Pursuant to a congressional request, GAO reviewed the Internal Revenue Service's (IRS) Taxpayer Service Integrated System (TSIS) to determine whether tests to measure TSIS effectiveness provided the information necessary to justify installing TSIS at all call sites.

GAO found that: (1) tests of two of the three TSIS projects to automate its taxpayer inquiry program, the Taxpayer Service Expert Assistance System and the Automated Taxpayer Service System (ATSS), during the 1989 and 1990 tax filing seasons did not conclusively demonstrate whether the automated system could improve accuracy and productivity; (2) the Expert System could not answer about 62 percent of the questions taxpayers asked because it lacked adequate information and it took more time to answer simple questions than manual assistance; (3) the Integrated Test Call Survey System (ITCSS) could not adequately evaluate the Expert System, since ITCSS measurements did not consider whether assistors used the system properly, or at all, in responding to taxpayers; (4) IRS found that assistors using ATSS correctly answered almost 94 percent of taxpayers' questions, while those not using ATSS answered questions correctly about 90 percent of the time, but according to ITCSS, those not using ATSS gave correct answers 10 percent more often than those who used ATSS; (5) less experienced personnel had higher rates of productivity when they used ATSS, but more experienced staff members' productivity suffered; (6) despite inconclusive test results, IRS planned to install the Expert System at four more locations for the 1991 tax season; (7) 1992 TSIS test results will lack sufficient information to decide whether IRS should install it at more call sites; (8) the Office of Management and Budget denied the IRS fiscal year (FY) 1992 budget request for $41 million to expand TSIS, since IRS did not show the system to be cost-effective; and (9) IRS now plans to extend the development and installation of TSIS through FY 1997, rather than FY 1994, as planned.