-

Workweek Newsletter {beacon}

Blake Madden

Sep 24th, 2026

{cta_url_read_in_browser = community_base_url + "/library/" + article_id + "?utm_source=newsletter&utm_medium=email&utm_campaign=" + edition_slug + "&utm_content=read_in_browser"}{cta_url_read_in_app = community_base_url + "/library/" + article_id + "?utm_source=newsletter&utm_medium=email&utm_campaign=" + edition_slug + "&utm_content=read_in_app"}{cta_url_join_conversation = community_base_url + "/library/" + article_id + "?utm_source=newsletter&utm_medium=email&utm_campaign=" + edition_slug + "&utm_content=join_conversation" + "#comments"} {if profile.vars.member_status == "lead" || profile.vars.member_status == "unfit"} Read in browser {else}{if profile.vars.member_status == "fit"} Read in Boardroom {else}{if profile.vars.member_status == "member"} Join the convo {else} Read in browser {/if}{/if}{/if}

In partnership with

Sorry, couldn’t help myself with that subject line. Cheeky!

As this gets published today, Oracle is concluding the final day of its Health and Life Sciences Summit in Orlando chock full of discussion on the future of the company and healthcare, during which Seema Verma, EVP and GM at Oracle Health and former CMS Administrator among other advisory roles, is playing a prominent position. Oracle Health also had a slew of announcements alongside the event - everything from a new oncology EHR to new AI-enabled solutions for rev cycle, life sciences, and more with the ultimate goal of connecting clinical research and care delivery working with enterprises on an outcomes based pricing model.

Prior to this event I got in a quick chat with Seema Verma at Sessions. So between the event happening in Orlando, Sessions, and everyone discussing the future role of the EHR and innovation, I thought this was a timely, yet digestible dive into how Seema & Oracle view the world.

Anyway, Seema was great, I enjoyed my convo with her, appreciated her time, and we dove head first into all things Oracle Health.

P.S. VBC leaders: by 2027, prospective risk adjustment and audit-ready documentation stop being differentiators and become the price of staying in the game. I hope you’ll join me and leaders from Privia Health and Navina on 9/30 to see how the best VBC orgs are building their winning playbooks now. Save your spot.

PPS - UT vs. other UT (you can decide which is which) game prediction at the bottom. Hook ‘em.

Was this email forwarded to you?

Subscribe Here

Talking Shop with Seema Verma

Seema Verma on Walled Gardens, Point Solutions, and Why Medical Coding Might Not Survive

Quick context for the 3 (maybe fewer) of you who need it: Seema helps head up Oracle Health and Life Sciences, which puts her over Oracle's EHR (the old Cerner business), its clinical trials business, and its payer tech portfolio, and she's the one overseeing the rollout of Oracle's next-gen, AI-powered EHR. Before Oracle, she ran CMS from 2017 to 2021.

Executive Summary

Coding. Why do we still code? When I asked what she'd change if she were back at CMS, Seema went after medical coding. Coding exists because we used to run on paper and needed a way to translate a doctor's notes into something a reimbursement system could read. With ambient listening, AI can transact the payment straight from the encounter recording. And CMS rules built around codes and documented prior auth submissions could end up dinging providers for doing the efficient thing.

"Maybe we don't even need coding going forward. But CMS ties everything to coding. So even when AI is capable of transacting, issuing the payment, and doing the prior authorization, the system still requires it."

Rev cycle leaders probably care a LOT about this one. (So does every coding vendor with a board meeting next quarter.)

Walled gardens and front doors. Seema, of course, wouldn't name Epic or any competitor in the space. She didn't have to. Her read on the walled garden debate goes back to her CMS days, when innovators kept telling her they had solutions ready to go and no way to plug them into what health systems already ran.

"It seemed like there was always a walled garden, or a front door they had to go through. At the end of the day, our customers need to make the decisions about innovation. It's their system and their patients."

Oracle's EHR boss calls the EHR a point solution. Asked what the EHR looks like once AI is fully baked into care delivery, the head of Oracle Health told me the EHR has always been a point solution, and it'll stay one. Personally I find that framing fascinating as it implies that the EHR itself can be traded out. ‘(Over my dead body’ says every health system CIO who has spent $2B implementing an EHR over 5 years. By the way, you could also call that the sunk cost effect and/or the Judy psy-op. Half joking).

Oracle's real bet is an enterprise AI platform sitting across the EHR, ERP, HR, supply chain, and rev cycle. Coming from the person who inherited the Cerner book of business, that's a hell of a line.

"It's really about the enterprise solution around AI - an AI platform that allows the enterprise to use AI across all of their systems, whether it's ERP, human resources, supply chain, their rev cycle, bringing it all in a coordinated fashion."

System of record, meet system of intelligence. Seema's core thesis is that AI is only as good as the data in front of it, and legacy EHRs were built to keep records.

"Why does AI hallucinate? When it doesn't have access to real-time data. So you need a system that can access enterprise data - not just what's in the EHR, but what's in your supply chain, your financial systems, human resources."

Here's our full conversation, lightly edited for length and clarity. My questions / commentary in bold, Seema’s responses underneath.

