Scaling Health AI: Four Places, Many System Clues

Health care is using AI more often, but getting it to work at a large scale is still a new problem. Many projects are still small tests or early use cases. The research looked at how health systems can move from one-off tools to methods that can be built, adopted, used, and expanded across care settings. It focused on lessons that might travel from one system to another.

The work compared four places: Catalonia, Norway, Singapore, and Queensland. The research used 60 documents, 34 interviews, and 5 focus groups. Across those activities, 50 people took part, including strategic decision-makers, policymakers, and lead clinicians. The questions covered why systems wanted change, what needs were emerging, what kinds of AI tools were being tried, what benefits people expected or saw, what problems showed up, and which parts seemed useful elsewhere.

The analysis happened in two passes. First, each case was reviewed on its own to find repeated themes. Then the cases were compared using the Technology, People, Organization, and Macroenvironment framework. That lens helped organize system-level factors, but the review also left room for new themes to appear from the data.

The findings point to a practical check: before scaling AI, a health system has to look at its own digital plans, funding, culture, history, and legacy technology. A new AI tool may fit one setting and struggle in another if those background pieces are very different. The useful question becomes not only whether AI can work, but which system conditions have to line up before it can grow.