Equipping Tax Service and Professionals with AI: The Promising Future

By Revanza Almaas, a Directorate General of Taxes officer

In 2023 everyone was suddenly talking about generative artificial intelligence. As we get into 2024, the intensity of discussions about generative AI is going to increase. In the US, law and accounting firms have been researching generative AI and its impact on the future of professional services. Several law firms have launched their own in-house generative AI solutions. Countless academic & professional conferences and events have recently hosted discussions on taxes and AI. Also, much of the popular media is busy thinking about the possibility of AI automating all jobs and further advances in AI risking the loss of control over our civilization.

The Restrospect on 2023

A year ago, in early 2023, ChatGPT appeared among the public. The public also issued various comments. Academics and professionals said that GPT (Generative Pretrained Transformer) has limitations. Some people liken this weakness to “stochastic parrots.” Stochastic parroting is an algorithmic process that “arbitrarily assembles sequences of linguistic forms… based on probabilistic information about the way linguistic forms are combined, but without understanding their meaning.”

This means that AI is good at imitating human language but cannot understand the meaning of the sentences it makes. This raises concerns about potential bias, lack of transparency, and the production of false or misleading information, particularly information about our tax system.

A revolving concern is that if these models are used uncritically, real social harm will result. From this perspective, the LLM (Large Language Model) is considered inadequate so that it arguably can’t be trusted to help real professional work. In the context of tax law, stochastic criticism is leveled against LLMs, that AI can produce tax advice that appears convincing but is legally erroneous or does not reflect current regulations. Of course, until this system is improved with some intention, for the government, before implementing it on the official websites, it is very important to carry out an evaluation of the LLM output for legal validity to reduce the possibility of giving incorrect advice.

At the same time in early 2023, a number of leading AI developers issued warnings about the challenges faced in ensuring AI safety, reflecting the uncontrolled progress of AI. Geoffrey Hinton, labelled as the godfather of AI, worried that the powerful capabilities and speed of AI in the future could pose risks to humanity. The worry he parables is that AI might find a way to kill people.

Of course we face confusion from these two opposing points of view. One believes that generative AI is just a stochastic parrot that doesn't understand what it's saying, while the other is worried about AI's excessive intelligence.

For me, there is a genuine reason to be optimistic about the benefits of generative AI for tax professionals. Of course, nevertheless, it is necessary for us to attend to the ethics of generative AI in tax practice. We need to appreciate the efforts being underway to test and ensure the security and ethics of AI in tax laws. Tax administrations in several countries are starting to try to adopt AI to provide better taxpayer services and compliance. For example, the OECD's 2023 tax administration report highlights the widespread use of AI virtual assistants by tax authorities to improve service quality.

Tax Systems and Generative AI

Startups and publishers, particularly in the US, are starting to focus on using generative AI to enhance their tax and legal research capabilities. Companies are starting to produce tools that enable tax professionals to leverage generative AI to analyze tax-related content. As new capabilities arrived throughout 2023, it will become increasingly clear that new applications of generative AI in the tax research process and taxpayer services will make both things faster and more effective.

People who have conducted experiments on AI responses to a range of tax law research questions highlight the importance of careful use of these LLM-based AI tools in tax research, exposing their promising potential in providing accurate and relevant responses, while emphasizing their limitations. However, they highlight the need for continued development and refinement of these AI tools to increase their reliability and usefulness in tax research.

Last year, I underestimated how quickly AI would develop. Now, I dare to predict the extraordinary things that AI systems (LLM) will be able to do in 2024. In the field of tax law, generative AI is expected to experience significant progress this year. The improvements seen in 2023 by companies like OpenAI and Gemini will likely continue in 2024, making the underlying AI models even more powerful. However, even with these advances, there are challenges that must be overcome before AI can be widely used in tax fields. Here are some of the main issues:

• Maintaining confidentiality of tax data;
• Ensuring tax professionals know how to use available AI tools; and
• Making sure AI works well with existing tax practices.

Another challenge is that AI may develop faster than we think, meaning we will have to continually adapt how we use it. There are also concerns that governments will not be able to keep up with these advances, creating a gap between what the private sector can do with AI and what governments can do with AI in the area of tax law.

Generative AI is increasingly becoming a necessity for tax experts, with the hope of leading to more efficient and effective tax research as well as helping taxpayers face their tax obstacles. However, AI will not be able to replace the nuanced insights and opinions that experienced professionals bring to their work. As we use these new tools, we must remain focused on the value they add, while recognizing the importance of keeping our work people-centered and ethically grounded.

Looking ahead, the chatters regarding the integration of generative AI in tax law are not just about the technology, but also about how we choose to adapt to changing possibilities and enhance our capabilities as professionals. The obvious challenge that arises is how to use these tools responsibly and continue to provide the insights and expertise they rely on. The future of tax work is here, and it's up to us to respond and shape it.

*) This article is the author's personal opinion and does not reflect the attitude of the agency where the author works.

The content on this page may be copied and reused for non-commercial purposes. However, we kindly request users to give credit to the source by providing a link back to the original page. Thank you for your cooperation.