Chatbots Were Supposed To Help Taxpayers But IRS Can’t Prove They Work
IRS chatbots were supposed to reduce phone pressures. But the Treasury Inspector General for Tax Administration, or TIGTA, found the IRS cannot prove they work.
The IRS built its chatbot and live-chat programs to do something taxpayers have wanted for years: make it easier to resolve routine tax problems without having to sit on hold. But a recent TIGTA report says the agency expanded those tools before it had reliable data showing whether they were working, whether taxpayers were getting answers, or whether live chat was actually reducing pressure on IRS phone lines. The result is that taxpayers who came to the IRS website looking for help may have left needing to contact the IRS again.
The IRS first piloted live chat in November 2017, expanded authenticated live chat in June 2019 (made permanent in June 2025), and added an automated ACS chatbot, available in English and Spanish, to offer preprogrammed responses for taxpayers in December 2021. (ACS stands for Automated Collection System, the IRS function that handles many inquiries involving balances due and delinquent returns.)
In January 2022, the IRS deployed an ACS unauthenticated voice bot that answers common questions (the authenticated voice bot was added in June 2022 and allowed taxpayers to create and manage installment agreements and make payments through self-service applications).Authentication is important because, after identity verification, an IRS employee can answer account-specific questions and take actions such as helping taxpayers establish payment plans or obtain account information. Unauthenticated chat or bots, by contrast, are limited to general tax information.
The Taxpayer First Act, enacted in 2019, required the IRS to modernize, and that necessarily included expanded online services. Among other reforms, the law required the IRS to develop a strategy that included expanded online services, telephone callback options, and improved employee training. That was followed by the Inflation Reduction Act and its modernization plan, which also emphasized a reimagined taxpayer experience. The goal was to push routine questions away from the toll-free line and give taxpayers an easier, faster path to answers. And early signs were promising (you can read about my experience in 2023 here).
As with any tech rollout, it’s important to measure the tool’s success. But the IRS didn’t analyze its chat applications or have meaningful performance measures. It did, however, keep reports. And when TIGTA reviewed ACS chat statistical reports from February 2023 through December 2024, it found outliers and discrepancies serious enough to call the reports’ reliability into question. How serious? One metric showed a live assistor handling as many as 603 live chats simultaneously, a figure the IRS agreed was wrong because the system was supposed to cap concurrent chats at three.
One of the issues may be a faulty handle time calculation. Handle time measures how long it takes a live assistor to service one chat. IRS IT officials told TIGTA the metric was being miscalculated,........
