A comprehensive framework to assess AI’s value and clinical outcomes is lacking. Public–private partnerships (PPPs) are vital for demonstrating AI’s practical, scalable value in health. Such collaboration is mutually beneficial: policy-makers gain insights into costs and potential returns on investment, while the private sector can be appropriately compensated and benefit from understanding regulations, government priorities and policy frameworks. As a part of this there is a need to move towards adapting existing methods into more flexible approaches to take into account the outcomes of AI tools, such as capturing real- world evidence and tracking benefits. Complexities in securing AI financing A dual approach to financing AI in health supports a gradual and structured rollout of AI technologies. But the approach means that financing AI transformation is not straightforward, as AI investment requires two different time horizons: Navigating the “valley of death”10 Short-term public–private investments are crucial for initially managing change, generating evidence and demonstrating the value of AI in health. These investments are essential for showcasing AI’s effectiveness and establishing proof of concept. However, beyond such short-term support, multi- year investments spanning two to three years are necessary to ensure sustained progress. In the initial stages, various stakeholders, including payers and funders, focus on education and upskilling to ensure that all parties are adequately prepared for AI integration. These short-term investments help build the foundational infrastructure and training necessary for AI adoption. Crucially, these investments should aim to implement multiple solutions in parallel, rather than isolated point solutions, enabling a whole-system transformation. For instance, according to Accenture, AI in health could save the US health economy $150 billion annually by 2026.11 These investments are essential for innovation to get through the “valley of death”; however, they carry risks, and stakeholders should consider a blend of different financing options, both public and private. Additionally, a significant challenge is in establishing appropriate revenue models to ensure that early adopters are compensated for their initial investments and ongoing costs. This includes considerations for procurement, reimbursement and other financial mechanisms relevant to different regions and health systems. Moreover, it is not just about the technological solution but about a new way of working – addressing “how” the transformation happens as well as “what” is implemented. This involves removing constraints such as procurement processes and ensuring the right capabilities are in place, including training and PPPs to enable rapid assessment and innovation cycles. Securing long-term market-based financing Long-term market-based financing aims to support sustained investment and scaling, ensuring continuous positive impacts on local economies. This phase will allow AI technologies to be effectively scaled and maintained, providing lasting benefits. To facilitate ongoing development and innovation, a new costing model is needed. This model should detail the allocation of funds and identify who is responsible for financing different stages of market development, especially for early innovators and research teams. It should include guidelines for covering costs associated with research, development, implementation and scaling of AI technologies to ensure a clear and sustainable financial pathway. Studies indicate that countries investing in AI technologies in health have received a return on investment (ROI) of 10–15% annually over a five-year period.12 Furthermore, continuous investment in AI technologies could lead to savings of up to 10% in healthcare spending.13 Following the initial change management and value demonstration phase, financing for these long-term activities should rely on traditional market principles. Effective reimbursement and revenue models must be in place to ensure that providers and developers still feel they have an incentive to implement and improve AI solutions. This sustains innovation and operational efficiency over the long term. Additionally, one critical aspect often overlooked is the necessity of multimodal data training sets, which are essential for many AI applications. Such datasets are not only crucial for accelerating AI but also present a significant opportunity for governments to generate revenue or share in novel intellectual property (IP). However, these business models, often managed by government payers, can pose risks, as they must ensure that ongoing investments yield the anticipated benefits. Additionally, incorporating concepts such as health technology assessments (HTA) and health economics and outcomes research (HEOR) can help evaluate the value and impact of AI solutions, ensuring that investments are justified and beneficial in the long run. There is a need for convergence of HTA standards across borders to allow for a globally integrated approach. Investment strategies must be designed to ensure equity throughout the entire health value chain, from development to delivery of solutions. This includes preventing the exacerbation of existing disparities and supporting local production initiatives as advocated by the G20 intergovernmental forum,14 the World Health Organization and other bodies, including establishing local networks for AI-driven health research, development and validation. The Future of AI-Enabled Health: Leading the Way 16

The Future of AI Enabled Health 2025