Health systems and payers are entering a new era of artificial intelligence adoption—but their buying strategies haven’t caught up yet given how fast the technology and market are moving. Many are still purchasing AI the same way they bought software a decade ago: one tool, one use case, one department at a time.
The result in today’s fast-moving market ends up being a mishmash of vendors, pilot purgatory, mounting integration complexity and unclear return on investment.
To move past experimentation into enterprise‑scale deployment, healthcare leaders must rethink how they evaluate, procure and govern AI across their organizations. That shift requires moving beyond point‑solution evaluation toward intentional system design. Increasingly, buyers are no longer asking whether a solution works, they’re asking where it fits, what it depends on and how it compounds value over time.
AI can no longer be treated as an application decision; it needs to become an architectural one. Leaders must stop evaluating tools in isolation and start designing a cohesive enterprise AI architecture.
Organizations have the potential to find success through a new framework for AI procurement, one that recognizes three distinct layers: core enterprise platforms, foundation model platforms and specialized context-aware, use‑case innovators. For startups, this architecture view is no longer optional as they think through their go-to-market strategies. It is quickly becoming the lens through which enterprise buyers evaluate relevance, durability, moats and long‑term strategic fit.
Layer 1: Core infrastructure and enterprise platforms
Core infrastructure and cloud platforms now form the foundation of the healthcare enterprise stack, led by hyperscalers such as Microsoft Azure, AWS and Google Cloud. These providers deliver the compute, storage, security primitives and data services that underpin modern AI capabilities. Sitting on top of this infrastructure layer are longstanding systems of record and enterprise application vendors—most notably Epic, along with platforms like Workday, ServiceNow and Salesforce—which continue to serve as the digital backbone of most healthcare organizations.
Together, these vendors increasingly embed AI directly into the enterprise stack, leveraging privileged access to data, workflows, identity and compliance frameworks that health systems and payers already trust. Hyperscalers in particular are becoming foundational not just as infrastructure providers, but as policy‑setting actors defining how data can be stored, moved and secured and, increasingly, how AI workloads are governed at scale.
For startups, this layer defines the gravitational center of the enterprise. Core platforms and cloud providers shape procurement constraints, deployment patterns and architectural boundaries long before a point solution is evaluated. They determine where data live, how models are accessed and what “enterprise‑ready” actually means. While this layer offers the lowest‑risk adoption path for buyers, it compresses opportunity for vendors whose value propositions overlap too closely with native functionality or cloud‑embedded services. Startups that fail to account for this reality risk being sidelined by platform evolution rather than displaced by direct competitors.
Layer 2: Foundation model platforms
The emergence of foundation model platforms introduces both opportunity and strategic tension. New offerings such as OpenAI’s Frontier and Anthropic’s Cowork point toward a future where enterprises can operate secure, organization‑specific AI environments that sit between core systems and end‑user applications. And these offerings are becoming more and more woven into the Layer 1 infrastructure (e.g. Microsoft’s recent Copilot Cowork announcement that is powered by Anthropic’s Cowork offering).
These platforms function as orchestration layers, enabling custom application development, centralized governance, reusable prompts and workflows, and controlled access to proprietary data. For startups, this layer could increasingly represents the control plane of enterprise AI. Products that align with it—rather than attempt to bypass it—may benefit from faster procurement cycles, deeper enterprise embedding and clearer paths to scale.
At the same time, foundation platforms raise existential questions. If they become the default operating system for enterprise AI, startups must decide whether they are building on top of, alongside or in competition with them. That choice has implications for product architecture, deployment models, data ownership and even pricing strategy. Startups that fail to articulate this positioning clearly may struggle to earn enterprise trust, regardless of model performance.
Layer 3: Specialized AI startups
There remains an expanding universe of AI startups attacking discrete operational pain points—from prior authorization and claims management to revenue cycle optimization and clinical documentation. Many of these companies deliver fast, measurable ROI and solve problems that core vendors are slow to address.
But the bar is rising. Enterprise buyers are increasingly evaluating Layer 3 solutions not just on functionality, but on architectural compatibility. How does this product integrate with core platforms? Does it leverage foundation model environments already approved by the organization? Can it scale without introducing fragmentation, redundant infrastructure or governance risk?
As Layers 1 and 2 mature, startups that succeed will be those that position themselves as architectural complements rather than isolated tools. Integration strategy, data boundaries, security posture and governance alignment are becoming as important as algorithmic sophistication.
The strategic imperative
Healthcare AI procurement is no longer about picking winners one use case at a time. It’s about orchestrating across layers—aligning enterprise platforms, foundational capabilities and specialized applications into a coherent system that can evolve.
This shift demands a new level of enterprise awareness from startups. Winning vendors will understand not just their buyer’s problem, but their buyer’s stack. They will design products that slot cleanly into existing and emerging enterprise architectures and communicate that fit with clarity and discipline.
Healthcare’s digital frontier is no longer about whether to adopt AI, but how to organize it. Procurement strategy has become a core enterprise capability. The ability for startups to understand where they fit within that strategy may determine whether they scale, stall or disappear.
Keith Figlioli is managing partner at LRVHealth, a venture capital platform that invests in technology-based businesses in the healthcare sector on behalf of health systems and payers.
Editor's note: LRVHealth’s investments include, but are not limited to, AI startups, digital health solutions, technology-enabled services, clinical services, diagnostic tools and medical devices.