AI needs a new spine

I write less about EHRs and interoperability, and given your history, I consider you an expert on both. How would you assess where things stand right now? It feels like there's a ton going on, and it's impossible to write about anything without AI coming up.

AI is only as good as the data it has in front of it. You've got lots of great models being developed right now, including models built specifically for health, and that's all fine and dandy. But at the end of the day, whether we're asking AI to help with coding or clinical decision support, it needs context on what's happening in that real-time scenario. AI is just like our brain. It has to have access.

So we go back to the fundamentals of interoperability. Interoperability is about having the right data at the right time and making sure that data is complete. We've been talking about it for years, but it's becoming even more essential now, because for AI to deliver on its promise, it has to have that access. I'm really proud of the work we did at CMS, and I'm proud of what the administration is doing now, because that's fundamental.

If you look at where healthcare is right now, and you've written about a lot of these problems, AI gives us the capacity to solve longstanding problems. But if you don't have interoperability, that's going to be a deficit.

The second thing I'd say is that everybody's operating off an EHR, and the EHR was built as a system of record. It handles some workflows, but for the most part it's keeping the data. It's not built as a system of intelligence.

I'm hearing that term so much.

And that's what Oracle is trying to do. I don't call it a theory, because we've done this transformation in other industries. For AI to work, you have to solve for it at the infrastructure level. You need interoperability and access to real-time data so those models work. People talk about hallucinations. Why does AI hallucinate? When it doesn't have access to real-time data.

So you need a system that can access enterprise data - not just what's in the EHR, but what's in your supply chain, your financial systems, human resources. Longitudinal data. You have to have an enterprise strategy. That's why we're bringing all of our products onto an AI data platform. First, it brings your data together. Then it lets you match that data with the AI models you have in real time, and it's built on the AI database at a foundational level.

The EHRs that existed in the past weren't built for that. We try to bolt solutions onto old, outdated technology, and that never works over the long term. Just look at the history of technology. Maybe it's okay in the short term if you're doing one thing over here and one thing over there. But when you're talking about many different AI agents working together, it has to be on a platform, and it can't live in one part of the system.

The agent has to work across the EHR, supply chain, ERP, the entire system. It has to retrieve data not just from the EHR, but on the latest clinical treatment. What's the literature saying? What's the insurance company saying? What's CMS saying? You need real-time access to multimodal data, all at once, for AI to be effective.

Walled gardens and front doors

On your panel earlier, you alluded to the walled garden approach versus an open ecosystem. One of your competitors has clearly picked a side. How do you break down that dynamic?

The industry is crying out for an open ecosystem, and I don't say that just because I'm wearing my Oracle hat. I heard it all the time as administrator. Health systems would say, "We have this innovative solution over here." Companies told us they had all kinds of ideas that could move the industry forward, drive better outcomes and access, but they couldn't integrate those solutions with what they had. We need to move away from that.

Healthcare is facing a real crisis. We have a big affordability problem, so we've got to become more efficient. And I don't think the best ideas are going to come from one company. We have to create a platform that allows for innovation.

At Oracle, we're building that open ecosystem. We bring our products onto an AI platform that brings your data together and lets you build your own AI products, buy others, and do it in a coordinated way that's also secure. Cybersecurity is critical here. As we talk about all these different solutions, we have to have that foundation of security.

With Epic facing antitrust scrutiny, do you think their business model stifles innovation?

I won't comment on a competitor, but what I will say is that when I was administrator, I heard from a lot of innovators who were very frustrated. It was hard for them to do business and bring innovative solutions to the marketplace, because it seemed like there was always a walled garden, or a front door they had to go through.

At the end of the day, our customers need to make the decisions about innovation. It's their system and their patients, and they need to have choice about what products they use. We should all be competing to deliver better care, better quality, and better outcomes.

Open vs. walled garden is one axis. How are you differentiating technologically against other EHR competitors?

What's differentiating is that this EHR has been built from the ground up for AI. It's also being built with safety in mind, so we've layered in monitoring, transparency, and human in the loop. It's built for the future era. We're not trying to bolt solutions onto an old, antiquated EHR.

Now, we continue to support Cerner Millennium, and we'll keep doing that while we bring a new product to market. But we don't think you can bolt solutions onto technology from the past and expect it to deliver the outcomes the industry desperately needs.

Narrowing to the EHR specifically: how do you keep a thriving innovation ecosystem while also building AI-native products yourself?

It's an open ecosystem, an open platform. We're going to build the base product: the EHR, the patient portal, rev cycle. But because it sits on a platform, customers can say, "You know what, I actually want to build this one myself," or "I like this product from another vendor," and build it right into their ecosystem.

That level of openness matters because we're all driving toward the same transformation. For Oracle, this is a mission. We have a mission to patients and to healthcare that's higher than competing with another company. We want to drive better outcomes, and if a competitor's product does that, we would certainly not want to prevent our customers from using best-of-breed solutions.

What's left of the EHR

Say AI is fully implemented in healthcare. What's the role of Oracle, and of the EHR, in that future state?

