Loading summary
A
Hello everyone and welcome to Becker's Healthcare Podcast. I'm Scott King, thrilled today to be joined by two very special guests. First, we'll start with Rick Peng, Digital Ventures lead in Memorial Sloan Kettering Cancer Center's Office of Entrepreneurship and Commercialization. And we also have Natalia Somerville, Director of Decision Intelligence at Memorial Sloan Kettering Cancer Center. Thank you both so much for joining. I know we have a lot planned for a great discussion on all the innovative things you're up to over at Sloan Kettering. How you both doing?
B
Doing good. Very excited to be here. Thank you so much for having us and yeah, definitely very happy to share some of our experiences.
A
Yes, no, thanks so much to both of you for joining and, and we'll go ahead and get started here with this important discussion now. Now, Rick, I'll get started with you. I know you know, MSK is, is often seen as a leader in applying AI in clinical settings. How is AI currently being used across the system and where do you, where are you seeing the greatest impact in regards to patients today? Yeah, so AI is definitely being applied across the board here at msk. It's being applied in not only clinical care, but also in the research space. It's in the clinical spaces. It's used in use cases such as assisting clinicians in their visits with patients. Key examples, our partnership with companies like Abridge that help us manage transcription documentation of patient visits in the research space. We do a lot of internal innovation as far as AI development models that can predict responses to therapy and that assist in the development and discovery of new therapeutics and oncology. We also partner with external companies in this space as well to co develop new AI that can be applied in these situations. And of course we have a large network of pharma partnerships that we that can leverage these technologies to accelerate the development and bring to market new drugs that serve unmet needs in cancer. And of course on top of that, we also apply AI as tools across the enterprise to help staff at MSK, outside of just clinicians and researchers, to do their jobs as effectively as possible as well. So really across the board. Yeah, I appreciate the details there on all the AI usages, Rick. Now Natalia, I know you co create MSK's AI governance operating model, which is obviously a huge accomplishment. But what problem was MSK trying to solve when you were developing the framework?
B
Yeah, so a couple of years ago here at msk, our leadership created what was called the AI Task Force, where kind of given all the movement and progress happening AI and healthcare, our leadership wanted to understand what are the important aspects and what should we be looking for, what are the valuable use cases. So one of the work groups was governance. And that work group was led by our Chief Informatics Officer, Dr. Pit Stetson. At the time and, and during that group I was supporting the lead and we had conversations with important players, let's say within the hospital, like legal and clinicians who had been working the ethics and technologists. And that's where there was obviously kind of an understanding between everybody that we are starting to deploy more and more AI and we need to be able to make sure that it's done safely and also that we can scale it up because we totally understand that if maybe a couple years there was a couple models deployed, this is growing exponentially now. So we want to be able to do that responsibly and support that innovation and that scaling up of AI technologists.
A
Absolutely. And Rick, I want to ask you about the AI governance as well. You know, in a high stakes environment like cancer care, why is governance, it's not optional when it comes to AI adoption. And why is that? Yeah, I mean, I think there's obvious, you know, ramifications on, you know, how patients receive care, how research is conducted, not only in terms of, like the clinical impact on patients, but also, you know, everything from, you know, ethics and scientific integrity as far as all that work that we do. And I, and I think, you know, broadly, like, the need for governance in AI is particularly important as we think about, you know, organizations like MSK are very large. There's a lot of different stakeholders involved, including external collaborators. And I think AI requires connectivity across all these different groups, whether they're at the points of applying the AI, at the points of developing the AI, at the points of supplying the inputs to AI, including data, where we really need all these pieces to be working in coordination with each other to make sure that what comes out the other end is created in a way that optimizes its intended performance as well as is being applied responsibly across all the different groups that need to be using it. Natalia, what qualities, I'm sorry, should health systems look for an AI governance partner to ensure responsible adoption while still supporting some innovation?
B
Yeah, I think it's definitely flexibility because as we have learned, and you just mentioned, we have to be able to balance those two goals, be able to deploy AI responsibly, but also making sure that we are not seen and we are not blocking any, any innovation performed by our clinicians. Our clinicians do amazing research and that is partially why our institution is so well known, because of the research they do. So we want to make sure to encourage that research. So the way we partner and the way we see these processes is, is making sure that there are different paths, that if there is an AI model or AI system, that the governance committee is looking at a time that we have different paths, that we are flexible, understanding that the research initiatives may have a faster path. They are working towards proof of concepts and developing the science versus tools that may go into workflow, workflows right away, let's say predicting mortality. And we want to deploy that within two months, then we definitely want to have a different path, much more focused on reviewing the details. So yeah, I believe that that flexibility on adjusting the processes and our AI life cycles across the different types of implementation is key.
