US Agencies See 9x Rise in GenAI Use Cases in 2024, Yet Deployments Lag

Rising AI Use Cases in Federal Agencies

Over the past year, there has been a noticeable increase in the number of artificial intelligence (AI) use cases across U.S. federal government agencies. However, the implementation of these technologies remains slow due to several challenges. These include insufficient funding, limited access to computational resources, outdated policies, and a workforce that is still adapting to the rapid pace of AI development.

A recent report by the U.S. Government Accountability Office (GAO), mandated by Congress, examined the current state of AI deployments within federal agencies. The findings revealed a significant surge in AI usage. Out of 12 agencies reviewed, 11 reported an increase in AI applications, with the Department of Defense being an exception due to security restrictions.

According to the GAO, federal agencies documented 1,110 AI use cases in 2024, nearly doubling the 571 cases reported in the previous year. The growth was even more pronounced for generative AI, which saw a ninefold increase, from 32 use cases in 2023 to 282 in 2024.

Of the 282 generative AI use cases identified, approximately 56% were in the process of being initiated or were under acquisition or development. Around 41% were either in implementation or already operational, while the remaining cases had been retired.

The majority of AI projects focused on internal mission support, such as improving communication, streamlining processes, and enhancing operational efficiency. Public-facing services like chatbots and virtual assistants ranked second in popularity. Medical applications came in third, with the Department of Veterans Affairs using AI to automate medical imaging processing and the Health and Human Services Department employing AI to combat the re-emergence of polio.

Notably, NASA was the only agency reviewed in 2024 that reported no generative AI use cases, despite listing related activities in its 2023 inventory.

Common Challenges Faced by Federal Agencies

Despite their substantial budgets and high-level support, federal agencies face many of the same obstacles as private enterprises when it comes to implementing AI. Of the 12 agencies surveyed by the GAO, eight struggled to obtain the necessary compute resources, with seven citing a lack of funding as a primary issue. Six agencies also reported difficulties in finding or training a skilled AI workforce.

Another shared concern among agencies is the potential for AI tools to produce biased data or generate inaccurate information. Five agencies cited this as a reason for caution in adopting AI. Additionally, the Defense Department and NASA expressed concerns about the "black box" nature of most generative AI systems, highlighting the need for greater transparency.

Government agencies also face unique challenges due to their slow-moving bureaucratic structures. Many are hindered by layers of red tape and regulatory hurdles. Ten out of the 12 agencies noted that existing federal policies may not account for or could impede the adoption of generative AI. Cybersecurity regulations, in particular, pose significant challenges, as compliance can be time-consuming and complex.

The pace of AI innovation also presents a challenge, as government agencies often struggle to keep up with the rapid changes in technology. Several agencies have reported delays in acquiring desired AI products due to the lengthy FedRAMP approval process, which ensures compliance with government cloud security standards.

Steps Toward Improvement

While the GAO did not provide formal recommendations in this report, it highlighted that agencies are using existing frameworks, such as those developed by the National Institute of Standards and Technology, to guide their generative AI efforts.

Candice Wright, the GAO director for science, technology assessment, and analytics, emphasized the importance of policy and practice approaches for agencies looking to deploy generative AI systems. She also encouraged agencies to learn from successful teams to better plan for potential challenges.

Wright noted that the issues identified in the report will require ongoing attention as agencies continue to develop and implement AI solutions. She stressed that these challenges will not resolve themselves and must be actively addressed moving forward.