Is AI in healthcare scalable and sustainable?
AI-based innovation is increasingly seen as a strategic lever to strengthen clinical decision-making, patient engagement, and health system performance. Still, moving from promising research outputs to scalable and sustainable implementation remains a major challenge.
This was the focus of the workshop “Bridging Research and Market: Financing and Deploying AI in European Health Systems”, organized by SDA Bocconi as part of the Horizon Europe CINDERELLA project. The workshop brought together experts from clinical practice, entrepreneurship, engineering, health policy, venture capital, and digital health implementation to discuss how AI-based healthcare innovations can move from research settings into routine care.
The discussion highlighted several conditions that are essential for scalable and sustainable AI innovation in healthcare.
Successful implementation depends first on clinicians’ trust, AI literacy, cultural aspects and perceived value of AI. Professionals need to see these tools as augmenting, rather than replacing, clinical work. Involving them early in co-development and research can build familiarity, reduce resistance, and make adoption easier. At the same time, AI tools must be embedded into routine patient pathways and designed for usability in real-world clinical settings.
Scale-up is often constrained by fragmentation at multiple levels: between developers and adopters, across healthcare providers (for example in terms of data infrastructure), between public and private settings. This reduces the transferability of AI solutions from one institution to another and may widen inequalities in access to innovation. More scalable models require multidisciplinary teams, shared standards, and stronger industrial–research–clinical partnerships from the outset.
Evidence generation must evolve to reflect the specific characteristics of AI. Randomized controlled trials remain important, but AI technologies evolve rapidly and require continuous, real-world evidence generation, supported by implementation-oriented evaluation. This approach may also be better aligned with the pace of venture capital, as investors typically seek scalable business models within relatively short time horizons that may not correspond to the time needed to validate and deploy AI-based healthcare tools.
Clear governance frameworks are needed to better define responsibilities, including medico-legal accountability. Regulation remains essential to ensure safety and efficacy, but it should not create disproportionate barriers to innovation, particularly for small-medium enterprises and academic spin-offs. Public procurement processes should also be made clearer and easier to navigate, to ensure that they do not become a major bottleneck in the innovation process.
Funding mechanisms need to evolve as well. Research grants can successfully support development and pre-market validation, but additional mechanisms, including but not limited to transitional grants, are required to finance implementation, long-term maintenance and recognition of the AI broader dimensions of value.
Conversations like the one hosted at SDA Bocconi are essential to nurture this agenda. Bringing together such diverse stakeholders around the same table creates the conditions for a genuine exchange of perspectives — and it is precisely the opportunity to share them that can drive more implementable, inclusive, and sustainable approaches to AI innovation in healthcare.