Healthcare systems are under pressure to do more with less, and artificial intelligence has become a central part of that conversation. For Fairfax, Va.-based Inova, the path forward involves a deliberate, measured approach to its AI strategy rather than a rush to adopt every new tool. The organization has spent years figuring out how to fund and deploy technology safely, according to Chief Data and AI Officer Jon McManus. “That doesn’t change overnight,” he said.
Inova currently has 67 different AI features in production, with continuous monitoring of value, safety, and reliability. The organization has evaluated more than 400 different AI features to date, a process that helps separate useful capabilities from market noise.
How Inova Decides Which AI Tools Get the Green Light
The organization’s governance structure starts with an AI ethics and oversight committee, which McManus describes as having a robust setup. A team of five AI product managers handles idea management and business proposals, acting as the front door to governance. The process weighs risk, liability, policy changes, training needs, cost, and value-creation components.
McManus says the core question is simple: “Is this OK or not to do?” The decision rests on whether a feature is ethical, safe, reliable, and mechanically sound. The committee is chaired by a physician and includes 24 members from across the organization, including seven physicians, administrators, nursing leaders, and representatives from supply chain, legal, privacy, and bioethics. There’s also a component for the voice of the community and patients.
Related: Government scientists test AI research assistants
It’s a structure that sounds complicated, but McManus says it’s been streamlined into an easy-to-understand process. The goal is to centralize the approval decision without centralizing implementation. Inova encourages its workforce to explore AI opportunities, while the interdisciplinary committee answers that single question on behalf of the organization.
The approach makes sense given the scale of the challenge. When a health system can have over 1,000 vendors in its portfolio, someone has to sort through what’s actually worth pursuing. Inova’s answer is a formal process that treats the entire AI portfolio as a balanced set of investments, some of which may not generate financial returns but still affect quality of care.
Taking, Shaping, and Making AI Solutions
Inova divides its AI inventory into three categories. The first is “take,” which refers to approved medical devices or clinical decision support algorithms cleared by the U.S. Food and Drug Administration. These are implemented as vendor products without modification.
The second category is “shape,” which covers most of what the organization does. This involves buying a product from a vendor and configuring it with low-code tools to fit Inova’s workflows. Almost all Epic AI features fall into this category. The third is “make,” where Inova builds truly custom solutions using its modern data stack.
Related: Global wellness market to hit 11 trillion
Currently, about 10% of Inova’s AI inventory is “take,” 80% is “shape,” and 10% is “make.” That distribution matters for hiring decisions. McManus notes that many organizations think they just need to hire AI engineers, but Inova has invested in the skills required for the 90% of its inventory that involves taking and shaping products.
On the technical side, the organization runs its stack on Microsoft Azure, with data positioned primarily in Databricks. The infrastructure includes FiveTran for ingestion, dbt Labs and Databricks for transformation, Astronomer for orchestration, and Microsoft Fabric and Power BI for delivery. DataHub handles metadata management, DataRobot covers data science, and GitHub manages code versioning.
Inova is also working on an enterprise knowledge management platform. The idea is to orchestrate institutional knowledge once in a unified appliance, almost like a data warehouse for knowledge. This would support continuous API utilization whenever an AI feature needs a knowledge-based component, rather than redoing retrieval augmented generation for every single feature.
Monitoring is a key part of the strategy. The organization tracks quantitative and qualitative elements of accuracy, performance, drift, and safety. It also uses industrial-level monitoring for network logs, with partners including Signal 1, Rubrik, and CrowdStrike. This helps watch for agent use that falls outside approved protocols.
Related: Hospital Groups Urge Becerra to Delay Info‑Blocking Rule
For workforce upskilling, Inova has created proprietary distribution portals to slow down the manifestation of change. The idea is to keep things static enough for people to build literacy, even if the backend changes frequently. The organization has a detailed communication plan that includes AI town halls and an AI champions network in development.
One notable detail: Inova updated its patient safety reporting system to include a designation for AI. Team members can anonymously share a patient safety event and list AI as a partial or primary contributor. That allows the organization to facilitate root cause analyses and support effective responses.
Government scientists are actively testing AI research assistants to see how these tools can support complex healthcare data analysis and patient management tasks. [1]
The global wellness market is projected to reach 11 trillion by 2027, illustrating the massive economic potential of sectors that integrate advanced technology with human well-being. [2]