With over 300 CE marked AI applications in radiology alone, it is only obvious that the number of AI applications for clinical use has grown exponentially. As adoption increases, however, implementing and maintaining individual point solutions can become cumbersome for hospitals. Managing multiple vendors, integrations, contracts, updates, performance monitoring processes, and governance requirements places a growing operational burden on clinical, IT, and procurement teams.
As a result, hospitals are looking beyond single AI applications. They are seeking AI infrastructure that enables them to deploy, manage, and scale multiple AI solutions in a safe, efficient, and sustainable way. This shift has contributed to the rise of AI platforms: infrastructure solutions designed to consolidate multiple AI tools into a single, manageable environment, helping hospitals with deployment, integration, and oversight.
But with a lot to choose, how do you find the platform that fits your organisation? Platforms can differ substantially in how they position themselves, how they structure contracts, which functionalities they offer, how much customisation they allow, and how well they align with a hospital’s broader AI strategy. For hospitals, understanding these differences is essential to identifying a platform that fits their organisational needs, technical environment, procurement processes, and long-term ambitions for AI implementation.
To bring transparency and clarity into this complex ecosystem, Health AI Register will soon introduce a dedicated domain for AI platforms. In this three-part blog series, we are going to discuss the key differences between AI platforms across three dimensions; positioning and contracting, functional capabilities and customisation and strategic features. For starters, let’s delve into positioning and contracting!
Positioning: who built it and what is the scope?
AI platforms aren’t built by the same kinds of companies, and where they come from has an impact on how the products are designed, positioned, and used. Some platforms are developed by established PACS or RIS vendors, where the main advantage may be seamless integration into existing imaging workflows. However, hospitals may also consider the risk of reduced vendor neutrality or increased dependence on a single ecosystem. Other platforms are built by dedicated AI platform companies, whose primary focus is often AI deployment and orchestration. These companies may be ahead of the curve in terms of platform functionality, flexibility, and support for multi-vendor AI environments. OEMs, such as major imaging equipment manufacturers, have also entered this space, although building and maintaining AI platforms has proven challenging for some. In several cases, OEMs have chosen to partner with or white-label existing platform providers rather than develop a fully independent solution. Adjacent players from the broader imaging ecosystem have made similar attempts with mixed results: Bayer, for example, entered the market with Calantic (built on Blackford) but subsequently withdrew, illustrating that sustaining an AI platform business is not straightforward even for well-resourced healthcare companies. Finally, some platforms originated from AI product companies that expanded their offering into a broader platform, often building on their experience with specific clinical applications while gradually extending towards wider infrastructure capabilities.
The strategic focus of AI platforms may also differ. Some take a curated approach, offering a single, carefully selected AI application per clinical use case. Others give hospitals access to a broader range of AI tools for the same clinical problem, so they can pick the solution that best fits their needs. In some cases, platforms also include AI applications developed by the platform provider itself. This can create an interesting dynamic, particularly when these in-house solutions are offered alongside competing third-party applications within the same environment. For hospitals, this raises relevant questions around neutrality, transparency, and how applications are selected, prioritised, or presented. A growing number of platforms also support hospital-developed AI applications. This is increasingly relevant as larger and academic institutions invest in building their own algorithms and look for ways to deploy them in clinical workflows alongside commercial solutions; for these hospitals, the ability to host in-house AI on the same platform can be a decisive factor. Neither approach is inherently better. The right fit depends on whether a hospital values consistency and simplicity, flexibility and choice, or a balance between both.
Lastly, AI platforms may split in terms of clinical scope. Several platforms are designed exclusively for radiology, others extend to broader medical imaging and some are even positioning themselves for hospital-wide use across multiple specialties and modalities. For hospitals considering an AI infrastructure partner this distinction matters considerably.
Contracting and deployment: where does the complexity sit?
On the contracting side, platforms fall broadly into two models. In a centralized contracting model, the platform provider serves as the single point of contact for all financial, privacy, and security agreements, including those with the underlying AI vendors. For example, this contracting model could significantly reduce administrative overhead for hospital procurement teams. In a decentralized model, the hospital itself manages separate contracts directly with each individual AI vendor for those same arrangements. The platform may still provide technical integration, but the contractual relationship with each AI vendor sits with the hospital.
In addition, pricing models vary quite a bit between platforms. Some charge hospitals separately for access to the platform itself, in addition to the cost of the individual tools running on it. Others roll everything into a single fee, which can make it less clear what hospitals are actually paying for behind the scenes.
Finally, deployment options differ between cloud, on-premise, and hybrid models. Cloud-based platforms could potentially be easier for hospital IT teams to manage, but they can raise questions around data storage and internet reliability. On-premise setups keep everything inside the hospital’s own systems, giving organizations more control but also more responsibility for capacity, maintenance and support. Hybrid models combine elements of both, offering a layer of flexibility.
What this means in practice
For hospitals evaluating AI platforms, these two dimensions, positioning and contracting, largely determine who you are working with and how that relationship is structured. Before comparing features, it is worth clarifying what type of vendor relationship, contracting model and deployment model fits the hospital’s operational and governance context.
Stay tuned for the next post in this series, in which we look at what platforms actually do: their functional capabilities, including how they handle data orchestration, clinical workflow integration, and reporting.