theCUBE's Coverage of Google Cloud at HIMSS25 | Ramaswamy Rajagopal, Cognizant & Aashima Gupta, Google Cloud

Clips
News
More from theCUBE's Coverage of Google Cloud at HIMSS25

Ramaswamy Rajagopal

AVP, Healthcare Strategy

Cognizant

Aashima Gupta

Director, Healthcare Solutions

Google Cloud

AI’s future in healthcare: A balancing act of innovation and responsibility

AI in healthcare delivery is already showing considerable promise. However, with higher stakes than ever, concerns mount over data privacy, regulatory compliance, explainability and the human element in artificial intelligence-driven decision-making.Google Cloud’s Aashima Gupta and Cognizant’s Ramaswamy Rajagopal talk with theCUBE about how AI in healthcare is transforming patient care, enhancing data security and reducing administrative burdens

play_circle_outline Navigating Data Privacy Challenges in Healthcare with Regulation, AI, and Fairness for Trust with Stakeholders

play_circle_outline Navigating Data Sharing, Privacy, and HIPAA: AI Best Practices for Healthcare Organizations

play_circle_outline Empathy and human in the loop in AI decision-making processes.

play_circle_outline Role of AI in reducing administrative burden and burnout in healthcare.

Info
Transcript

Ramaswamy Rajagopal, Cognizant & Aashima Gupta, Google Cloud

Ramaswamy Rajagopal

AVP, Healthcare Strategy Cognizant

Aashima Gupta

Director, Healthcare Solutions Google Cloud

Ramaswamy Rajagopal, ADP Healthcare Strategy at Cognizant, and Aashima Gupta, director of Healthcare Solutions at Google Cloud, join theCUBE at HIMSS 2025 to discuss the future of healthcare innovation and Agentic AI. The conversation, led by theCUBE Research analysts, explores data privacy and security, as well as the complexities of building trust across various healthcare stakeholders. Gupta and Rajagopal share insights into the evolving landscape of AI in healthcare.

Key takeaways from the discussion include the importance of accurate data in AI ... Read more

explore Keep Exploring

What are the key themes to consider when building and deploying AI solutions in healthcare? add

What is the importance of self-regulation in healthcare organizations and how does it relate to the need for regulation in certain use cases, such as AI in clinical diagnostics? add

What is the role of the human in the loop when it comes to bringing empathy to life in AI solutions? add

What is the potential impact of using AI in the administrative tasks of an organization, particularly in the healthcare sector? add

bolt Powered by CUBE AI

Ramaswamy Rajagopal, Cognizant & Aashima Gupta, Google Cloud

search

Rebecca Knight

>> Welcome back everyone to theCUBE's coverage of HIMSS 2025 here at The Venetian in Las Vegas. I'm your host, Rebecca Knight. We are here with Google and its ecosystem talking about the future of healthcare innovation and the future of Agentic AI. With that, I have two great guests for this next segment. We have Aashima Gupta. She is the director of Healthcare Solutions at Google Cloud. Thank you so much for returning to the show. Aashima.

Aashima Gupta

>> Glad to be here.

Rebecca Knight

>> And another CUBE veteran. We have Ramaswamy Rajagopal, ADP Healthcare Strategy at Cognizant. Thank you so much. A CUBE veteran.

Ramaswamy Rajagopal

>> Great to be here. Great to be here.

Rebecca Knight

>> So Ramaswamy, I'm going to start with you. Let's dig in and talk about efforts around data privacy and security to transform healthcare delivery. Why don't you begin offering a broad overview of some of the issues and the challenges in terms of complexity, accuracy, explainability? There are a lot of them that come between building trust across stakeholders.

Ramaswamy Rajagopal

>> So at Cognizant we've been building a lot of AI solutions over the last 12 to 18 months and based on our experience, we see four or five major themes which come to the top of the table to build and deploy a solution that can be productionized and be able to roll out to members and patients and things like that. Because think about healthcare, it impacts people's life. You don't want to get things wrong. So if you look at starting with data, it's important for any of these solutions to have the right set of data to support your models or training your models and things like that. If you don't get your data right, you get into a lot of complexity when you start testing it and rolling it out. The second most important thing is regulation. The regulation continues to evolve and we see how do we prepare yourself for today, but also how do we basically prepare for tomorrow? You have to continue to be able to challenge yourself to say what's going to come in the pipeline and how do you adjust your model for the future? So because we see that as a big... Everybody's learning a little bit on that regulation aspect of it. The third is explainability. This is a big part of it, especially in healthcare, when you are rolling out solution, you want to be able to build trust into the ecosystem and one of the big, big build trust into the ecosystem is to have an explainable solution that you can roll out to your member or your patient to say how the model came to the answer that you came out with. And a lot of it is how do you capture the pathway and got to the results and be able to showcase that. The fourth is healthcare inherently focuses on certain data sets, but fairness and bias is something which you really need to concentrate a lot because beyond taking the straight approach to solving a problem, you have to look at the different demographics that play into it. SDOH is a big part of it. How do you get that built into your model as well? And last but not the least, we know that these solutions are going to evolve and so we need to be prepared to support that.

