Generative AI in tax
August 2024 | TALKINGPOINT | CORPORATE TAX
Financier Worldwide Magazine
August 2024 Issue
FW discusses the use of generative artificial intelligence in tax with Jennifer Deutsch, Richard Littleton, Jeffrey York and Jason Sawyer at Deloitte.
FW: How would you characterise the potential impact of generative artificial intelligence (GenAI) on corporate tax departments? Are you seeing widespread adoption of GenAI for this purpose?
Deutsch: Generative artificial intelligence (GenAI) possesses significant potential to spark a transformative shift in the operations of corporate tax departments by enhancing automation, data accuracy and data analysis. The technology may offer benefits in tax policy research and data-related tasks, serving as a personal assistant to tax professionals and changing their interaction with data. GenAI is set to transform traditional tax work and become an essential tool for tax professionals, driving a significant shift in the overall function of corporate tax departments.
Littleton: GenAI is here and is only going to continue to expand its applicability across organisations. From a tax perspective, GenAI is going to transform the way tax professionals traditionally do their work and challenge the status quo. For example, with GenAI, tax departments are going to have the capability to analyse large data sets in their entirety. As organisations refine their enterprise GenAI strategies, integration of financial data with GenAI models will change how professionals interact with data. The transformative potential of GenAI finds resonance in diverse tax department functions – tax planning, forecasting, tax credits, withholding tax, tax controversy, sales and use tax, direct tax, tax classification and adjustment automation. It is crucial for tax professionals to stay informed and adapt swiftly. The GenAI market is changing rapidly and will continue to evolve and serve as an enabling technology, providing solutions to business issues and fostering efficiency.
“The benefits of GenAI can only be fully realised with careful planning, the right infrastructure, and skilled human resources to manage and interpret its outputs.”
FW: Could you outline some of the typical applications of GenAI within the tax function? What aspects are corporate tax departments focusing on?
Littleton: We are standing at the dawn of GenAI adoption. In the realm of tax, initial adoption has been driven by chatbot-type solutions that are primarily geared to natural language processing, both on the input and output sides. As tax professionals interact with these, they are realising the value and requesting expanded features to interact with their data. The trend is shifting from ‘tell me’ to ‘show me’, with increasing efforts to build small proofs of concepts or minimal viable solutions for business case and return on investment validation. The next stage is autonomous automation where AI solutions will execute key functions in near- or real-time. This will transform tax professionals from doers to reviewers, enabling a new perspective on data analysis. While considering typical GenAI applications, it is vital to understand that tax technology is an ecosystem of integrated applications rather than a single construct. GenAI solutions will comprise various components like centralised data sources, data integration, user interface, output formats, workflow and connected processes.
York: It is helpful to divide currently available GenAI functionality into two categories. The first is general functionality, designed to streamline routine tasks like email drafting or document summarisation. While these tasks may not always be tax-specific and may need professional review, GenAI can help start these processes, eliminating the burden of starting from scratch. The second category is special-purpose use, which requires additional design for specific needs. This functionality involves using plain language to query data sets or documents, thereby enabling users to find required data more efficiently. In a tax context, a professional can use GenAI to swiftly locate data on a tax return. While corporate tax departments are mostly focused on the second category, both functionalities can significantly improve efficiencies and add value to the tasks of tax professionals.
FW: In what ways can GenAI transform tax processes? What benefits does it offer?
York: Corporate tax functions are often exploring technologies and processes that can optimise their operations and enhance their work quality. GenAI is an instrumental tool in this quest, leveraging the power of its underlying data models, such as large language models (LLMs), to build on previous technologies and accomplish goals that were previously challenging. For example, pairing GenAI with optical character recognition technologies has the potential to enhance data extraction accuracy from scanned documents, while reducing both the initial and ongoing costs.
Sawyer: Risk mitigation is another transformative application, with GenAI being capable of pinpointing potential risks and red flags in tax data for proactive risk management. It further extends support in maintaining regulatory compliance amid ever-evolving tax laws and regulations and can generate necessary audit reports and documentation. However, it should be noted that the benefits of GenAI can only be fully realised with careful planning, the right infrastructure, and skilled human resources to manage and interpret its outputs. The ultimate result is a streamlined tax process that enhances strategic tax planning and improves business outcomes.
