The Future-Ready Tax Function: Harness the Power of Generative AI

Welcome to the latest issue of Talking Tax, where each month I explore a major opportunity, trend, or topic in corporate taxation. Be sure to subscribe to Talking Tax for my latest views on tax strategy and tax policy.

Have a question about your business’s tax strategy? Get in touch with me at Mkbecker@bdo.com. Now, let’s Talk Tax.

Since OpenAI co-founder Greg Brockman demoed GPT-4 in March by answering questions about an individual’s tax return, tax professionals have been trying to determine how best to leverage generative artificial intelligence (AI) technology. Many tax departments are now considering buying or co-developing a secure and responsible generative AI specific to their industry needs, like an in-house GPT, to automate tasks and access efficiencies.

In that context, generative AI would consist of a chatbot interface using machine learning (ML) via a large language model (LLM) to perform natural language processing (NLP) tasks. AI models of this kind have existed for some time, but their availability and sophistication have expanded rapidly in recent months and paved the way for a range of useful business applications.

Just a few examples of practical use cases for generative AI include tax risk management (e.g., to identify errors, non-compliance, or fraud), tax planning and analytics (e.g., to identify patterns and trends or highlight opportunities for savings), and process improvement (e.g., to reduce repetitive tasks like data entry or reformat data for enhanced processing by generative AI as an in-house knowledge base).

There are four key ingredients to developing and implementing generative AI effectively in the tax department:

• The first step is rigorous organization of the data necessary for training and using an in-house AI model.
• Second, you need talented tax professionals to act as data stewards who can harness the AI tools and leverage useful insights from data.
• Third, you need the malleable systems (e.g., cloud, APIs, transformation tools, etc.) to collect and process relevant data from across the organization. 
• Fourth, to protect your employees, your data, and society at large, governance needs to extend beyond functional applicability to include responsibility and security, in addition to collaboration with IT to operate and optimize systems and integrations securely.

Good Outputs Require Good Data

Generative AI needs a strong data foundation to work well, which entails robust data governance and master data management to help ensure data hygiene and synthesize usable data sets. It’s important to bear in mind that there are complex data science requirements to address before a business can access value with generative AI in the tax function.

Tax departments are inundated with data from various sources, which can require many hours of repetitive, manual work to sort and cleanse for analysis. Those time-consuming tasks detract from more strategic work because tax professionals must prioritize baseline activities like ensuring compliance.

Generative AI can significantly reduce that manual work and free up internal resources while also distilling insights for strategic decision-making — once the model has been trained, tested, and closely monitored for accuracy on a maturity curve. AI can also draw on broader and deeper pools of data to uncover insights that might otherwise have gone unnoticed.

The Value of Talent and Data Stewardship

Using generative AI in the tax department can reduce manual tasks, but it still requires experienced, knowledgeable professionals to review the findings and reap the benefits of the model. Strong data stewardship is critical for an AI-enabled tax function. That should include the following:

• Developing documented policies and procedures for data proportionality, adequacy, minimization, accuracy, completeness, security, accessibility, and more.
• Routinely assessing the model for refinements and enhancements, including automation of any repeatable processes, such as collection, maintenance, and synthesis of data, as well as ongoing training and updating of reference materials.
• Conducting recurring audits or in-depth analyses of data sets and outputs from the AI model to help mitigate potential bias and other risks.

Enablement and adoption are also vital to the success of an in-house GPT system for tax. Tax professionals and talent across the workforce will require training in prompt engineering and optimal ways of using the model to get the information they need. Leveraging advanced technology and training staff in how to use it can also be a tool for recruitment and retention by reducing repetitive, manual tasks and developing skills in high-level, strategic functions.

Working in conjunction with IT professionals and data scientists, tax practitioners can help ensure firmwide data stewardship that supports a strategic, technology-enabled approach to the tax function and unlocks value across the organization.

The Importance of Malleable Systems

Technology leaders for the organization should conduct an assessment of the business’s AI readiness to determine overall digital maturity and any systems that may require upgrading or modernization. The broader AI strategy should reflect key business goals and support a strategic approach to decision-making that includes the tax department. With nimble, adaptable systems, organizations can access the full range of benefits from generative AI and foster a culture of continuous innovation.

Developers can also build features that help staff understand how the AI model uses the underlying data to generate responses. A “show-your-work” function can demonstrate how the model arrived at a specific answer for each prompt, which can help tax professionals analyze processes for accuracy. The model can also be programmed to respond when it does not know an answer to a prompt instead of attempting to synthesize an answer that may include errors.

Ensuring Responsible and Secure AI

There are many governance considerations to assess as part of implementing responsible, secure AI models. It's critical to prioritize ethical AI practices that help guard against potential bias, including a rigorous process of algorithm selection and model training. It’s also important to maintain a focus on security to protect AI models against cyber threats and vulnerabilities along with continuous monitoring.

Addressing these objectives requires collaboration with IT to plan, implement, operate, and optimize systems and integrations. IT professionals and experienced advisors can also help define policies and identify tools that help mitigate bias and support responsible AI. As the regulatory environment for AI continues to evolve rapidly across multiple jurisdictions, organizations should take a proactive approach to compliance by staying up to date on new guidelines, best practices, and proposed regulations as well.

If further steps are needed to address internal systems maturity, an organization can still harness the benefits of generative AI through outsourced AI-as-a-Service (AIaaS) options, such as BDO Digital Genesis. But in the longer term, it is prudent to invest in building an in-house AI model to yield actionable insights, better inform strategic decision-making across the firm, and help remain competitive.

The Future-Ready Tax Function

Rather than replacing tax practitioners, the information-processing power of generative AI can supplement tax professionals’ experience and critical thinking to supercharge the tax function. At a time when most businesses are evaluating how to leverage efficiencies and insights from AI, it’s critical for tax professionals to seek ways that AI can help elevate tax departments — especially those that are more tactical and compliance-oriented — to become more strategic by driving innovation and sharpening the competitive edge. Otherwise, the business will be more susceptible to disruption and may be hampered in pursuing strategic objectives.

Generative AI is here to stay. By focusing on how to enable and enhance AI models through good data, talent, and systems, organizations can empower tax strategists and realize impactful benefits for the business.

Tapping an experienced third party can help assess your options and take a strategic approach to AI and the tax function. Contact us to discuss how your tax practice can gain better insights with generative AI.