The IRS Could Leverage Examination Results in Artificial Intelligence Examination Case Selection Models and Improve Processes to Evaluate Performance

Date Issued

Monday, May 19, 2025

Submitting OIG

Treasury Inspector General for Tax Administration

Agencies Reviewed/Investigated

Internal Revenue Service

Report Number

2025-308-022

Report Type

Audit

Agency Wide

Yes

Number of Recommendations

Questioned Costs

Funds for Better Use

Report updated under NDAA 5274

Open Recommendations

This report has 3 open recommendations.

Recommendation Number	Significant Recommendation	Recommended Questioned Costs	Recommended Funds for Better Use
1	No	$0	$0
The Chief Tax Compliance Officer should require the SB/SE and LB&I Divisions’ Commissioners, in partnership with the Chief Data and Analytics Officer, where appropriate, to use governance processes to ensure that examination performance results are part of the monitoring and continuous refinement of AI classification (deployed) and return selection models (if deployed).
2	No	$0	$0
The Chief Tax Compliance Officer should require SB/SE and LB&I Divisions’ Commissioners, in partnership with the Chief Data and Analytics Officer, to refine AI models discussed in this report by incorporating ensemble machine-learning when appropriate.
3	No	$0	$0
The Chief Tax Compliance Officer should require SB/SE and LB&I Divisions’ Commissioners, in partnership with the Chief Data and Analytics Officer where appropriate, to establish a measurement plan with appropriate metrics to monitor the Form 1040 return classification and LAR and LPC models to ensure that they are achieving the expected benefits and correct any model drifts.