The EHR has always been a point solution, and it will continue to be that. It's really about the enterprise solution around AI - an AI platform that allows the enterprise to use AI across all of their systems, whether it's ERP, human resources, supply chain, their rev cycle, bringing it all in a coordinated fashion.

In an interview earlier today, you said AI's value depends on redesigning workflows rather than layering tools on top. How are you redesigning workflows from the ground up, the same way you're rebuilding the EHR?

At the end of the day, it's humans using this technology, and we always want a human in the loop. So we have to think about how the work changes, because we have incredible demand right now. We have an aging population, people are having trouble getting routine doctor appointments, and that's going to impact quality of care.

There's a real opportunity with technology to redeploy our workforce so they're not doing manual, repetitive work, like arguing with insurance companies. I don't think the insurance companies want that. Payers don't want it. Providers don't want it. How do we automate processes to reduce friction on both sides and reduce administrative costs? That's where the sweet spot is.

Coding's expiration date, and regulating at AI speed

Given your prior role, where do you think CMS is headed? What do you like about what they're doing, and if you were back in the seat with all the power, what would you change?

You have to think of CMS as both regulator and payer. So they've got to think about what they're paying for when it comes to AI, and it's really hard for government to keep up with the private market. AI is moving so fast. You have to make sure the policy is flexible enough to accommodate change.

The sweet spot for CMS is always to stay on the side of outcomes: these are the outcomes we want to achieve. You may use AI, you may not. You've also got the FDA involved on the safety and efficacy of products - that's their jurisdiction. CMS has to think about what it's paying for and what the safeguards are. They have the conditions of participation, which are really about guardrails and safety.

But going back to business process transformation, it's going to change the way we work. So some things CMS requires today may not be reasonable anymore. Let me throw a few examples at you.

Take rev cycle and coding. The reason we have coding is that we used to be on a paper system. The doctor would write their notes, and we needed a way to translate what happened in that interaction into a set of codes tied to reimbursement. With AI and ambient listening, it's possible to take the recording of the encounter and completely transact the payment from it. Maybe we don't even need coding going forward. But CMS ties everything to coding. So even when AI is capable of transacting, issuing the payment, and doing the prior authorization, the system still requires it.

Here's another example. CMS's proposed rule on prior authorization said, essentially, we want to encourage providers to use modern technology in prior authorization, and you have to show that you're using it. Well, the reality is the AI may say, "You don't need to do this. This doesn't require prior authorization." You don't want to ding a provider for not submitting a prior auth because the AI told them they didn't need to. You see what I'm saying?

It's going to change how we work, and CMS is going to have to think about what it's requiring. What it requires now may not be necessary in the future.

So it's constant iteration.

Constantly iterating. And the problem with regulations is they're usually behind where the market is.

Healthcare is conservative by nature, which is a massive mismatch with the pace of AI adoption, especially after last weekend's "we need to slow down" moment from the AI labs.

Which is understandable, and that's why one of the things we did at Oracle was put out our AI governance and safety approach. This is how we develop our AI. Here are the guardrails. Here's all the safety. We did that months ago because we wanted to give our customers certainty about how the AI works and how we monitor it.

How much do regulators need to evolve, given where AI and healthcare are right now?

It's a really tough issue. On one hand, we've got to put patients first and make sure we have guardrails and safety in mind. At the same time, you don't want to inhibit innovation. The world's changing. Other parts of the world are starting to accelerate, not just on clinical care but on clinical research. It's a fine line between accelerating innovation and limiting it, and for regulators, it's finding that sweet spot.

If we think about AI in healthcare, we've been grappling with the same problems for 50 years. If you've been to one health conference, it's kind of the same thing. We're talking about the same problems over and over again, and we haven't been able to solve them. Here's this incredible opportunity to finally solve them.

So there's hope, because right now there's nothing else on the table. Government is getting to the point where it can't afford healthcare. You're seeing premiums go up. The problem's getting worse, not better. AI gives us a solution, but we have to do it in the right way, in a way that's secure and safe for patients.

My quick take

Sure, the open ecosystem pitch is easier to make from the challenger's seat. Google ran some version of it against Apple for a decade (both of which, notably have dealt with their fair share of antitrust shenanigans). But what Seema described from her CMS days, innovators stuck outside a front door, lines up with what vendors keep telling me about UGM season…when one Epic ~~overpromise~~ keynote can reshape an entire product roadmap. So I don't think she's just talking her book.

I'd push back on maybe a couple things. One is the framing of an EHR as a point solution Hallucinations aren't only a data-access problem, and I'd love to see the evals showing an enterprise data platform fixes them. But her broader point holds: an agent that can see the EHR, supply chain, HR, and the payer's rules is going to make better calls than one stuck inside the chart.

The coding stuff was fascinating as well. I don’t know if anyone at CMS has the actual cojones to do it though. Our entire rev cycle apparatus sits on a translation layer we built for paper charts… and a former CMS administrator just told me it might not need to exist. This is the same person who launched Patients over Paperwork at CMS. Is part 2 coming?

Thanks again to Seema & the Oracle team for the time!

WORKWEEK