A
Natalia, how does MSK's AI governance operating model help ensure that AI tools are safe and accountable and aligned with clinical standards?
B
Yeah, that's a great question. So over the last, I want to say two years, we developed these processes and an AI lifecycle. So this is a tool that kind of grounded, grounds the AI research and deployment that happens. And AI lifecycle is the sequence of steps that you follow, such as you have an idea, you have to validate that the idea has value within the organization and then design, deployment, testing all those different steps. I do want to call out that actually two years ago, our life cycle was our original life cycle. We did use Duke Health systems life cycle. They are very advanced in this area and then we modified it. What makes sense for us, what I was mentioning is this life cycle allows us to create steps depending on what stage the AI system is. So for example, if the AI system that we are looking into or governing at the moment is in pilot mode, then there are certain requirements that we ask. We ask for the pilot plan to be very explicit and presented to the committee. We're asking for a monitoring plan in advance of the deployment versus if the AI system is earlier, let's say they're just kind of finished design and going into development. There's other sorts of questions that we ask. So depending on the life cycle, we have these different touch points with the model owners and model sponsors to make sure that it goes through in a safe way and we're able to suggest any modifications as early as possible if needed.
A
And Rick, what risks do health systems face if they deploy AI tools without a clear strategy or framework in place? Yeah, so I think like for health systems, if AI isn't deployed in a way that's strategically directed. I think there's a couple levels of risks we. One is an operational risk of impacting the adoption of these technologies. If there isn't a clear idea of what workflows or what stakeholders need to pick this up and build it into practice, it can really limit the extent to which the AI is being applied in the first place. I think ultimately longer term it could impact the ROI of these types of solutions, especially if there's significant investment in them, whether, whether it's through purchasing external solutions or investing in internal development of capabilities. If there isn't the right adoption or if it's not being applied in the most impactful use cases, reduce your roi, which I think can not only hamper the impact of the particular AI solution in question, but also just AI across the board as well. And one special consideration for organizations like MSK as well is like msk, definitely it's point of pride that we are one of the leading cancer centers in the US and in the world. And so I think application of AI also needs to be done in a way that fully takes advantage of the differentiated expertise that MSK has. I think a lot of times where that comes into play, play especially is in terms of using AI solutions, ensuring that there's the appropriate level of governance, but also input the development of those AI tools, whether it's through leveraging MSK's unique know how expertise of its clinical and research experts to really to, to really, you know, infuse the AI solutions that we do apply with the, with the sort of secret sauce at MSK so that, you know, when applied, it really reflects the level of quality that MSK has come to be known for in the world of cancer care and research. Yeah, thanks so much, Ray. Maybe diving into that, that secret sauce, Natalia, that the process with AI over at msk, you know, you know, even after tools are approved and deployed, the work's not done right. So why is continuous monitoring of AI systems essential?
B
Yeah, that is, it's very important, as you mentioned. And the way I talk about monitoring is actually twofold. One is monitoring the safety. Has there been any adverse events, for example? Because as Rick was mentioning, there's several aspects that could potentially go wrong. So we want to be able to monitor any AI adverse events as well as kind of the technical model itself. Is there any drift? For example, if there is a pattern change in the data, then we need to be able to acknowledge that in the model if it doesn't do it automatically. So depending on the type of model, we need to be able to see that every quarter, every month. It really depends on how the model is structured. At the same time, the second part that is also very important, that sometimes it's talked about less actually than the technical monitoring is precisely what Rick was mentioning. We also want to make sure that the tool is actually being used and beyond being used, that the tool is bringing the expected impact. If we are deploying an AI tool to support, for example, discharges, the nurses with their discharge efforts, and the original goal was to reduce the patient length of stay such that they can go home as soon as possible and enjoy kind of more comfort settings, then we want to see that the length of stay is actually decreasing. So we want to be able to see those metrics that we are targeting as well as usage. We don't want to be maintaining tools that nobody's using. So those are the two aspects and they are typically monitored differently and then kind of different platform and visualizations, but both are key to be able to say that the AI is being used properly.
