EY helps clients create long-term value for all stakeholders. Enabled by data and technology, our services and solutions provide trust through assurance and help clients transform, grow and operate.
At EY, our purpose is building a better working world. The insights and services we provide help to create long-term value for clients, people and society, and to build trust in the capital markets.
AI uncovered – Part 1: summarization and legal insights in tax with AI
In the second episode in our special series, ‘GenAI in Tax: The next wave,’ Tax Partners, Manoj Rathi and Giselle Barboza explore the transformative impact of Artificial Intelligence on tax processes and its role in distilling complex tax documents into essential insights. Discover how GenAI's advanced algorithms simplify the intricate world of tax legislation and learn about the critical human element in the AI-assisted future of tax research. Tune in for an engaging discussion on the synergy between AI and tax expertise.
GenAI uses natural language processing to transform voluminous tax documents into concise, relevant summaries, enhancing tax professionals’ efficiency.
Training on large, diverse, and pre-validated proprietary datasets improves the accuracy of GenAI-generated tax summaries.
Future AI advancements will provide personalized drafting styles and intuitive interfaces, integrated with other AI services to enhance precision in tax work.
GenAI is your co-pilot in tax, not auto-pilot; it aids but also requires validation and review by professionals.
For your convenience, a full text transcript of this podcast is available on the link below:
Manoj Rathi: Greetings to all our viewers and welcome to the second episode of our special series ‘GenAI in Tax: The next wave.’ I am Manoj Rathi, Tax Partner with EY and I have Giselle Barboza with me, a Partner, Tax, EY India who specializes in transfer pricing and is our Tax Talent Leader. She is also very passionate about the use of GenAI in tax, and together we will weigh in on Artificial Intelligence (AI)'s spectacular ability to crystallize complex tax documents into distilled wisdom. So, stay with us as we unravel the machinery behind AI's complex engines, its relevance in various applications, and the importance of the human-in-the-loop.
Giselle, thanks for joining us.
Giselle Barboza: Pleasure to be here, Manoj.
Manoj Rathi: Let me start by asking you to share your take on what AI summarization involves, and how can it be relevant from a tax perspective?
Giselle Barboza: As tax professionals, we have both been in the business for more than 10 years. What it (AI) really does is look at large documents, which could be case laws, regulations like those from the OECD, or it could the tax code itself. It has the ability to summarize in a very clear and concise manner, and to take out what is irrelevant. Let me give you an example. If I am reviewing a case law and I need it summarized into the key points; AI has the ability to do that. That is not to be mistaken with précis writing, where you are just extracting. AI has the ability to understand what is relevant and decipher tax legal jargon. So, it is a very compelling technology. I have seen young professionals adapt to it, and they are turbocharged using GenAI.
Manoj Rathi: A lot of tax data that tax professionals are consuming today includes large, complex case laws or paper books that could run into hundreds or even thousands of pages. How does AI discern relevant information from these complex, large documents?
Giselle Barboza: GenAI essentially uses natural language processing algorithms. It processes volumes of data and, essentially, looks at patterns. For instance, today, if I were to ask someone who is not a tax professional about their ‘principal purpose’, they would probably tell me about their principal purpose in life. But if I ask a corporate tax professional about the ‘principal purpose test’, they will understand the context. GenAI, when trained on tax datasets, understands what the principal purpose test. If it is trained on transfer pricing documentation, it understands what DEMPE (Development, Enhancement, Maintenance, Protection, and Exploitation) is, what FAR (Function, Assets, and Risks) is. So, what is compelling is that it recognizes natural language with context and pattern recognition and is able to summarize appropriately.
There are essentially two methods of summarization: one is the extractive method and the other is the abstractive method. As the word suggests, in the extractive method, GenAI takes words verbatim from the text, removes the irrelevant parts, and you get almost like a précis. The abstractive method is quite marvellous because it involves a bit of creativity. GenAI is able to look at pages of data and summarize like a human would in natural language. Those are essentially the two kinds of methods.
Manoj Rathi: That is very interesting, Giselle. I would also like to know how one could put in more measures to ensure that AI more accurately reflects the content and the nuances from a tax perspective.
Giselle Barboza: When we work with GenAI in tax, we require two elements. We are very keen on being accurate as possible, and also up to date, because there is so much evolving legislation and jurisprudence. So, essentially, what makes it more accurate is working with large and diverse datasets. The second is that, in a closed AI environment where you put in data that is already pre-validated, the accuracy would be enhanced. So, if I were to look at AI tools which are run on proprietary data, they would be far more accurate than if they were learning off the internet.
That is one point. The other is that businesses need to be extremely cautious about the kind of governance they are building around their GenAI. How are they dealing with data? How are they dealing with privacy? There is a lot of work to be done in this space because, while it is compelling, there is a lot of responsibility that comes with it.
Manoj Rathi: I agree. From a governance framework perspective or data security perspective, these are very critical elements in context of what we are talking about. How should tax professionals go about vetting data to ensure higher adoption of GenAI in tax?
Giselle Barboza: I am asked this question quite often: how do we rely on GenAI? My answer is that this is a tool in your arsenal. As a tax partner, even today, if I get the best document and I have to sign off on it, I still need to validate. So, I would request everyone to think of GenAI as your assistant, as your co-pilot; it is certainly not auto-pilot. Validation, reviews, and examining the process output are absolutely critical and will continue to be for tax professionals.
Manoj Rathi: That makes a lot of sense because we are talking about a technology that can become phenomenally smarter with every incremental layer of learning, which makes the vetting and the learning layers even more critical. As we look forward to the future, what advancements do you foresee in AI summarization over the next few years?
Giselle Barboza: That is a pretty loaded question, and I am not a technologist, but in my humble opinion, if I were to look at what the capabilities could be for tax and legal, I would say it is certainly going to get more precise. There will be the ability to customize as per your persona. So, if I draft in a certain style, it will be able to mimic that. It is going to be far more intuitive, and it will be enhanced through integration with other AI services. So, there is lots that is going to happen. It may take an entire podcast to discuss that, but we all need to keep abreast of what is happening.
Manoj Rathi: That makes a lot of sense, Giselle. Very insightful. Thank you for clearly guiding our viewers through the seemingly complex web of tax documentation, helping them understand the journey from natural language to sophisticated documentation, and the synergy that AI and the human-in-the-loop can bring to tax and legal research.
Thanks again, Giselle, and thanks to all our viewers for joining us for this insightful conversation. Stay connected as we go deeper into the realm of AI in our next episode. Thank you and have a good day.
If you would like to listen to our podcasts on the go:
Stay ahead of the digital tax curve with EY's Tax Technology and Transformation (TTT) network. We help businesses manage tax big data, comply with regulations, and make data-driven decisions.
EY DigiGST solution is one of the largest ASP- GST Suvidha Provider offering a secure cloud platform to meet all Goods and Service Tax (GST) compliance needs of businesses.
