Strengthening the 2025 Federal AI Use Case Inventory for increased government transparency

About OMB

The Office of Management and Budget (OMB) is responsible for advancing and implementing the president’s agenda across the executive branch. Among the agency’s many responsibilities, OMB is mandated by the Advancing American AI Act, Executive Order 13960, and OMB Memorandum M-25-21 to collect, report and publish an annual Federal AI Use Case Inventory.

Each year, OMB issues instructions to agencies then coordinates with them to collect, validate, and publish their reported AI use cases. This valuable data asset serves as the primary source for truth for agency Inspectors General investigations and Government Accountability Office investigations on AI use, as well as helps the public understand how AI technology is currently being used by the federal government.

The challenge

The Federal AI Use Case Inventory is a relatively new requirement, with year-to-year changes and refinements in guidance, and publishing it requires collaborating with every federal agency to gather data. Each agency collects information for their own AI inventory differently, and OMB has limited data collection tools and automated data validation capabilities. As a result, the published use case inventories in years prior to 2025 have historically had many invalid responses, duplicate rows, incorrect classifications, and inappropriately scoped submissions. Although OMB works hard to ensure response quality, it is infeasible for its small team with limited bandwidth to review thousands of submissions across more than 50 agencies and extract qualitative insights from the vast and varied set of submissions.

As a result, prior AI use case inventories’ limitations meant that federal technology leaders rarely consulted it when setting government-wide or agency-specific AI policies.

The approach

Seven USDC Data Science Fellows spearheaded all aspects of the 2025 inventory data review, processing, and consolidation, with guidance from stakeholders within the OMB Office of the Federal Chief Information Officer.

The Fellows started by analyzing the prior year’s inventory, noting shortcomings in the dataset, and presenting their findings to OMB. Based on this work, OMB stakeholders requested support for the 2025 data call, citing substantial government-wide goals. According to OMB, the AI use case inventory is “one of OMB’s most visible data projects” and “operationalizing the AI use case inventory is one of the Federal Chief Information Officer’s top priorities.”

The team of Fellows created a data validation and processing pipeline to automatically detect and fix data validation issues in agency submissions, including misclassifications and misformattings. The Fellows also leveraged LLMs in parallel to pre-screen every non-private response for intelligibility, jargon, and errors. After review, OMB sent the Fellows’ feedback back to agencies for recommended incorporation into final publication.

After agencies published their own AI use case inventories online, the Fellows created pipelines to download, extract, transform, and join each agency’s variously-formatted datasets together to centralize one government-wide consolidated inventory. These scripts generated programmatic metadata related to data cleaning and transformation, which were included alongside the consolidated dataset for public reproducibility and transparency.

Finally, the Fellows collaborated with OMB to analyze trends in the 2025 response set. They calculated descriptive statistics across agencies and reporting years, and utilized language models to identify topic clusters and perform qualitative analysis on the free response questions. Specifically, they helped identify trends of governmental AI use to better guide AI strategy at the federal level. For example, analyses showed that common AI use cases across agencies included managing FOIA requests, public comments, and federal grants management.

The impact

The Fellows consolidated and analyzed over 3,600 standardized use cases into the 2025 Federal AI Use Case Inventory. Their work strengthened the inventory’s role as a key resource for government-wide AI oversight, transparency, and decision-making.

Many federal agencies reported the Fellows’ work as highly accurate, helpful, and able to be incorporated into their own final publications. The 2025 Federal AI Use Case Inventory enables agencies to compare use cases and learn from peers across government, and serves as a foundation for scaling proven use cases.

The Fellows also collaborated with OMB on their lessons learned to codesign further improved reporting instructions for 2026. Public users of the inventory can now access more accurate information, which empowers researchers, civil society, and government technology leaders to better understand the federal government’s AI posture via a more reliable window into federal AI adoptions.

Most broadly, this work advanced government-wide transparency around a quickly-evolving technology. By recommending improvements to data quality and consolidating over 50 agency data sets into a common format, USDC Fellows made the inventory more usable and accessible for agency leaders, OMB, and the public.

The 2025 Federal Agency AI Use Case Inventory can be found on the OMB GitHub.