“Companies should start small, set realistic expectations and focus on actual pain points within the tax function, such as time-consuming processes.”
FW: On the flipside, what limitations and risks need to be considered when using GenAI for tax purposes? What are the potential implications of ‘hallucinations’, for example?
Sawyer: While GenAI holds promising potential for enhancing tax processes, it is important to acknowledge its associated limitations and risks. Key among these are concerns around data privacy and security, given that GenAI systems process large volumes of data, often sensitive in nature. The effectiveness of GenAI is also closely tied to the quality, accuracy and completeness of the data supplied. Therefore, any inaccuracies or inadequacies in data can lead to erroneous analysis and predictions. Regulatory compliance presents another challenge, with the dynamic nature of tax laws and regulations necessitating frequent updates to GenAI systems. Furthermore, the implementation and maintenance of GenAI demand substantial technical expertise and infrastructure, which may pose technical challenges.
Littleton: Governance, risk and controls are all valid considerations when implementing GenAI solutions. With GenAI being a relatively new technology, there are a lot of unknowns and regulations that are rolling out globally and at varying paces. These are all important considerations and play a big part in executing ethical AI strategies and ensuring tax GenAI solutions get designed to mitigate these concerns. To help these concerns, we are seeing more activity in the market on small LLMs that are built for purpose.
“From a tax perspective, GenAI is going to transform the way tax professionals traditionally do their work and challenge the status quo.”
FW: To what extent do you believe GenAI can dispense with human involvement, and automate every task performed by tax professionals?
Deutsch: While GenAI can automate many routine and repetitive tasks currently performed by tax professionals, it will not eliminate the need for human involvement. Critical tasks such as decision making, strategy planning and complex problem solving still require human skills. Additionally, human oversight is necessary to ensure that the AI system is functioning correctly and ethically. Finally, human interaction is crucial for tasks that require a personal touch, such as client relations and interaction with tax authorities.
York: The prospect of a future where GenAI can efficiently collect and prepare data and workpapers, fill out and submit tax returns, and even manage and reply to tax notices is thought-provoking. However, GenAI tools do not apply human judgement to any process the way we traditionally think about it. They merely process tasks within the framework and limitations of the model on which they are trained. Even if models become well-tuned and highly accurate, high-risk or high-value activities will continue to warrant human involvement. In addition, while it is tempting to look only at the areas where GenAI might reduce or remove the need for human involvement, there are plenty of areas where human judgement and decision making seem nearly impossible to replace. For example, would you prefer to debate tax law with AI during an audit or seek a human for resolution? While we will likely see more GenAI processes in tax linked together over time, there are clearly some activities that benefit most from human interaction.
FW: What advice would you offer to companies looking to incorporate GenAI into their organisation’s tax function? What steps do they need to take to ensure the process is as seamless as possible and delivers the anticipated efficiencies?
Deutsch: Companies should focus on key themes to effectively use GenAI in their organisation. First, deepening their understanding of GenAI and the various vendor and specialised solutions available. Second, using readily available productivity tools to improve efficiency, particularly in areas ripe for quick gains. Third, collaborating across functions to identify enterprise solutions that can integrate with their tax functions. They should formulate a strategy that not only considers unique tax requirements but also explores how GenAI can meet these needs. This should involve identifying opportunities for automation and pinpointing areas where GenAI can deliver added value. Successful execution requires appropriate infrastructure, including data storage, processing capabilities and the right AI software. Data quality and security are paramount due to GenAI’s reliance on data. Companies need to ensure their data is accurate, comprehensive and clean. Finally, they should establish a robust governance structure to monitor GenAI’s performance and manage potential risks. This approach will help promote effective and secure use of GenAI.
