Part One: Major Shifts in Healthcare AI: From Foundation Models to Clinical Reality
Welcome to the first installment of our five-part series, Healthcare AI and Infrastructure Dependency. In this series, we explore why AI success in healthcare depends on more than just powerful technology; it requires the systems, infrastructure, partnerships, and governance that allow innovations to operate safely, scale effectively, and deliver lasting value. We’ll break down the complex ecosystem of healthcare AI, showing how a trusted infrastructure creates compounding advantages for organizations, clinicians, and patients alike.
Q: What major shifts are you seeing in healthcare AI with new entrants?
The addressable market for healthcare AI infrastructure is expanding.^1 ^2 Every clinical and administrative workflow in healthcare is now a candidate for AI augmentation. The foundation model companies are building general-purpose AI capabilities that will reshape every industry, including healthcare. These systems are pushing aggressively with large-scale models and cloud infrastructure designed to power clinical, operational, and patient-facing applications. In demos, they summarize clinical notes, generate prior authorization letters, and extract billing codes from medical records.
Likewise, healthcare’s major electronic health record (EHR) vendors are increasingly positioning AI as a way to expand beyond their traditional roles in documentation, billing, and coding to influence the entire care journey, from intake and triage to clinical decision support, care coordination, and follow-up. AI is often embedded directly into their platforms in an effort to become the central operating system for clinical workflows, using automation and predictive tools to touch nearly every step of patient care.^3
The result is a growing collision between AI-first infrastructure providers and established health tech platforms, each aiming to define the layer that ultimately controls how intelligence is deployed across the healthcare system.^4
EHR vendors possess significant advantages in workflow integration and regulatory familiarity; however, they can face architectural constraints. Many platforms were originally designed to prioritize documentation, billing, and compliance. Consequently, these structures can present challenges in rapid iteration and large-scale deployment of modern AI. As a result, AI capabilities currently embedded within EHRs are often optimized for incremental improvements, such as streamlining note generation or coding efficiency, rather than facilitating entirely new models of care delivery.^5
On the other side, foundation AI and cloud companies bring powerful models and rapid innovation cycles. Their disadvantage is healthcare itself: clinical workflows are complex, highly regulated, and deeply embedded in hospital operations. Without direct control of the EHR and the surrounding care processes, AI platforms can struggle to move from promising models to real-world clinical adoption.
In practice, this creates a strategic gap where healthcare EHR companies control the workflow but may lag in AI capability, while foundation model companies lead in technology but lack control of the clinical environment where that technology must operate. That gap is shaping the next phase of competition in healthcare technology.
Q: How do healthcare incumbents and AI-native companies relate in the integrated AI healthcare ecosystem?
The growing tension between healthcare incumbents and AI-native companies reflects a fundamental divide. On one side are the AI foundation models powering speed and efficiency. On the other is the infrastructure embedded in the care model that provides the operations, oversight, and scale required to deliver care.
AI foundation models can’t operate in isolation. They require the infrastructure, control systems, and workflows that allow them to operate safely and effectively at every layer.
Likewise, EHRs maintain the core infrastructure and established workflows of patient care. Because many of these foundational systems are legacy-based, the AI models they deploy may be limited in scope and struggle to innovate quickly.
IKS Health built the foundational platform. While the AI model is the core technology, the integrated system that enables it to operate safely and effectively consists of the EHR infrastructure, the hospital ecosystem, and the clinical workflow it plugs into. It also includes the compliance infrastructure it must satisfy, the human oversight that catches its errors before they reach a patient, and the integration across multiple interdependent services. By providing this trusted infrastructure, IKS Health is able to protect and expand the margins to care for clinicians and patients.
Q: How does this infrastructure ensure patient safety in AI?
You can build more powerful technology, but without control of the operational environment, AI foundation model companies struggle to deploy reliably. In healthcare, if AI hallucinates and is not managed, it poses a patient safety event. A wrong medication dose, a fabricated lab value, or an incorrect diagnosis code that changes a treatment plan are not edge cases; they are the failure modes that every health system executive, every payer, and every regulator is laser-focused on preventing.^6
IKS Health deploys AI models effectively across complex clinical systems, scaling safely and reliably across the entire connected care journey. Healthcare organizations and clinicians need reliable integration into workflows. Without these systems, even the most advanced AI models risk irrelevance.
This isn’t a theoretical concern; it’s the reason that healthcare AI adoption, despite enormous enthusiasm, remains cautious, incremental, and deeply dependent on a trusted infrastructure like IKS Health.^7
Sources:
^3. HIMSS24: How Epic is building out AI, ambient technology for clinicians (Fierce Health)
^4. Microsoft expands AI capabilities to shape a healthier future (Microsoft)
^5. The hidden costs of legacy data infrastructure in healthcare (Innovaccer)
