GAO flags AI procurement gaps
A GAO review found agencies face persistent challenges when acquiring AI, including gaps in talent and a failure to systematically capture lessons learned from prior procurements. The report noted agencies increased AI use from 2023 to 2024 but still lack repeatable intake and review practices that create a usable acquisition record. (executivegov.com)
Federal agencies are buying more artificial intelligence, but the Government Accountability Office said many still do not keep a usable record of what worked and what failed. (gao.gov) The report, released April 13, 2026, said agencies more than doubled reported artificial intelligence use from 2023 to 2024 and used multiple buying models through fiscal year 2025. Government Accountability Office reviewers examined 13 acquisitions across the Department of Defense, Department of Homeland Security, General Services Administration, and Department of Veterans Affairs. (gao.gov) Artificial intelligence procurement can mean buying software outright, paying for a service that delivers outputs over time, or using agreements outside the standard Federal Acquisition Regulation process for more advanced work. Government Accountability Office officials said agencies also faced trade-offs between agency-directed buys and vendor-driven offers that arrived before agencies had written detailed requirements. (gao.gov) Officials told the Government Accountability Office they struggled to get technical reviewers such as data scientists into source-selection work and had trouble estimating artificial intelligence costs. The watchdog said selected agencies were not systematically collecting lessons learned, even after Office of Management and Budget guidance told agencies to share acquisition knowledge across government. (gao.gov) (whitehouse.gov) That gap comes as artificial intelligence use inside government is rising fast. In a separate July 29, 2025 review, the Government Accountability Office said reported artificial intelligence use cases across 11 agencies rose from 571 in 2023 to 1,110 in 2024, while generative artificial intelligence use cases jumped from 32 to 282. (gao.gov) The policy backdrop also shifted after that earlier procurement memo. The Office of Management and Budget’s memorandum page shows the Biden-era September 24, 2024 acquisition memo was followed by a new April 3, 2025 memo, “Driving Efficient Acquisition of Artificial Intelligence in Government,” alongside a same-day memo on accelerating federal artificial intelligence use. (whitehouse.gov 1) (whitehouse.gov 2) The Government Accountability Office’s recommendation was narrow and procedural. It said the General Services Administration should require officials to systematically collect lessons learned from artificial intelligence acquisitions, including best practices on contract clauses, and submit them to the General Services Administration-managed repository for governmentwide sharing. (gao.gov) The four agencies in the review agreed with the recommendations, according to the report’s appendixes. The finding leaves the government with a basic procurement problem: agencies are buying more artificial intelligence while still building the playbook after the contracts are signed. (gao.gov)
