HHS outlines strategy to expand AI adoption across the department
The Department of Health and Human Services released a plan to deploy and centralize artificial intelligence across the agency, with an initial focus on internal use. The strategy prioritizes shared AI resources, governance for new tools, and applications in public health. Collaboration with the private sector is planned, along with identifying priority conditions that AI could help address.
The move fits a broader federal push to cut costs and speed up technology adoption. Separate HHS divisions have already signaled AI use: the FDA plans to use agentic AI to support premarket reviews, and the CDC plans to analyze public health data with AI.
What HHS plans to do
- Centralize tools and data under "OneHHS": Create a shared inventory of AI tools and resources across divisions.
- Stand up governance: A senior leadership board will meet at least twice a year, streamline approvals, monitor tools, and set risk protocols for high-impact AI.
- Train the workforce: HHS will equip employees at all levels to use AI responsibly and effectively.
- Fund R&D: Invest in science programs that use AI in biomedical research and development.
- Support clinical decision support: Promote AI tools that augment care delivery and operations.
"For too long, our Department has been bogged down by bureaucracy and busy-work; even the most productive public servants are mired in paperwork and process. ... It's time to tear down those barriers," said Jim O'Neill, deputy secretary of HHS.
Targets that matter to providers
HHS highlighted outcomes as the scoreboard for AI adoption. Examples include fewer hospital readmissions, lower sepsis mortality, fewer unnecessary ED revisits, and improved infant and maternal health outcomes. The goal is measurable gains, not shiny tools.
Governance and risk
HHS will begin with a baseline assessment of its current AI tools to evaluate maturity. A centralized database and oversight board will guide approvals, monitoring, and risk management, including specific protocols for high-impact AI.
This is needed. AI can introduce inaccurate outputs and bias, which carry clinical and operational consequences. A common governance model and shared playbooks can reduce duplicated effort and uneven standards across divisions.
Context inside and outside HHS
The strategy follows an executive order from President Donald Trump and OMB direction to speed agency AI adoption. Within HHS, individual groups are already moving: the FDA is exploring AI to support reviews, and the CDC plans broader use of AI for public health analytics.
HHS is also restructuring-laying off 10,000 workers and consolidating divisions from 28 to 15-which makes standardization and shared services more pressing.
What healthcare leaders can do now
- Prioritize high-value use cases: Start with readmissions, sepsis detection, ED flow, care coordination, and maternal health outcomes that align to HHS priorities.
- Tighten data quality and access: Map data sources, close gaps, and set versioned datasets for model evaluation and post-deployment monitoring.
- Adopt a simple risk tiering model: Classify AI by impact on patient safety and operations; require stronger validation and human oversight for higher tiers.
- Build a lightweight AI review path: Define evidence thresholds, bias checks, and clinical sign-off before go-live. Instrument every tool for drift and performance tracking.
- Upskill teams: Train clinicians, analysts, and IT on prompt use, validation basics, and safe workflows; document standard operating procedures.
- Align contracts to governance: Bake in data rights, auditability, model update cadence, incident response, and vendor transparency.
- Close the loop: Tie AI projects to clear metrics-length of stay, sepsis mortality, readmissions, ED revisits-and retire tools that don't move the needle.
Key takeaways
- HHS will centralize AI efforts under OneHHS, supported by a leadership board and common risk protocols.
- Initial focus is internal, but private-sector collaboration is planned, especially for priority conditions.
- Success will be measured by clinical and operational outcomes, not pilot volume.
If you need a quick primer for staff development, see AI upskilling options by role at Complete AI Training.
For reference on how regulators are approaching AI, review the FDA's materials on AI/ML-enabled medical devices: FDA AI/ML in medical devices.
Your membership also unlocks:
