By now, many federal agencies have published their respective 2025 AI use case inventories, statutorily mandated by the Advancing American AI Act. A pair of April 2025 AI-based OMB memos outline the latest reporting requirements for agency AI use case inventories, “Each agency (except for the Department of Defense and the Intelligence Community) must inventory its AI use cases at least annually, submit the inventory to OMB, and post a public version on the agency's website…. agencies must document implementation of the minimum practices in Section 4(b) of this memorandum for high-impact uses of AI and be prepared to report them to OMB…”

Note that high-impact use cases refer to those instances where the AI output may have a legal, binding, or significant effect on an individual or entity’s rights, safety and/or security.

Below is an assessment of two health?focused federal agencies and how their use of artificial intelligence has evolved in both volume and variety over the past few years. Although additional health?oriented bureaus and agencies operate within the federal landscape, HHS and the VA often represent the leaders in federal healthcare.

*Previously referred to as rights-impacting and safety-impacting

Both health-centric agencies increased in AI use cases year over year since 2023, with the number of AI use cases at the VA growing 186% since 2023, and 174% at HHS.  This is understandable as the push for, and familiarity of AI adoption continues to grow in the federal government.  

Examples of high-impact cases at VA include integrating voice bots across VA call center lines, search and summarization of patient records for clinicians, and improving the ability to identify veterans at high-risk for suicide.

Interestingly, despite having a larger number of total use cases, HHS only identifies two AI programs in its 2025 inventory as high-impact. These include sponsor identity verification at ACF and automating the processing of remote identity proofing (RIDP) verification tasks at CMS. It is unclear whether this low number reflects HHS’s definition of high-impact use cases or a reluctance by the department to pursue risk?based AI pilot initiatives. A closer look at the HHS data set reveals that the department labels 13 additional cases as “Presumed high-impact but determined not high-impact.” These cases include assistance in a grant referral process at NIH, identifying waste, fraud and abuse at CMS, and accelerating identification of persons, vehicles and events at FDA.

Additional observations are below. Note that each department provides varying information within their respective AI use case inventory data set. Thus, there may be some reflections available for one agency and not the other.

Additional observations:

• HHS lists 49 use cases as “retired” in 2025, while VA lists 72 use cases as retired.
• VA lists 30 use cases as developed in-house, the rest are either purchased from a vendor or hybrid.
• HHS lists 81 use cases as developed in-house, with the rest identified as purchased from a vendor or hybrid.
• VHA holds 80% of VA’s 2025 high-impact AI use cases, with many centered on clinician assistance and predictive diagnosis.
• About 68% of 2025 HHS AI use cases are already deployed or in pilots, rather than in a conceptual phase, revealing a measure of AI maturity at the department.
• Nearly 40% of VA AI use cases are in the pre-deployment stage.
• Most HHS use cases cite “None of the Above” for demographic variables used.
• According to the 2025 HHS AI use case inventory, Deloitte and Palantir led the vendors listed within the department’s AI inventory with 19 and 15 use cases, respectively.

Observations:

• HHS high use of natural language processing (NLP) and generative AI technologies reflect the department’s typical handling of large volumes of regulatory documents, public comments, and scientific data.
• HHS signals a growing adoption of Agentic AI use cases over VA.
• Nearly 83% of VA computer vision use cases are deployed, signaling a maturing for the technology at the department.
• Most generative AI use cases at the VA cite a pre-deployment status, signaling the technology’s early-stage presence at the department.
• Most generative AI use cases at the VA cite a pre-deployment status, signaling the technology’s early-stage presence at the department.

Observations:

• Health & Medical dominate in use cases at both departments: VA use cases in this topic area center on ambient scribing/clinical summarization, while HHS use cases focus on clinical and biomedical summarization and identification.
• Administrative Functions use cases at both departments largely center on content generation and knowledge retrieval.
• Generative AI represents nearly 50% of VA use cases classified as Procurement & Financial Management, which include document analysis assistance, closeout automation, and compliance checks, among others.
• NIH and CDC represent the bulk of HHS Administration Functions use cases, while FDA represents nearly 50% of Health & Medical use cases.

Overall, the growing volume and diversity of AI use cases at both HHS and the VA illustrate the accelerating integration of artificial intelligence across the federal healthcare landscape. While each department’s approaches and maturity to AI differ, particularly in the identification of high-impact cases and the rate of deployment, the HHS and VA’s latest inventories reflect a growing commitment to AI adoption. Contractors can expect areas such as generative and agentic AI to grow among the health entities to expand areas of clinical and administrative workflows.

Sources:

• VA AI Use Case Inventory – Updated 7/29/2024
• VA AI Use Case Inventory – Updated 12/16/24
• 2025 VA AI Use Case Inventory
• 2025 HHS AI Use Case Inventory
• 2024 HHS AI Use Case Inventory
• 2023 HHS AI Use Case Inventory