment. The list is expected to grow and include government agencies as well.

The Panel and Introduction of the Speakers

We have a terrific panel today to discuss these issues and many more:

Sharda Shiruru, a retired partner from EY, an independent board member, startup mentor, and digital transformation leader.
Caitlin Tharp, an associate at Steptoe & Johnson who writes and speaks frequently about the intersection of tax and new technologies.
Mindy Herzfeld, a professor at the University of Florida Levin School of Law, who teaches a class on AI and tax and is also of counsel for the taxation, mergers, and acquisitions practice at Potomac Law Group.

Discussion on AI and Tools in Legal and Tax

Mindy: I am teaching a class on AI and tax. When thinking about the different AI tools, it might be useful first to understand that everyone is already incorporating AI into their practice in some form. The question is how far one wants to go ahead.

For instance, there's a significant industry called "legal tech," designed for legal practices. Tools used in this industry are significant for processing large amounts of data, contract review, due diligence, etc. Specific tax tools, like Blue Jay Legal, which is more of a research tool, provide answers to legal questions based on a considerable database of tax authorities.

Another tool we have used in my class is ChatGPT, which, while impressive, poses its challenges. For example, we used it to determine if a financial instrument was debt or equity. ChatGPT initially did not distinguish well between these, but with further probing, it managed to provide acceptable responses. However, the risk of ChatGPT "hallucinating" information is a concern, as it can sometimes generate incorrect or fabricated responses.

Sharda: Regarding AI in tax practices, I have observed it transforming the way we gather, clean, and input data. At EY, we embarked on automation projects around back in 2015. We focused on using robotic process automation tools to handle repetitive tasks, such as preparing sales and use tax returns, which significantly improved efficiency. Over time, we expanded into more complex areas using machine learning to detect errors and outliers in the data.

Potential Pitfalls and Ethics

Caitlin: There are potential pitfalls. ChatGPT is not a research database but a large language learning model, meaning it can generate well-written but sometimes incorrect or fabricated text. This issue became apparent when a fake case citation generated by ChatGPT was submitted in a legal setting, leading to real-world consequences.

Guardrails are necessary to avoid such pitfalls. For example, ALAS (the Attorney's Liability Assurance Society) suggests prohibiting its use for substantive client legal work due to concerns about competence, confidentiality, and the authenticity of the information.

Mindy: It's essential to distinguish the types of AI tools and how they are used. There’s privacy and bias concern, especially when considering Tax Administration. The IRS already uses AI for auditing, but the lack of transparency raises questions regarding oversight and bias in selecting cases.

AI and Tax Administration

Mindy: AI tools could be a magic pill for Tax Administration, although there are challenges. For instance, my student reverse-engineered the IRS’s interactive tax assistant algorithm and found ChatGPT provided similar answers, often more elaborately. This raises questions about transparency and accuracy for AI tools that the IRS uses.

Privacy Concerns

Caitlin: ChatGPT learns from user inputs, which raises significant privacy concerns. Information entered into the model could be shared unintentionally with other users. This potential leakage could breach client confidentiality and attorney-client privilege, which LSAS finds problematic under professional conduct rules.

Mindy: Firms making significant investments in their private AI tools are an attempt to mitigate these concerns, as captured in the PwC-Harvey partnership example. However, there’s a debate over whether this makes AI tools accessible primarily to big firms, possibly creating an unbalanced professional landscape.

Future Workforce and Training

Caitlin: AI could replace some or most of the tasks performed by first and second-year associates, such as form filling and standard research. However, human judgment remains crucial, and AI may require verifying outputs provided by such technology.

Mindy: The shift requires new professionals to jump early into more sophisticated roles. This is why integrating AI knowledge into law school curricula is critical. OpenAI and generative AI included in courses will enable upcoming professionals to seamlessly transition into roles that require higher judgment and skills from the onset.

Keywords

Keyword

Artificial Intelligence (AI)
ChatGPT
Tax Practice
Legal Tech
Generative AI
Data Analysis
Tax Administration
Privacy Concerns
Automation
Legal Research

FAQ

Q: What are the potential benefits of AI for tax professionals? A: AI can strengthen data analysis, forecasting, help stay abreast of new regulations, reduce human error, and free up practitioners for more sophisticated work.

Q: What are some of the pitfalls of using AI in legal practice? A: AI tools like ChatGPT can sometimes generate incorrect or fabricated responses (hallucinations). They also pose privacy risks since inputs can be unintentionally shared.

Q: How can AI be integrated into law firms without compromising privacy? A: Significant investments into proprietary AI systems, like the PwC-Harvey partnership, ensure that client data is processed securely and confidentially within internal databases.

Q: Will AI replace the tasks performed by junior associates? A: AI can perform many repetitive tasks typically done by junior associates, but human judgment and review are still essential. Law schools and training programs need to adapt to train new professionals to work with these tools effectively.

Q: What does AI mean for the future of Tax Administration? A: AI can enhance efficiency and data processing capabilities of agencies like the IRS, but it raises concerns about transparency and bias in auditing and other processes.

Q: How should tax professionals and legal practitioners prepare for the AI revolution? A: Staying informed about AI technologies, incorporating them into practice, and constant learning to understand their capabilities and limitations is crucial.