Blueprint 2026: Bridging the Ops-Tech Gap

What does it actually take to close the gap between the technology available today and the way health care organizations actually operate? At Blueprint 2026, CertifyOS's Anshul Sheopuri sat down with Samir Deshpande—former CIO of Humana and Capital One—and Rajat Puri, CEO of Primera and founder of a 27,000-person health care operations organization, to find out. Here are five ideas that cut through.

1. Start with the Customer—Every Technology Decision Flows from There

The single most important reframe Samir Deshpande offered is one the industry largely ignores: the end customer. In health care, the three fundamental needs are access, quality of care, and affordability. According to Deshpande, almost every technology decision should be traced back to one of those three—but in most health plan IT conversations, they rarely are.

His framework is practical: establish a three-year horizon anchored to customer outcomes, then ask which parts of the technology stack need to be modernized to enable that future—and which just need to be stabilized. The answer isn't always transformation. Sometimes it's targeted improvement. The point is to make that distinction deliberately, not by default.

For an infrastructure layer like provider data management (PDM), that logic is especially important. PDM may not create a direct customer benefit on its own—but it's the rails on which better care navigation, accurate directories, and smoother credentialing all run.

"Start with what the customer needs and have at least a three-year plan. That'll have a level of clarity for where within that ecosystem you want to make choices."

— Samir Deshpande, Former CIO, Humana

2. The Buy-vs-Build Debate Has a Clear Answer—with Two Important Caveats

Both panelists were direct on the buy-vs-build question: buy first, almost always. Engineers love to build. Operators love to run things at scale. But health plans are not software companies, and treating them as such is expensive.

Rajat Puri learned this firsthand—his previous organization spent hundreds of millions building a homegrown PDM solution, only to spend more money for years trying to make it competitive with purpose-built vendor tools. The math never worked.

But Deshpande offered two caveats worth remembering;

1. Vendor pricing can become predatory over time. He cited vendors who raised prices by 500% and 1,400% with no new functionality because customers had built critical infrastructure on their platforms and had no easy exit.
2. When AI makes it genuinely fast and cheap to build narrow, purpose-fit solutions, the calculus shifts. The buy-first rule holds for complex platforms. It may not hold for lightweight, specific-use automation.

The smarter framing, both agreed, is buy-or-partner. Rather than acquiring a tool, co-developing a product roadmap with a vendor means the health plan's scale shapes the innovation, and the vendor's cross-industry visibility improves the product for everyone.

"Always buy if a solution exists and it's a really good solution. Build only if there is no solution, or your needs are truly, truly unique compared to the rest of the industry."

— Samir Deshpande, Former CIO, Humana

3. Compliance Isn't the Enemy—But the Conversation Needs to Happen at the Right Level

One of the most persistent friction points in modernizing health care operations is compliance: the demand for screenshots, manual sign-offs, and verification workflows that technology has long since made unnecessary. The industry has a hashtag for it now: #NoScreenshots.

Deshpande's take was nuanced and practical. Compliance organizations aren't obstinate by nature—they're doing their job. Junior compliance staff are paid to ensure the letter of the regulation is followed. Senior leaders have the authority and context to engage at the level of intent: what was this regulation actually trying to protect against, and does a modern API-based approach satisfy that intent better than a screenshot?

That conversation has to happen proactively, at the right level of seniority, with the regulator. When it does, the answer is usually yes—APIs and audit trails are more reliable than screenshots. But without that engagement, the default answer stays no.

"Being able to take the conversation to the level of what is the underlying requirement and why does it actually exist—allows you to get to the heart of what you're trying to meet."

— Samir Deshpande, Former CIO, Humana

4. AI Is Real—but the ROI Comes from Workflow Automation, Not Task Automation

The AI conversation at Blueprint had more nuance than the average pitch deck. Deshpande drew a distinction that matters: traditional predictive AI—pattern recognition on structured data—has been around since the 1980s and is still dramatically underutilized in health care compared to financial services. Generative AI is newer and faster-moving, but health care is at roughly the second inning.

The framework for evaluating AI vendors that emerged from the panel: clean data first. Any AI system built on inconsistent, unverified, or siloed data will fail—and most of the provider data that feeds these systems is exactly that. The second test is monitoring: can you tell, one month after deployment, whether the model is still doing what you thought it was doing?

Anshul Sheopuri added the sharpest framing: the ROI from AI doesn't come from automating individual tasks. It comes from automating entire workflows end to end. A tool that takes one step out of a ten-step process delivers marginal value. A platform that redesigns the workflow delivers transformation. As AI SDKs increasingly talk to each other and agentic processes handle multi-step operations, the distinction between task tools and workflow platforms will define which vendors survive.

"The ROI is going to come from workflow automation, not task automation. Think about the small part of the process you're going to implement, but design it as an end-to-end workflow you can automate."

— Anshul Sheopuri, CertifyOS

5. Provider Data Should Be a Utility—and Building It That Way Requires the Whole Room

The session closed with the clearest call to action of the day. Deshpande's closing argument: provider data management is a utility—in the same category as the electrical grid or the charging cable standard. No one competes on it. Everyone depends on it. And right now, the industry is paying 50 different organizations to solve the same problem 50 different ways.

The path forward has two steps:

1. The senior leaders in the room need to align on a common standard. That's a coordination problem, not a technology problem, and it can only be solved by the people with the authority to commit to it.
2. Once a standard exists, a small number of companies can compete on who executes it best—driving efficiency, quality, and reliability through genuine market competition rather than redundant reinvention.

Rajat Puri framed the organizational implication plainly: if he were rebuilding his 27,000-person operations organization today, he wouldn't. He'd start with AI and automation at the center, hire an AI engineering team alongside every operations function, and build toward an organization where judgment—not repetition—is what humans provide.

"PDM is a utility. The only way it can become one is if the leaders in this room decide to come together and create a standard that everyone can comply with."

— Samir Deshpande, Former CIO, Humana

Want to read more about Blueprint 2026? Find out what three of health care’s sharpest minds had to say about fixing provider data.