Q: Hi Keith. I’m Rohit Mahajan, CEO and Managing Partner of Damo Consulting and BigRio, and also the host of the Big Unlock podcast. Welcome to the Big Unlock podcast. I’d love for you to introduce yourself, Keith—share your background, your role, and what motivates you on a daily basis.  

Keith: That sounds great. I’m Keith Morse. I am a pediatric hospitalist, which means I’m a physician who takes care of children admitted to Lucille Packard, Stanford Medicine’s Children’s Hospital. I’m a clinical informaticist, focused on studying and optimizing the use of technology and data systems for care delivery within a health system, specifically within informatics. My role is Medical Director for Enterprise AI, where I lead the team deploying and evaluating AI and large language models within care operations at the children’s hospital. I’m also an educator. I serve as the Program Director for our Clinical Informatics Fellowship Program, a two-year training program for physicians who have completed their specialty training and want to gain expertise in clinical informatics. 

The program prepares them for roles as CMIOs, Chief AI Officers, or positions in industry. Finally, I’m a researcher. I conduct research in AI and also serve as the Co-Site PI for Stanford’s participation in PEDSnet. PEDSnet is an 11-institution EHR data-sharing consortium that supports large-scale investigations to improve pediatric health.

Q: That’s awesome. I was reading about your background, and as we discussed previously, you did your MBA in healthcare administration and then chose to become a physician.
What motivated you in that direction? Do you still practice and see patients, and how does it work within the Stanford Medicine health system?

Keith: Certainly. I studied Econ and business as an undergrad, and where I was, they offered a combined MBA program. You add on an extra year to your undergrad training and you can start to get a sense of MBA-type training courses. I was doing my pre-med courses then as well, and I had a couple of years between finishing the MBA and starting med school. The job that I had was writing SaaS code for a consulting company, analyzing Medicare and Medicaid data for federal and state government agencies. 

That was essentially a data scientist role before the term “data scientist” was commonly used. And I loved it — understanding what data you have, how you can use it to answer questions, how you summarize it, and how you present it back to the requester. Those are such core things. I was in medical school in Philadelphia at Thomas Jefferson and then did a pediatrics residency in Phoenix, Arizona. During both of those times, I got involved in a couple of research and operational projects where I served as the data scientist. Writing R code, but also getting a sense of, instead of just being a data consumer as a data scientist, I was also a data producer. 

When you start delivering care, you are the person standing at the bedside, talking to the patient, making decisions, and also writing a note, ordering labs, and ordering follow-up. All of that is the beginning of the types of data that eventually get billed and trickle into databases that get used. In residency, I started working with our CMIO to analyze the data that was available through our EHR and really started enjoying it. 

Then I joined Stanford for their Clinical Informatics Fellowship and have been on faculty since then. We are fortunate here at Stanford Children’s to have a forward-looking CMIO, Dr. Natalie Pageler, who supported both myself and a small team to start building out our organization’s capabilities in operational AI back in 2020. We have been working at this for the last five years or so. Obviously, when ChatGPT hit, we got a whole lot busier, but it wasn’t the beginning for us. The processes we have in place now are built on that early investment.

Q: That’s pretty early to start in 2020. That’s a solid five years of experience with various kinds of AI implementations. How do you approach the business problem? Do you have specific use cases, and what do you do with them? How do you decide where to put your energy, time, and money?

Keith: I’ll answer this in three parts. First, I’ll talk about our process for identifying use cases, and then give a brief overview of two use cases that are relatively mature. First off, I love talking about use cases. In some ways, it’s a mythical concept. When operational AI folks get together at conferences or meetings, people sit around and whisper, “What are your use cases?” The reason is it’s such an important topic because we are, in essence, asking: In what ways have you found that AI is valuable? Where is the juice actually worth the squeeze? It’s easy to have research papers or proof-of-concept pilot projects showing that AI is theoretically useful. But when you have to make it work for an enterprise indefinitely, it’s a much different problem to solve.

We think about where AI can bring value without it costing $50 million or taking five years to implement. These are the types of considerations we focus on. This is actually a really hard question because three separate areas of expertise must align to arrive at good use cases.

The first is understanding AI technology in isolation. What is a large language model? What is a deep neural net? What is logistic regression? What can reasonably be expected of that technology? 

The second is understanding what infrastructure is available at my organization to support those tools.
It doesn’t help to have a sophisticated AI tool if you don’t have the data available, the compute power, or if the data isn’t available at the needed cadence for the algorithm. Another tricky part is that AI infrastructure is invisible. You can’t walk into a room and see where the AI lives. It exists in the ether. You have to be plugged into the organization’s IT structure to understand your current infrastructure. And it’s always evolving. We’re growing our infrastructure, making investments. It’s different now than it was two years ago, and it will be very different five years from now. 

The third and by far most important — and the hardest — is that workflow expertise does not solve nebulous, non-specific problems. It must solve a specific problem for a specific human, in a specific job, in a specific part of their workflow. We summarize all of that by saying “workflow,” where in a person’s workflow, AI can potentially be useful. The challenge is that healthcare is a diverse, complicated entity. My health system is different from the health system down the street, different from those in Cincinnati or Texas. 