Rebecca Knight

>> So Aashima, Ramaswamy has just laid out a slew of some of the major challenges. I want to dig in with you about regulation and requirements because a lot of this, all of healthcare is about sensitive patient data, about some of the most intimate problems and issues that people are facing. How can organizations balance this data sharing needs, also the strict privacy requirements and HIPAA regulations? What are some of your best pieces of advice here?

Aashima Gupta

>> So I would say from the regulation perspective, healthcare is a regulated industry, the sensitive patient data at stake, and this is critical to get that part right. But I would say, I'll go back to what I shared earlier, having that self-regulation is important for organizations. Before you even actually look for a new regulation to come in, how many organizations have best practices, AI principles, how are they going to secure the data? How are they going to be privacy preserving? And I'm seeing a lot more best practices sharing in the industry. For example, we worked with... I'm the industry co-chair of Coalition of Health AI and we have built a constant color model card. It's like when you have a nutrition label in a yogurt, you see the ingredients on it. On your data set, what does the model card, what are the data? What is the difference, diversity of that data set? You need to be much more open, transparent about that and thousands of health systems across the globe are adopting that as a standard. But that's just one approach. In answering your question, the enterprise grade privacy security is table stakes to build that trust. And then from the regulation perspective, self-regulation is enough. It's a necessity. It's not enough. Depending on the use cases, if you are building AI that touches patient, clinical diagnostics, we believe regulation will be very important. And that's our stance that AI is too important to not be regulated.

Ramaswamy Rajagopal

>> So that's it-

Aashima Gupta

>> It is a changing environment as well too.

Ramaswamy Rajagopal

>> Exactly. That's a great point she brought up about self-regulation, right? Don't wait for somebody to tell you what to do because one of the key aspect is how do you make this work is through effective change management and communication and things like that. You don't want to be rejecting a prior authorization for some patient and then it bounces back into of a problem by itself. But if you use your self-regulatory mechanism to identify these things early enough and be able to explain that solution and say, how do you determine that? We have been working with pediatric oncology in the University of Yale and one of the important challenges that they have is that when you go through these treatment for oncology, there is no straight path. There are multiple paths for the solution. And when the physician recommends a certain... Make a decision to recommend certain treatment, when it gets to a health plan, they look at it from a different lens. So when you're building these tool sets to support those decision making, you need to be at the end of the day, be able to get a little bit of an empathetic view as well especially in this case, a pediatric patient. So you don't want to be making decisions which are going to be impact. So that's why what Aashima brought out was about self-regulation, beyond regulation to be able to say, how can I make these decision making tool sets that much more strong that we can roll out and be confident that it works?

Rebecca Knight

>> Well, as you said, healthcare affects people's lives, but it often affects them at the worst point in their lives after they've gotten a really hard diagnosis or their child has been diagnosed with a pediatric cancer. And so it makes the explainability issue all the more pertinent because people need to understand how was this diagnosis made? What kind of care can I expect to receive? I mean there really is this black box issue in terms of AI. Can you talk a little bit about how we can make Gen AI models more explainable and transparent?

Ramaswamy Rajagopal

>> It's all around the data sets that you use for training. There is always a continuous improvement process to build on, but there are a lot of things that you can roll out today that keep the regulation aspect of it out of it. In this particular example which I talk about, they are going to use the tool set to be able to explain to the patient and the parents in this case, what are the different effects of this particular diseases. First educating about the disease itself. And the second is basically showing them what are the different care pathways that they can take and literally walking them through how the care will impact during the course of the treatment. So when you look at it that way, Gen AI becomes a very powerful tool to support that But it requires a lot of training and lot of real content and that's why data becomes a really important aspect to support that.

Aashima Gupta

>> I think just to interject, as is being said, and this use case is not clinical diagnostics, right? We are not there yet where Gen AI diagnosing a symptom or disease. It is much more about being an assistant, being a collaborator, being a helper, being an assistant to a patient or a parent who's navigating that care journey today. We all know the drill. We diagnose with something., I need to see a specialist. My first thing is which insurance, my network, out of network, how far are they? That navigation today is fragmented. It's super hard. I think that's where we are at healthcare. We have a lot of this initial problems to solve before we even actually go to clinical diagnostics and huge space to innovate there.

