US federal agencies reported 3,611 AI use cases across 41 agencies in the 2025 inventory. 445 were labelled high-impact, and Veterans Affairs (215) and Justice (114) account for 329 of them. Health and Human Services reported the most use cases (447), followed by NASA (425) and Veterans Affairs (367). 1,040 use cases were deployed, 440 were in pilot and 1,479 were in pre-deployment.
Hiring a Chief AI Officer or Head of AI Governance?
Get a free one-page role brief after a 30-minute call: scope, reporting line, a pay range from public filings, and where the candidates are.
Get my role brief →What does the table show?
Each US federal agency publishes an annual inventory of its AI use cases, and OMB consolidates them into one file. We counted the 2025 file by agency, using the inventory's own labels for impact and stage. Where an agency left a label blank, the count shows as Not published.
| Agency | Use cases | High-impact | Presumed, then not | Deployed |
|---|---|---|---|---|
| Health and Human Services | 447 | 2 | 13 | 167 |
| NASA | 425 | 1 | 3 | 26 |
| Veterans Affairs | 367 | 215 | 0 | 138 |
| Energy | 340 | 29 | 5 | 119 |
| Justice | 314 | 114 | 32 | 163 |
| Interior | 247 | 0 | 0 | 40 |
| Homeland Security | 238 | 55 | 49 | 110 |
| Commerce | 223 | Not published | Not published | Not published |
| Agriculture | 162 | 4 | 2 | 39 |
| Treasury | 129 | 4 | 3 | 26 |
| Transportation | 70 | 0 | 0 | 16 |
| State | 60 | 3 | 0 | 19 |
| Securities and Exchange Commission | 60 | 0 | 0 | 20 |
| Tennessee Valley Authority | 59 | Not published | Not published | Not published |
| Education | 56 | Not published | Not published | Not published |
| Federal Deposit Insurance Corporation | 50 | 1 | 0 | 16 |
| General Services Administration | 49 | Not published | Not published | 17 |
| Labor | 41 | 1 | 0 | 21 |
| Federal Reserve Board | 38 | 0 | 0 | 16 |
| Small Business Administration | 34 | 0 | 0 | 0 |
| Social Security Administration | 33 | 9 | 0 | 27 |
| Environmental Protection Agency | 29 | 2 | 1 | 7 |
| National Science Foundation | 17 | Not published | Not published | 2 |
| Pension Benefit Guaranty Corporation | 17 | Not published | Not published | 2 |
| Federal Housing Finance Agency | 16 | 0 | 0 | 12 |
| Federal Trade Commission | 16 | 0 | 0 | 11 |
| National Archives and Records Administration | 14 | 0 | 0 | 4 |
| Housing and Urban Development | 11 | 0 | 0 | 5 |
| Federal Communications Commission | 6 | 0 | 0 | 0 |
| Federal Energy Regulatory Commission | 6 | 0 | 0 | 1 |
| National Credit Union Administration | 6 | 5 | 0 | 6 |
| National Indian Gaming Commission | 5 | 0 | 0 | 5 |
| Election Assistance Commission | 4 | 0 | 0 | 1 |
| National Endowment for the Arts | 4 | 0 | 0 | 0 |
| National Transportation Safety Board | 4 | 0 | 0 | 0 |
| Nuclear Regulatory Commission | 4 | 0 | 0 | 2 |
| Commodity Futures Trading Commission | 3 | Not published | Not published | 1 |
| Farm Credit Administration | 3 | 0 | 0 | 0 |
| Surface Transportation Board | 2 | 0 | 2 | 0 |
| Occupational Safety and Health Review Commission | 1 | 0 | 0 | 1 |
| Office of Special Counsel | 1 | Not published | Not published | 0 |
| All 41 agencies | 3,611 | 445 | 110 | 1,040 |
Download: all 41 agencies with every impact and stage category (CSV). The Farm Credit Administration left 1 of its 3 impact labels blank; the CSV shows the split.
Which agencies report the most high-impact AI?
Veterans Affairs labelled 215 of its 367 use cases high-impact, and Justice 114 of 314. Homeland Security follows with 55, then Energy with 29. Together those four account for 413 of the 445 high-impact use cases. Health and Human Services reported the most use cases of any agency, 447, but labelled only 2 as high-impact.
What does "presumed high-impact, but not high-impact" mean?
OMB memorandum M-25-21, issued April 3, 2025, lists kinds of AI use that are presumed to be high-impact. Agency officials can rebut that presumption with written documentation to the agency's Chief AI Officer. The 110 use cases with this label are ones where that happened, 49 of them at Homeland Security and 32 at Justice.
Why this matters for AI leadership hiring
M-25-21 requires each covered agency to designate a Chief AI Officer and to publish its AI use case inventory every year, with extra risk management for high-impact AI. Those high-impact determinations run through the Chief AI Officer. The same work, deciding which AI uses carry real risk and documenting why, is what boards and regulators increasingly expect from AI governance leaders in companies and nonprofits.
What the count includes and leaves out
It counts the use cases in OMB's consolidated file of individually reported use cases. A separate file of commercial off-the-shelf AI products is not included. The Department of Defense does not appear in this file. M-25-21 does not apply to AI used as part of a national security system, so those uses are not here either. Labels are each agency's own; we did not recode or check them.
Frequently asked questions
What does "Presumed High-Impact, but Not High-impact" mean?
Under OMB memorandum M-25-21, some kinds of AI use are presumed to be high-impact. Agency officials can rebut that presumption with written documentation to the agency's Chief AI Officer. These are use cases where the presumption was rebutted. The 2025 inventory has 110 of them.
Why does the Department of Commerce show no impact or stage figures?
Its 223 rows in the 2025 inventory file carry no impact label and no development stage, so they show as Not published. Nine agencies reported no impact label for some or all of their use cases, 426 rows in total.
Methodology
Source. 2025 Federal Agency AI Use Case Inventory, file of individually reported use cases, from OMB's official GitHub repository, downloaded November 13, 2026. One row per use case; 3,611 rows.
Counting. Rows counted by agency, by the published impact field (High-impact; Presumed High-Impact, but Not High-impact; Not High-impact) and by the published development stage (Pre-deployment, Pilot, Deployed, Retired). Blanks are counted as Not published. Every agency's impact and stage counts add up to its total.
Refresh. Updated at this address when OMB publishes the next annual inventory.
Compiled by Husani Bastien. Tables were built with automated scripts from the official files listed below and checked by hand.
Sources
- OMB, 2025 Federal Agency AI Use Case Inventory, individually reported use cases (CSV) (US government data, public domain).
- OMB, M-25-21, Accelerating Federal Use of AI through Innovation, Governance, and Public Trust, April 3, 2025.
- Table and CSV on this page: Talent Echo Advisory Group, licensed CC BY 4.0. Please credit "Talent Echo, talentecho.net/data/federal-ai-use-cases/".