What It Actually Takes to Scale AI in Healthcare: Lessons from Sentara Health’s CTO

At ViVE, one theme was everywhere: AI.

But beneath the hype, a more important question kept coming up:

Why are so many healthcare AI initiatives still stuck in pilot mode?

To get a real answer, the Beat Podcast sat down with Jeff Thomas, SVP & CTO at Sentara Health — a leading integrated provider-payer system — to talk about what it actually takes to move AI from experimentation to impact.

Watch the full conversation

AI Isn’t the Problem — Outcomes Are

One of the most striking takeaways from Jeff:

“AI has to be efficient, effective, and it has to meet the outcomes…[otherwise] it doesn’t solve anything.”

It’s a simple idea — but one that cuts through a lot of the noise.

Healthcare organizations aren’t struggling to adopt AI because of a lack of tools. They’re struggling because AI initiatives often:

• Don’t tie back to measurable outcomes
• Add cost without clear ROI
• Fail to integrate into real clinical workflows

For Jeff and his team, success isn’t about deploying more AI — it’s about delivering better care, more efficiently.

The Real Bottleneck: Data Accessibility

When the conversation turned to AI strategy, Jeff immediately shifted focus:

“When you talk about AI, we keep forgetting about the data… the problem is, that data is locked.”

This is the core challenge.

Healthcare organizations are sitting on massive amounts of data — but:

• It’s fragmented across systems
• It’s difficult to access in real time
• It’s often too stale to be useful

And that creates a fundamental limitation:

“It doesn’t do us any good to look at data that’s three years old… a year old… maybe even a day old.”

For AI to work — especially in clinical and operational settings — it needs timely, accessible, and usable data.

Why Infrastructure Matters More Than Algorithms

There’s a tendency to focus on models, agents, and applications.

But Jeff makes it clear: the real work happens underneath.

“How do we have our structure… our data layer in a way that’s accessible for inference?”

In other words, AI success depends on having a data foundation that can:

• Support real-time access
• Enable AI inference without impacting production systems
• Avoid costly data duplication and sprawl
• Scale efficiently in the cloud

Without that foundation, even the most advanced AI initiatives will struggle to deliver value.

The Cost Challenge No One Talks About

Another critical theme: cost.

Healthcare organizations are under increasing pressure to:

• Reduce operational expenses
• Justify new technology investments
• Deliver more with less

At the same time, AI and cloud adoption can increase costs if not managed carefully.

As Jeff puts it:

“What are we doing to reduce those costs, but still deliver that care or better care?”

This is where many AI strategies break down.

It’s not just about enabling new capabilities — it’s about doing so in a financially sustainable way.

The Hidden Unlock: Reducing Tech Debt

One of the most counterintuitive insights from the conversation:

“The complexity is getting rid of the tech debt. The complexity is not building the new capability.”

Before organizations can fully take advantage of AI, they often need to:

• Rationalize legacy systems
• Simplify their architecture
• Reduce operational complexity

This “addition through subtraction” is what creates the space for innovation.

Connecting the Dots: Where Silk Fits In

Jeff briefly touches on a critical piece of the puzzle:

“There’s tools that I use like Silk… it allows me to do AI inferencing and present that data… [without] impacting my production.”

This speaks directly to the gap many healthcare organizations face.

To make AI work at scale, you need a data layer that can:

• Make data accessible in real time
• Support AI inference workloads
• Eliminate the need for costly data replication
• Reduce strain on production systems
• Control cloud and infrastructure costs

That’s exactly what Silk is designed to do.

Silk enables healthcare organizations to:

• Unlock real-time data access for AI and analytics
• Run inference workloads efficiently without impacting core systems
• Reduce cloud costs by avoiding unnecessary duplication and movement of data
• Modernize data infrastructure without adding complexity

From AI Strategy to Real-World Outcomes

The biggest takeaway from this conversation isn’t about a specific technology.

It’s about a shift in mindset.

AI success in healthcare requires:

• A focus on outcomes, not hype
• A foundation of accessible, real-time data
• A commitment to cost efficiency and sustainability
• The discipline to reduce complexity before adding more

Organizations that get this right won’t just experiment with AI — they’ll scale it.

See What This Looks Like in Practice

If you’re thinking about how to move your AI initiatives forward, this conversation is worth your time.

👉 Watch the full interview.

And if you’re exploring how to unlock real-time data for AI in your organization:

👉 Request a demo to see how Silk can help.