York: Implementing GenAI, like any new technology, requires a measured approach. Companies should start small, set realistic expectations and focus on actual pain points within the tax function, such as time-consuming processes. Rather than solely seeking problems to solve with GenAI, it is important to consider its general non-tax uses, like summarising meeting notes. Engaging with GenAI in everyday tasks helps users understand the technology’s strengths and weaknesses, aiding in change management. Companies should not overlook potential transformation in areas like data collection and analysis. Like any new process, it is vital to identify potential risks of unexpected results and build in review mechanisms while assessing risk. Finally, collaboration with corporate leadership is needed to define measurable outcomes from the onset, ensuring project success is clear and quantifiable.
“While GenAI can automate many routine and repetitive tasks currently performed by tax professionals, it will not eliminate the need for human involvement.”
FW: Looking ahead, do you expect GenAI to become integral to the tax function? To what extent is it a game changer for corporate tax planning and productivity?
Sawyer: As we look toward the future, GenAI is poised to become an essential part of tax functions due to the increasing data volumes, escalating complexity of tax regulations, and the need for precision and efficiency in tax processes. GenAI is expected to become a crucial component of tax functions through enhanced efficiencies and productivity by automating mundane tasks, freeing tax professionals to focus on higher value activities. It can significantly improve the accuracy of calculations and data processing, as well as reducing errors – a crucial benefit in tax functions where minor mistakes can have major repercussions. GenAI’s advanced analytics can support strategic tax planning by processing large data sets to generate insights, identify trends and aid data-driven decision making. It can also help ensure compliance and facilitate risk management by monitoring ever changing tax laws and regulations, and identifying potential risks in tax data. However, GenAI is a tool designed to assist, not replace, human professionals. The nuanced understanding, professional judgement and personal touch of human professionals are still vital in the tax function. Thus, an optimal blend of GenAI and human expertise may yield the best results in terms of efficiency, accuracy, strategic planning and compliance.
Littleton: As digital transformation trends continue, it is increasingly vital to integrate solutions like GenAI into the tax function. GenAI can automate routine tasks, support data management, and elevate strategic planning by enhancing productivity. Furthermore, GenAI’s potential to analyse voluminous datasets can provide actionable insights for tax leaders, aiding in effective planning. Such insights enable businesses to spot risks, identify opportunities and make data-driven decisions. However, the real game-changer is how effectively businesses integrate GenAI into their operations and align it with their strategic goals. Therefore, companies should devise a clear integration plan, understand their strategic objectives and ensure alignment to realise the full potential of GenAI in their tax department.
Jennifer Deutsch is a partner in Deloitte Tax LLP’s tax technology consulting practice, focused on the operations of tax and specifically on tax processes and technology. With over 26 years of experience in serving corporate tax clients, she leads teams serving clients as they undergo transformations in tax. Her data and technology experience includes consulting on overall data management approaches, tax use and sensitisation of ERP and consolidation systems, as well as implementation and integration of standalone tax applications. She can be contacted on +1 (213) 364 4829 or by email: jdeutsch@deloitte.com.
Richard Littleton has been part of Deloitte’s tax technology consulting practice since 2010 and leads the tax automation and data management market offering. The practice is focused on working with tax organisations to drive automation and process efficiencies in the tax function through technology enablement. Areas of focus include source system tax sensitisation, tax data management, tax provision and compliance implementation services reporting, and analytics and advanced technology deployments such as GenAI and machine learning. He can be contacted on +1 (214) 840 1781 or by email: rilittleton@deloitte.com.
Jeffrey York is a partner at Deloitte Tax LLP, focused on multistate data technology and tax transformation. He has over 22 years of experience working with clients to better utilise data, technology, analytics, process improvement and automation to help analyse risk and identify opportunities. His experience crosses a variety of tax types, including income and franchise tax, sales and use tax, property tax, excise tax, unclaimed property and credits, and incentives. He can be contacted on +1 (612) 692 7008 or by email: jeyork@deloitte.com.
Jason Sawyer has over 25 years of experience in technology design and development, specialising in digital transformation. He is the chief digital officer of Deloitte’s US tax practice where he leads strategic technology development projects which provide solutions for collaboration between clients and Deloitte teams. He also serves as the global tax GenAI transformation leader. He is responsible for developing strategic direction, development and delivery of AI solutions to global member firms and clients. He can be contacted on +1 (518) 320 0090 or by email: jasonsawyer@deloitte.com.
© Financier Worldwide