To Build Or To Buy? With Healthcare AI, Both May Be The Answer

When it comes to artificial intelligence in healthcare, hospitals and health systems are revisiting a classic IT dilemma: should they build their own solutions or buy from vendors?

At Keck Medicine of USC, the answer is: both.

Yesha Patel, PhD, associate director of data science and AI at Keck Medicine of USC in Los Angeles, will share insights on this hybrid approach at the upcoming HIMSS AI in Healthcare Forum in Brooklyn. Her session, titled “Case Study: A Blended Approach to AI Acquisition and Implementation,” will outline how Keck Medicine strategically combines in-house development with select vendor tools to integrate AI securely and effectively into its existing health IT systems.

Striking the Right Balance

“Keck Medicine has found an optimal balance between innovation, speed to market, and sustainability by using a blended build-and-buy approach,” Patel said in a preview of her session. “Healthcare has unique demands—patient safety, data protection, and clinician trust are non-negotiable.”

She emphasized that relying exclusively on either internal development or external vendors can limit an organization’s agility. By blending both, Keck Medicine has been able to tailor AI solutions to specific workflows while still leveraging proven technologies that save time and reduce development costs.

“There’s no one-size-fits-all system in healthcare AI,” she said.

Matching the Solution to the Problem

Some applications, such as predictive models for patient deterioration or streamlining administrative tasks, may work well with commercial tools that integrate with existing EMRs or third-party platforms. Other needs—especially those requiring deep contextual knowledge or access to proprietary data—are better suited for custom-built solutions.

“The key is being intentional,” Patel noted. “We evaluate each use case carefully: Do we have the right expertise internally? Is customization critical? What are the long-term costs and support requirements? These are strategic decisions, not just technical ones.”

Why Blended Can Be Better

So why does this hybrid approach work better? According to Patel, it allows for both agility and scalability while respecting the clinical and operational realities of a healthcare environment.

“Building from scratch sounds great in theory, but limited budgets, staffing constraints, and stretched IT teams make it tough,” she explained. “On the other hand, fully relying on vendors can lead to poor integration with EHRs and workflows, limited customization, and low clinician adoption.”

With a blended strategy, health systems can selectively use vendor tools while maintaining control over core data models and integration points.

Faster Development, Better Alignment

For example, a hospital might buy a commercial NLP tool to extract insights from radiology reports, but then develop its own rules engine to apply those insights based on internal triage protocols.

“That way, we speed up development without sacrificing clinical relevance or compliance,” Patel said.

Managing Complexity and Technical Debt

One major challenge Patel highlighted is the buildup of technical debt when implementing multiple standalone AI solutions.

“When you piece together tools from different vendors—like one for sepsis prediction, another for radiology imaging, and another for note processing—you end up with a fragmented system,” she said. “Each tool has its own infrastructure, data formats, APIs, and support models. They don’t always work well together.”

Over time, this patchwork creates operational inefficiencies and high maintenance costs.

“We were spending significant resources just to keep everything talking to each other,” she said. “Every update risked breaking something else.”

A Unified, Scalable Vision

Ultimately, the real issue was long-term complexity.

“Each new point solution added technical debt and increased our reliance on multiple vendors,” Patel said. “It made our environment more fragile and harder to scale.”

That’s why Keck Medicine shifted toward a more deliberate, blended strategy—ensuring that any external tools fit into a unified architecture and support long-term sustainability.

Source: Healthcare IT News