According to a September 21 report, the Internal Revenue Service (IRS) takes about 20 months to process a case, and is still working through a backlog of identity theft cases left over from the pandemic. 

The report by the Treasury Inspector General for Tax Administration (TIGTA) found that case volume with the IRS’s Identity Theft Victim Assistance (IDTVA) unit nearly quadrupled in fiscal 2021, largely due to the pandemic. 

According to a FedScoop report, TIGTA says receipts remain “significantly elevated” even today due to the growing prevalence and sophistication of this kind of fraud.

In these cases, identity theft occurs when someone files a tax return using another person’s name and taxpayer identification number to claim a fraudulent refund.

The IRS started fiscal 2025 with 473,608 open IDTVA cases and ended the year with 315,740. TIGTA warned that the delay creates financial hardship for taxpayers and costs the government money through interest paid on delayed refunds.

What the Report Found

The IRS aims to resolve identity theft claims within 120 days of receipt. TIGTA reviewed a sample of 114 IDTVA cases closed in fiscal 2023 and found that only four met that goal. Average processing time across the reviewed cases was 655 days.

TIGTA estimated that the IRS paid approximately $124.2 million in interest on delayed refunds for identity theft cases closed between fiscal 2023 and fiscal 2025.

There was, however, a bright spot in the findings. Once a case was assigned to an assistor, the IRS moved quickly: 54% of taxpayer‑reported identity theft cases were resolved within 30 days of assignment, and a small share were closed the same day they were assigned.

That gap points to the real bottleneck: getting cases assigned in the first place, not resolving them once an assistor is working on them, according to the FedScoop report.

How the IRS Plans to Respond

TIGTA recommended that the IRS develop a process to assess the complexity of each case and use that assessment to speed up assignments. The IRS agreed with the recommendation.

Rather than pulling trained identity‑theft staff off casework to screen incoming cases, the agency said it will explore using artificial intelligence (AI) to help perform that evaluation. According to FedScoop, the IRS said it will determine next steps based on what that assessment finds.

What This Means for Data-Driven Case Management

This backlog is fundamentally a data and triage problem. Cases sit unassigned in part because the agency lacks a reliable way to sort incoming cases by complexity before a human reviews them.

That is where the IRS’s proposed use of AI becomes relevant. However, using AI to triage and route cases is likely to be effective only if the underlying case data is structured and consistent enough for a model to assess complexity accurately.

This is a core AI governance challenge: an AI system’s usefulness depends heavily on whether the data feeding it is labeled, current, and trustworthy.

The IRS’s own numbers show where the payoff is real: once a case reaches a human, resolution is fast. 

Any organization facing a similar backlog, where the bottleneck is intake and assignment rather than the work itself, should consider whether an AI‑assisted triage layer, built on properly governed data, could unlock similar gains without disrupting the part of the process that is already working.