In June Ashok Chennuru, chief data and AI transformation officer at national payer Elevance Health, wrote an article for NEJM AI in which he described AI as a “foundational design layer, embedded within workflows to support coordination, reduce administrative burden, and improve system coherence.” He recently spoke with Healthcare Innovation about these concepts of connected intelligence and shared data between payers and providers. 

Healthcare Innovation: Your New England Journal of Medicine article states that until now AI deployment has largely been focused on addressing individual tasks rather than operating across systems, and you wrote that the greater value lies in the capacity to support coordination across organizations. I was wondering if you could talk about that in context of Elevance’s Health OS backbone.

Chennuru: In order to truly address the complex problems in healthcare, the real opportunity is working end to end from the time an authorization is submitted by the provider in the doctor's office to the decision we make and the response back to them because there are several moving pieces involved when you look at it with that end-to-end perspective. 

How do we build that trust with providers?  When they submit an authorization, sometimes we don't make a decision because of a lack of complete information. But what if we tell them that ahead of time — before we make a decision vs. going back and forth? Or instead of us denying it,  then they provide more information and then we approve it? We are building an interoperability framework with agentic automation. We are working with one provider on a first-of-its-kind in the industry where we are going to use agents going between a big health system and Elevance. That way we can exchange information seamlessly, based on the trust between both organizations, with clear proof points that are beneficial for both us and the provider vs. doing it in a way where it only benefits one, because that's not sustainable.

If a provider submits a claim, and we deny because of lack of information or other reasons, and then they get the denial back and then submit an appeal, and then we overturn and pay them — when you look at that sequence, it's a cost to both the payer and the provider because whenever we do a denial, that leads to a cost on our side, too,  because we have to manually process it. That leads to cost on the provider side because then they have to use AI to create their appeal, and then we have another team that does grievances and appeals. Meanwhile, while we are doing this, there are a lot of calls from the providers asking, "Hey, what's happening?” So how do we get it right to begin with? We call it payment accuracy — getting it right the first time.

HCI: Elevance has something called Total Member View, allowing both the payers and providers to work from the same clinical picture, How difficult is that to achieve? And is that view embedded in the EHR of the clinicians?

Chennuru: First, let me give you some context behind the Total Member View. It is built on Health OS, which is the platform that aggregates all the medical records and clinical information and admit, discharge, and transfer data we get from the hospitals and the providers, and integrated with our administrative data, which includes claims, pharmacy, behavioral, and social data, too. We call it the Total Member View because it paints a longitudinal view. If a member goes to one particular health system, that health system only captures the services that were provided within their four walls and whatever the patient documents as part of their past medical history. But if a patient had a surgery 15 years back, we would have that from the claim, and since we connect to multiple systems, we focus on the person vs. just a clinical encounter at a particular location. 

HCI: Do you work with health information exchanges too?

Chennuru: Yes, we work with EMR vendors, with the hospitals, and with almost every health information exchange in the country.

You asked if thew Total Member View integrates into the EMR workflow. We have two paths there, and we let the providers decide. Some of them have population health management tools on top of their EMRs. So we give them direct access. We have close to 100,000 providers who log in on a daily basis into TMV directly. Some health systems would prefer that we give them the information, like gaps in care or the relevant information they need, because they want to control how it gets presented to their physicians. So we provide the data, and they integrate it into the provider workflow because they're trying to make it the same user experience for multiple payers, too. In a nutshell we meet the provider where they are — either we let them access our system or we give them the data that's relevant for their needs.

We share the same information with our members directly, too, through our Sydney app, because it's important to show the same information to our member and to the provider. Internally, we have a 3,000-plus care management team that uses the TMV, too. So everybody is leveraging the same data and the insights.

HCI: Has this been in use long enough that you can already see it reducing the friction that you described earlier? 

Chennuru: Absolutely. In certain systems where we have deployed it, within a couple of months we have seen lack-of-information denials drop. Because now we are not asking for the same medical records again and we are also digitizing this, so providers are not faxing us the medical record; we are getting them digitally. Plus, when providers send us 1,000 pages, the real information is usually there in one or two paragraphs. That's where we are leveraging AI to summarize the meaningful content. And if we are doing a denial, we always have a human in the loop, so we are not automating the denial process.

HCI: Could we talk about Elevance’s collaboration with Epic on its Payer Platform to address inpatient concurrent review? I think Elevance is the first payer working on that with them.

Chennuru: Let’s start with why inpatient concurrent review is important. When someone gets admitted to a hospital — with pneumonia, for instance, they could also develop sepsis. Real time decision-making is important so that the hospital can do what is needed to take care of that patient. In the past we used to go back and forth in phone calls or we would log into their Epic system and we don't know how to navigate that complicated Epic system. Now we get that data on a near-real-time basis and surface it to our team of nurses who can approve it right away because minutes matter there.

HCI: Were there issues that you had to work through with Epic to make this work smoothly?

Chennuru: We've been in partnership with Epic for almost eight years now. We are very much involved with the Epic technology and product team. They are dominant in the provider space —  95 of the top 100 health systems are on Epic and the other five want to go to Epic but they don't have the money to go to Epic. We want to help them figure out how to work with payers. We were very instrumental in shaping their product, but at the end of the day, it's all about proof points. If we try to go to the providers without the Epic partnership, then we have to do one provider at a time. Epic brings that scale and trust into the mix.

HCI: After working with Epic, have you piloted it with some healthcare systems?

Chennuru: Yes, with a few of them and we are scaling it. We have connectivity with over 150 Epic systems. Our goal is to scale it to all those systems in the next year or so.

We are looking beyond Epic, too. We are talking to Oracle, because they're pretty dominant with smaller health systems.

HCI: Are other payers looking to do the same thing with the Epic Payer Platform?

Chennuru: Yes, this is not proprietary to us. This is something that the whole industry benefits from. I would say national payers are at the forefront along with us. Our push is now how do we get it to all the Blues as well? 

HCI: Could you talk about Elevance's electronic prior authorization program as well? I understand you’re working with more than 30 health systems on that. There are some deadlines that the federal government has set that have gotten everybody working toward this.

Chennuru: One of the mandates is CMS-0057. We actually are live and are testing with all our ecosystem partners. We are one of the first to actually build it, and it requires extensive testing. We'll be ready for the deadline, which is January 2027. I spend a lot of time with our CMS partners talking about this from the payer point of view.

HCI: Are there other areas of Elevance's business where AI is showing promise? Other areas where agents can help?

Chennuru: Member experience is a big opportunity for us with AI. Proactively engaging the member in finding the right doctor, making sure that we do the last-mile connectivity, which is the appointment scheduling. If they log into MyChart, we have a partnership with Epic where we send the ID card and coverage information both to their care provider and to the member. The provider sees the coverage information and ID card within their workflow. We launched it about six months back. We exchanged 2.6 million ID cards in six months.