Rebecca Knight

>> Before we get to the...

Ramaswamy Rajagopal

>> Diagnosis, we are always... All of our solution today to Aashima's point, we are doing human in the loop, but there a lot of assist from these tool sets to support that.

Rebecca Knight

>> One of the things you talked about, Ramaswamy is the role of empathy in all of this. Can you talk a little bit about how the human in the loop and the role of making sure that there is empathy put into all of these decisions?

Ramaswamy Rajagopal

>> So the human in the loop ultimately brings the empathy to life. But the tool set itself is enabling that, showing because there is so much training that we can do to show what are the different, we can build that feeling into your AI solution by appropriate level of training. So one of the good things about these tool set is that it can get some of those aspects trained as well. So the empathy into the tool set. But ultimate what it gets to the patient is done through a human. So we are not letting out of the loop.

Aashima Gupta

>> Just to add to that, I think human in the loop, we believe in that approach because that's why you have extra certifies. It's assisting a clinician or a nurse. But let's be careful, human in the loop should not be a cop-out, meaning we cannot be adding more burden to the human who's now I have to check AI work. That's where hallucination, grounding and having the data strategy where your answers are grounded in your enterprise truth and there's a way to do that. And that's where the enterprise data set and data strategy comes in. Because to me, human in the loop should not create more work. And we have seen that many times these point solutions and that human in the loop should not become that. I'm adding more burden now. You check AI's work.

Rebecca Knight

>> Particularly at a time when we are facing a shortage of workers in the healthcare industry. And we are coming up on the anniversary of the start of the COVID pandemic out here in the United States at least. So we know that their burnout remains a big problem in this industry.

Aashima Gupta

>> 10 million healthcare workers shortage by 2030. It's real. It's here.

Rebecca Knight

>> Absolutely. And AI can perhaps...

Ramaswamy Rajagopal

>> Hugely impact that because when Aashima talks about assist, that's where a lot of these assists comes into play basically to take out a lot of the manually intensive work that they go through on a day-to-day basis. We see a powerful use in administrative area taking out efficiencies across. One of our CMO talks about it all the time. After he gets back from work, he spends about two to three hours just doing the administrative things. You wouldn't believe that there's $60 billion of money which is spent every year on clinical documentation just to support compliance. And so AI is a really powerful tool to enable all of that, whether it is summarization, chart summarization, all of those kinds of things. And even things around the peer provided collaboration, sending the documentation required for approvals and things like that. And so one day you could probably see a touchless claims processing. You go to a doctor, set up your schedule, pay a visit, they know which insurance you have. And basically once you are done with your visits, their documentation submitted to your health plan, then it gets approved and their payment is processed. Right? So everything, if you can make that whole loop seamless, the patient doesn't have the burden to basically carry all of these different things to the provider to show what they have.

Aashima Gupta

>> I think that term in healthcare called pajama time.

Ramaswamy Rajagopal

>> Yeah, exactly. That's a great point.

Aashima Gupta

>> 60 billion, like pajama time, people go home and I think Google also did a survey last year, 28 hours a week on documentation.

Ramaswamy Rajagopal

>> Right. Yes.

Rebecca Knight

>> Sitting in their pajamas on the couch when they should be enjoying themselves. But in fact, they're going through charts.

Ramaswamy Rajagopal

>> That's right.

Aashima Gupta

>> And charts and documentation, making sure the coding is right so then they can get paid. So all that to say that 28 hours on a weekly basis adds up. And when you talk about shortage, that's very helpful. Assistant taking care of... Taking healthcare. What is a burnout? It's reading and writing documentation over and over and over again. So finding out those roads, repetitive tasks, getting Gen AI to assist with that, I think that's the opportunity we all have.

Rebecca Knight

>> And make their jobs much more satisfying. Ramaswamy, I'm going to give you the final word here. If you could advise healthcare leaders one critical step for AI implementation, what's your best advice here?

Ramaswamy Rajagopal

>> I think they have to pick... I offer two advices. One is pick a lane, which basically that they can experiment and basically start using these tool sets and understanding how it is going to impact. And then second is basically see how you can marry up to the outcomes and basically build it more on a trustable environment and build that and then roll it out. Don't wait for things to happen. Get moving. I think there's a lot can be done with it.

Rebecca Knight

>> Excellent. Ramaswamy, Aashima, thank you both so much for coming on theCUBE.

Ramaswamy Rajagopal

>> Thank you. Thank you.

Aashima Gupta

>> Thank you.

Rebecca Knight

>> And thank you so much for tuning into this segment of theCUBE's coverage of HIMSS 2025. I'm Rebecca Knight. You're watching theCUBE.