AI/ML in Civilian Agencies’ FY 2025 IT Budget Requests

Last month, I published an analysis of cloud computing in the FY 2025 IT budget. That analysis, which is available here, showed the requested budgets for programs that are leveraging cloud technology in one way or another. Today’s post provides data from the FY 2025 federal IT budget request showing what the budgets are for programs that are leveraging artificial intelligence and machine learning (AI/ML). As was the case for July’s post, the budget numbers provided are for civilian agencies only.

Total Program Budgets, FY 2020-2025

Here is the data for total program budgets. The data OMB published in March 2024 stretches back to FY 2020.

As we can see, growth in the budgets of programs leveraging AI/ML peaked in FY 2023. Those budgets and requests have since fallen in FY 2024 and 2025 despite the extraordinary amount of hype around AI/ML following the advent of ChatGPT.

The drop is attributable to a decline in investment at U.S. Customs and Border Protection (CBP) and the Internal Revenue Service (IRS). CBP received $33M in FY 2023 and FY 2024 for its Autonomous Surveillance Towers (AST) program. AST uses AI/ML to detect, identify, and track items of interest within a given field of view. For FY 2025, CBP requested $0 for AST because it has already purchased the equipment it needs for them.

In the IRS’ case, one investment labeled Modernized Operations received $203M in FY 2023, but it did not receive or request additional funds afterward. Meanwhile, the Platforms and Applications (P&A) program received $182M in FY 2024 and requested $192M for FY 2025. P&A creates, delivers, and deploys new enterprise applications including AI/ML capabilities.

Top Ten Civilian Agencies by AI/ML-Related Program Budgets, FY 2023-2025

Ranking AI/ML-related program budgets by agency, we can see that much like with cloud computing, the Treasury, meaning the IRS, also leads investment in AI/ML by a wide margin.

The IRS’ leadership is centered in the P&A program mentioned above. It is deploying robotics process automation to handle mundane back-office tasks.

Top Ten Civilian Agencies by AI/ML-Related Program Budgets, FY 2025 Only

Narrowing down the requested budgets to FY 2025 yields the following.

Readers will notice that DHS is not represented in this view of the data. This is because of the end of the AST investment discussed above.

Final Thoughts

Wrapping up, this analysis sheds light on the gap between the reality of AI/ML use and the hype surrounding it. From reading media reports on the government’s interest in the technology one would get the impression that investment in it is going full bore. The budget data shows, however, that this is not the case. Investments in AI/ML are happening, but they are often popping up in one year’s request before disappearing in the year afterward. This tells me that many AI/ML-related efforts are probably pilot programs.

The requested budget totals are also low compared to a more mature technology area such as cloud computing. To recount from last months’ post, the FY 2025 request for programs leveraging cloud is $8.9B. For AI/ML, the total is $405M. Notice as well that some AI/ML-related agency requests, at the EPA and DOE for example, are in the low single digits. These data points suggest that many agencies are still only dipping their toes in the AI/ML waters. Wholesale investment on the scale of cloud computing or cybersecurity is still a few years away.