IRS Expands Compliance Risk Models With Economic Data

The IRS is accelerating its use of compliance risk models powered by artificial intelligence and machine learning to sharpen tax enforcement, according to the agency's Strategic Operating Plan. The shift signals a new era of data-driven audit selection that leans heavily on economic benchmarks and anomaly detection.

How the IRS Is Upgrading Audit Selection

For decades, the IRS relied on the Discriminant Function System, or DIF, and the Unreported Income DIF to score returns for audit potential. Those tools flag filings based on historical patterns of noncompliance and possible unreported income.

Now the agency is layering modern tax compliance analytics on top of that foundation. The plan includes a centralized anomaly-detection platform, stronger data integration, and supervised learning techniques. An IRS-hosted research paper found that combining supervised learning with anomaly scores increased audit-detected underreporting by an average of 8% per audit across business expense categories.

Economic Benchmarks Enter the Equation

The updated compliance risk scoring approach draws on a wider pool of external data than ever before. Regional economic accounts from the Bureau of Economic Analysis supply state and local GDP and personal income figures. The Bureau of Labor Statistics publishes wage data spanning occupations, regions, and roughly 400 industries. Census Bureau business statistics add establishment counts, employment, and payroll, sorted by geography and enterprise size.

Federal Reserve indicators such as industrial production and capacity utilization round out the macroeconomic picture. FinCEN's suspicious activity report statistics contribute to filing-trend data by industry and state. Together, these industry benchmarks provide enforcement analysts with a baseline for assessing whether reported income, deductions, or margins align with broader economic reality.

Complex Partnerships and Large Entities Draw Focus

The IRS has signaled that complex partnerships are a primary target of its upgraded analytics. The Strategic Operating Plan states that the agency is already using AI and advanced analytics to help determine which large and complex partnerships to examine. Treasury's Office of Tax Analysis supports that effort by maintaining large statistical databases and developing simulation models that assist with studies of tax compliance and compliance burden.

This focus reflects a broader enforcement trend. Multi-tiered partnership structures, pass-through entities, and high-wealth arrangements often involve layers of transactions that traditional scoring systems struggle to evaluate. Machine learning models trained on third-party and corporate-registry data can surface patterns across related entities that manual review would miss.

What This Means for Taxpayers and Practitioners

Returns that diverge sharply from regional or industry benchmarks may draw scrutiny even when the underlying position is correct. That risk is highest in volatile sectors affected by inflation, supply chain disruptions, or labor shortages, where legitimate margin swings can mimic noncompliance patterns.

The Government Accountability Office has cautioned that better model evaluation is essential so the IRS does not burden compliant filers while missing serious noncompliance. For tax professionals, the takeaway is clear: contemporaneous documentation, including contracts, payroll records, and evidence of local market conditions, is now as important as the return itself.

Documentation as a First Line of Defense 

Practitioners should also monitor enforcement priorities through the annual IRS Data Book, which reports examination activity and third-party data usage. Staying ahead of compliance risk scoring trends means understanding which economic indicators the IRS treats as baselines and how those baselines shift with macroeconomic conditions.

In a landscape shaped by AI-driven compliance risk models, thorough records do more than defend a position—they explain why an atypical return may still be accurate. Firms that proactively benchmark their clients' filings against publicly available economic data will be better positioned to respond if the IRS flags a return for further review.

Sources

• IRS Strategic Operating Plan
• IRS Audit Process Guidance
• IRS Data Book

‍By William Mc Lee, Editor-in-Chief & Tax Expert—Get Tax Relief Now