Inova Health System, based in Fairfax, Virginia, is carefully expanding its artificial intelligence capabilities while maintaining strong oversight. The organization balances innovation with governance to ensure safe and effective use across clinical and operational areas.
Governance as the foundation
An AI ethics and oversight committee evaluates potential use cases. This committee reviews risk, policy implications, training needs, and expected returns before approving any AI implementation.
Jon McManus, Inova’s chief data and AI officer, stated that the system currently runs 67 AI features in production. Continuous monitoring ensures safety, reliability, and value. “We examine the risk and liability of each feature, necessary policy changes, adequate training, cost, and value creation,” he explained. “Are we taking it, shaping it, or making it? Those represent three distinct disciplines within our AI portfolio.”
AI deployments fall into three categories: “take” (FDA-approved medical devices requiring no modification), “shape” (vendor-provided tools adapted for Inova’s workflows), and “make” (custom-built solutions).
The approach mirrors a growing trend in healthcare, where organizations increasingly use configurable vendor solutions instead of building everything internally. This method allows Inova to expand AI use without overwhelming its engineering teams while retaining control over tool integration.
A modern data stack enables AI orchestration
McManus emphasized that connecting data to models is essential for AI success. The infrastructure supports “agentic AI”—systems capable of performing tasks autonomously across the organization. The health system is developing an enterprise knowledge management platform to standardize retrieval-augmented generation (RAG) for AI features, eliminating redundant development for each new tool.
Related: New Therapies Pose Challenges for Insurers
Workforce adoption through literacy and trust
Since AI appears in many vendor tools, Inova faces a literacy challenge. “It’s in a lot of different places, and it looks different,” McManus said. “That creates confusion. How can employees become familiar when there are 100 ways to do things?”
Training includes AI town halls, a champions network, and formal courses integrated into Inova’s learning and development program. The organization also updated its patient safety reporting system to flag AI-related incidents, enabling analysis when AI contributes to adverse events.
McManus stressed the importance of stability in building trust. “Employees need time to learn, so interfaces must remain consistent. The place where users find AI tools should feel familiar, so they grow comfortable using them.”
Inova is preparing to scale federated agent building, allowing thousands of employees to create AI solutions for specific problems. “Enabling 2,000 people to solve 20,000 problems safely is the goal,” McManus said. “We’re nearing confidence in how to achieve that.”
As healthcare groups adopt new approaches, Inova’s structured method offers a model for responsible AI expansion.