A recent GAO report found that an Internal Revenue Service (IRS) pilot initiative employing artificial intelligence to pinpoint audit cases and detect noncompliance lacked proper documentation for crucial aspects of its technology-driven sample selection models.

“IRS is piloting a new process for sampling tax returns for NRP’s audits. The new process uses artificial intelligence (AI) to improve the efficiency and selection of audit cases to help identify non compliance. However, IRS has not completed its documentation of several elements of its AI sample selection models, such as key components and technical specifications,” the report said.

Thorough documentation would aid the IRS in preserving institutional knowledge, guaranteeing consistent implementation of the models, and enhancing transparency for future users of the process, the IRS suggests.

Additionally, GAO also stressed in its report that documentation can facilitate transparency for agencies by delineating the system’s purpose, its limitations, the design process, and what it is not intended for.

Previously, IRS Commissioner Danny Werfel has revealed that the IRS has multiple AI projects such as building’ chatbots and other solutions’.

“Right now, I believe that there are AI solutions that we have not yet leveraged that exist today that can help with some of these basic questions to the benefit of taxpayers. And on the other side of the equation, we are using AI today to do even more to unlock and spot this complexity,”  Werfel said while speaking at UiPath on Tour: Public Sector Event in April.