A
And Rick, government regulations always, obviously a big topic in healthcare, and it doesn't escape AI. So with AI regulations now evolving at the state level, how does an organization like MSK streamline compliance without slowing progress or innovation? So I think a big part of that is creating visibility across all different stakeholder groups. Part of that is on an operational level, the creation of the appropriate governance committees with multidisciplinary, multifunctional representation, which Natalia and her colleagues have really spearheaded over the past few years here at msk. A part of that as well is just facilitated through technology solutions that enable us to do this on an efficient basis. And so having the tools for these individuals within these committees to effectively communicate, have the write information at their fingertips so that when there need to be conversations or decisions around the, these AI solutions, that that's handled. And I think like at msk, I'm not directly embedded in these functions, but we also have very robust functions around regulatory and compliance. And I think a big part of it is MSK overall, you know, prides ourselves in, you know, staying abreast of all the latest developments, not in all the regulatory and legal spaces that impact the work that we do. And so our processes are, and our stakeholders and those functions are ever evolving to meet the latest requirements and make sure that we stay not only up to speed, but also think ahead in terms of where we need to be, not just this year, but also next year.
B
Yeah, and if I, if, if that's okay, I can jump in as well. As definitely as Rick is mentioning, the. The other aspect is also kind of being involved with and in communication with other organizations like us that are also working and implementing their governance processes. And part of chai, for example. MSK is part of chai. And that is a way that we try to keep, keep present in kind of the latest developments and as well from the government or legal aspects, any new technology developments. I do believe that it's kind of important that all these organizations that we are sharing with each other and kind of progressing together in this way.
A
Yeah. As you share what you've learned through this experience, Natalia, like what lessons do you, do you think from MSK's approach to AI governance, do you think can be applied to other hospitals or healthcare organizations?
B
Yeah, definitely. I love this question because one thing that is often also talked about is actually how operationally heavy is to manage the governance process. Because as I mentioned at the beginning, the AI kind of the AI implementations and systems and development is growing exponentially. But if we want to have a touch base with every model, every system that is medium to high risk, there's a lot of operational tasks that need to be done. You know, kind of check with the model owner, have them register their systems and verify their pilots and perform risk assessments. And so it's actually operationally very, very heavy. And so partnering with somebody who can help with that, those sort of operations can be very, very helpful. So for organizations that are beginning to look into this AI governance systems, I will definitely recommend two things. One is take a look at what has been done already. Many, many of us have published some of our systems. We are also always happy to talk about that kind of implement some of the processes and guidance that other organizations have done already and also look at platforms that can help with that operationalization of the governance that is typically bypassed at the beginning and that can really delay any sort of meaningful scaling that you can do with your governance process.
A
Well, Rick and Natalia, thank you so much for joining the podcast and just for a great discussion on MSK's, you know, innovative use of, of AI. I look forward to working with you both again soon.
B
Thank you so much for having us.
A
Thank you so much, Scott. It's great to be here.
Podcast: Becker’s Healthcare Podcast
Date: March 27, 2026
Guests: Rick Peng (Digital Ventures Lead, MSK Office of Entrepreneurship and Commercialization), Natalia Somerville (Director of Decision Intelligence, MSK)
Host: Scott King
This episode explores how Memorial Sloan Kettering Cancer Center (MSK) is leveraging artificial intelligence (AI) across clinical care and research, with an emphasis on scaling AI innovation responsibly and safely within a high-stakes cancer care environment. Rick Peng and Natalia Somerville delve into real-world AI applications at MSK, the development and implementation of a robust AI governance operating model, the importance of flexibility and continuous monitoring, and the operational realities of AI governance. The discussion provides actionable insights for healthcare organizations looking to drive responsible AI adoption without stifling innovation.
This episode offered an in-depth look into MSK’s strategic, flexible, and operationally rigorous approach to AI in cancer care. The conversation illuminated not only the benefits and challenges of AI adoption in healthcare but also equipped listeners with concrete lessons for governing AI responsibly at scale—emphasizing the need for tailored frameworks, ongoing monitoring, regulatory foresight, and a strong commitment to collaboration and operational support.