Skip to main content

AI Use in Tax Administration

Governments are leveraging artificial intelligence (AI) tools to streamline operations and transform public services delivery; opportunities to use them to improve tax administration are no exception. As part of its 2023 Strategic Operating Plan, the IRS identified AI models and data management as a priority to enhance operations and innovation ahead of 2025. As AI is increasingly integrated into tax administration, policymakers should establish guardrails to ensure secure and appropriate data use. This blog builds off BPC’s latest research examining efforts to improve tax administration through technology and data sharing.

Read Next

Only Scratching the Surface

Today, 65% of global tax administration authorities acknowledge AI’s use and integration in their daily operations. The IRS has successfully used AI to automate internal processes, improve taxpayer services, and detect potential fraudulent returns through various applications, including computer vision, natural language processing, and predictive analytics for fraud and verifications.

One of the most common applications of AI in tax administration is automating data retrieval and reducing manual entry of information. For instance, using the IRS’ Modernized e-File (MeF) system, optical charter recognition extracts relevant data from paper returns to upload into their databases, reducing the manual input required and the resulting backlogs. Experts see this AI use case as having the greatest potential to improve efficiency by eliminating many of the most time-consuming elements of the tax compliance process.

Once an electronic return is filed, MeF validates the information before it’s transmitted to the IRS. If a return fails this validation process, the system identifies where the error occurred so that taxpayers can better pinpoint the location of their error and, where possible, input information quickly to rectify the problem. Additionally, MeF offers an integrated payment option for taxpayers to directly authorize the withdrawal of funds from their bank accounts. To supplement these systems, taxpayers can also chat with the IRS’ customer service bots to ask questions and understand changes to tax law, which has reduced calls to the IRS. AI tools can help speed up these transactions, automate processing of information, and enable more convenient customer communication methods.

Another primary use of AI by the IRS has been to identify representative samples of taxpayer returns for annual audits, supplementing the manual process in which enforcement officers select returns for audits due to noncompliance concerns. The Department of Treasury recovered $375 million in fiscal year 2023 by using AI to mitigate check fraud and strengthen processes to reclaim potentially fraudulent payments, demonstrating that when used appropriately it can improve the agency’s efficiency and effectiveness.

Potential Public Policy Pitfalls

Lawmakers and tech leaders alike, while observing the potential of AI to improve operations and efficiency, are wary of the dangers associated with its exponential growth absent clear protocols and protections for taxpayers. When it comes to risks in tax administration, three overarching issues emerge:

  • AI systems require adequate safeguards to prevent fraud and abuse. While AI can be effective in detecting fraud, there are also concerns it will be used by scammers to aid in tax fraud or scams. Recent instances demonstrate that scammers illegally use AI to impersonate taxpayers to steal personal data or intercept tax refunds. A bipartisan group of senators wrote to IRS Commissioner Danny Werfel in May 2023 with concerns about the role of AI in conducting sophisticated tax scams, requesting that the agency identify plans to intervene in cybercrime, educate the public, and prepare for the evolution of these scams in years to come.
  • Preserving taxpayer trust through bias reduction and privacy protections. Tax forms contain sensitive and confidential information, so ensuring they are secured when submitted is critical to preserve trust, safety, and fairness in the system. Some IRS auditing practices have inadvertently yet disproportionately impacted low-income filers, making it paramount that simplicity not outweigh impartiality in AI’s use case. The IRS should ensure that dedicated staff are involved at every step to detect unintended consequences when AI is deployed in the compliance process.
  • AI should be a complement to—not solely a replacement for—workers. AI’s rapid introduction into the workplace has introduced concerns over its potential to disrupt or displace workers. As AI is further integrated into IRS systems and processes, the agency can focus efforts on training and educating employees in line with the priorities outlined in their Strategic Operating Plan.

AI’s potential can be harnessed while safeguarding against risks, ensuring a balanced approach that enhances efficiency without compromising security, fairness, or employment.

Takeaways: AI and the IRS

As AI becomes even more integrated into tax administration, it is crucial to establish measures to ensure its responsible development and use that uphold ethical principles, combat bias, and prevent trust erosion. This could encompass defining technical standards and designing organizational sandboxes to pilot new processes similar to other public sector use cases.

The Biden administration’s October 2023 executive order establishes a whole-of-government approach to responsible AI, and several bipartisan efforts seek durable policy that can adapt to the changing nature of AI tools. In May 2024, the Bipartisan Senate AI Working Group released a list of policy priorities for lawmakers, including upskilling and retraining workers who are at risk for displacement, investing in infrastructure and research, and creating clear privacy safeguards and codes of conduct.

Policymakers have a window to create guidelines on AI deployment for tax administration—from improving taxpayer assistance to screening large amounts of data to detect improper payments—with clear policies on data protection. Coupled with sufficient staff training, such guidelines could help ensure that federal agencies like the IRS can not only efficiently leverage AI but continue to adequately secure data as government services modernize in the years to come.

Support Research Like This

With your support, BPC can continue to fund important research like this by combining the best ideas from both parties to promote health, security, and opportunity for all Americans.

Give Now