IRS Weighs AI Triage for Identity Theft Backlog Nearing 257,000 Cases

The IRS is weighing artificial intelligence to triage a tax-related identity theft backlog that reached 256,988 unresolved cases as of July 25, 2026. The agency has not deployed an AI system for this caseload.

The Treasury Inspector General for Tax Administration, or TIGTA, found that taxpayers waited nearly 20 months on average for resolution, against a 120-day agency goal. The queue remains long. A separate figure puts the average time cases sit unassigned at 533 days.

That gap points to a bottleneck before an employee begins working a claim. TIGTA recommended faster complexity screening and assignment procedures; management agreed to assess whether AI could help evaluate cases and route them to staff.

The agency has not settled on a change. Its assessment will guide what it does next, while people with pending cases continue to face lengthy waits.

The longest delay comes before an employee takes the case

TIGTA’s findings distinguish the time cases spend waiting for assignment from the time an assistor spends handling them. The average unassigned wait was 533 days. That is nearly a year and a half before substantive casework begins.

The overall wait has been described in several ways across different reporting periods. TIGTA cited an average of 582 days as of June 2025, while current agency guidance gives 580 days. The broader average remains close to 20 months.

Once assigned, some cases move faster. The agency completed 54% of taxpayer-reported cases and 44% of cases it identified itself within 30 days after assignment. Those rates do not erase the preceding queue, where cases can remain untouched for months.

The proposed use of AI is limited to evaluation so far. The agency says it will consider whether a system could assess complexity and support routing, rather than using fully trained staff to screen every incoming file before assignment.

The inventory has shrunk, but it remains large

The number of open cases fell across three fiscal year-end counts reported by TIGTA. The reduction has not yet produced a short average resolution time.

The backlog grew during a pandemic-era rise in identity-related tax fraud. Reporting on the audit said the inventory had nearly quadrupled in 2021 before later declines. Even as the total fell, average resolution time remained about 20 months.

The latest count, 256,988 cases on July 25, 2026, is lower than the fiscal 2025 year-end figure. It still leaves a substantial number of taxpayers waiting for the agency to resolve claims involving misuse of their tax identities.

A watchdog’s sample found few cases met the target

TIGTA reviewed 114 closed cases and found they took an average of 655 days to resolve. Only 4 were completed within the 120-day target.

The sample shows how far actual case timelines can run beyond the stated goal. It also puts the aggregate wait in context: even cases that eventually close may spend far longer in the system than the target allows.

Those delays have a direct financial cost for the government. TIGTA calculated $124.2 million in refund interest tied to late resolutions during fiscal years 2023 through 2025.

Management has not committed to a new screening step

Kenneth Corbin, chief of the agency’s Taxpayer Services Division, acknowledged the need to move cases faster while questioning whether another manual screening stage would help with existing staffing.

“We agree that cases should move to resolution as quickly as possible.”

Corbin said the agency did not believe “adding a separate pre-assignment screening step would improve timeliness with current resources.” Assigning trained employees to screen cases before assignment could “duplicate work and reduce the number of employees actively closing cases.”

Instead, the agency said it would continue using case types, employee skill levels and other available tools to improve how work moves through the system. The AI review is an assessment, not a commitment to add a live automated step.

TIGTA said management agreed to evaluate case complexity and explore AI as an alternative to redirecting full-scope trained staff to screen cases. The agency’s next steps will depend on that assessment.

Commissioner Frank Bisignano has said the agency uses AI in some areas, but not for this backlog. National Taxpayer Advocate Erin Collins described the wait times as “unconscionable.” Neither statement changes the status of the proposed review: AI has not been put to work on these cases.

Delayed refunds have also carried an interest cost

The $124.2 million in refund interest accumulated across fiscal years 2023 through 2025 as cases took longer to resolve. TIGTA’s findings connect that cost to a queue that can hold cases for 533 days before assignment.

Management will decide what to do after assessing whether AI can help sort cases by complexity and improve routing. Until then, the agency’s existing case types, staff skill levels and other tools remain its stated approach.