Even within my organization, there are so many different workflows that no one person understands all of them. Our process for learning about these workflows — and this is something my team spends a lot of time doing — involves talking to people within our organization to help them tell us what problems exist and how AI could potentially help.
You would think that’s easy. You might think you can just talk to somebody and figure it out.
It turns out to be surprisingly difficult. The reason is this: if you imagine a standard organizational hierarchy — with director, supervisor, or executive oversight at the top, then managers, and then boots-on-the-ground staff — you find some patterns. This could be in Revenue Cycle, Supply Chain, Sanitation, or any other department. 

Usually, when you talk to the more senior folks, they are very excited about what AI can provide: speed, efficiency, safety, uniformity. They are generally onboard. Being a manager is different than being a boots-on-the-ground employee. Leadership often doesn’t understand what happens on a day-to-day basis in their teams. That’s not a dig on leadership; it’s just a fundamentally different job. People who oversee big teams — it’s too much to know. You can’t look to leadership to tell you where the problems are. Also, leadership is not the group whose jobs will be directly supported by AI. It’s going to be the people working on the ground. You have to talk to the folks whose workflows are potentially going to be impacted by AI to understand exactly what their workflows are and where AI could be helpful. Usually, the way we do this is to start at the top and work our way down. We get buy-in from leadership, then get passed to middle managers. They might suggest three or four areas that are worth exploring. We then meet with each of those teams and ask specifically what it is they do, what data they look at, what problems they encounter, and how AI could potentially support their work. Our hit rate in those meetings is relatively low. Most problems are not solved by sophisticated AI. There could be other, simpler solutions that work just as well. Often, our biggest takeaway is directing teams to existing data and reporting tools that can solve their problems without the need for advanced AI. But through this process, we do identify good use cases, and that informs our future efforts. 

The main takeaway is that no one outside your organization can credibly provide a guaranteed use case because they don’t know your AI infrastructure or your specific workflows. I don’t even know all the specific workflows within my own organization, so how could a third party or a hospital elsewhere know? There is a long-term role for internally identifying use cases, because they aren’t easily transferable from other institutions. One area where this is starting to shift is when AI is embedded within your electronic health record (EHR).

For example, our EHR is Epic. We use some tools they provide, like drafting responses to patient messages. This is a use case gaining traction nationally because Epic controls or manages all three areas I mentioned earlier: First, Epic understands that drafting patient messages is within the capabilities of a large language model. Second, all the infrastructure needed to use this drafting tool exists within Epic — no extra resources are needed. Third, the workflow for responding to patient messages is entirely within Epic, meaning they have a good understanding of the process. If the workflow required multiple steps outside the system — reading something in Epic, going to another machine, speaking to a patient, and coming back — it would introduce complexity and variability. But responding to a patient message is done entirely within Epic, so the workflow remains visible and consistent. That’s why use cases like this are more broadly transferable — because all the necessary components are self-contained.

Q:  Could you please share with the audience your experience with some of the large language models and where you have been successful in implementing them?

Keith: We have a couple of implementations, and we tend to publish most of the work we do, so a lot of it is available in the literature. One of our major use cases is using machine learning to help identify confidential content in teen notes. There are laws in the state of California and in most other states that protect teens seeking care for certain types of sensitive issues like mental health, sexual health, and substance use. In California, there are explicit laws that prohibit providers who care for adolescents from informing the teens’ parents that they are receiving this care. The reason for these laws is that there is strong evidence showing teens are more likely to seek care for sensitive issues if they are assured confidentiality. The challenge is that federal rules also require health systems to make patient records available to teens and their parents.

The 21st Century Cures Act requires that health information be available without undue effort. Basically, this means parents must be able to access and review records, usually through a patient portal within the health system. Anyone who takes care of adolescents faces a challenge: the federal government says you have to share all of the patient’s notes with the patient and their parents, but state laws say you can’t share portions related to three sensitive topics — mental health, sexual health, and substance use. We address this with what we call a “confidential note type,” where providers can document sensitive information. These confidential notes can be excluded from what’s shared with parents, and that system works well. Confidential information in these notes that we are theoretically sharing with the parents and families. This is a great application for large language models because it essentially processes large volumes of text to identify certain definable concepts. This is one of the projects we worked on before large language models came out. We developed a bespoke NLP model to help identify these concepts. We have replicated that analysis using large language models, and the way that we use it now is as a retrospective audit and feedback mechanism. We process the notes of different divisions, and then we can see who in the division is documenting in a way that’s potentially problematic. We can bring that back to the individual and also the leadership and say, “Hey, just FYI, we noticed that your notes contain a lot of information that would run afoul of the California state privacy laws.” 

We can then help providers improve their documentation. Often, what it is there is some sort of smart phrase or automatic pulling in of patient information that’s considered confidential. We just sort of edit or help update somebody’s note template, and the problem goes away . It’s those types of nudges that we think are helpful and really get at what we want, which is to be able to share patient information with the teens and their parents. We want this stuff to get out, we just need to do it in a way that is thoughtful and responsible.

Q: How about any other use cases?

Keith: Hospital operations. We have one that is not pediatric-specific, and it’s around tracking and quality metrics. This is work that we haven’t published quite yet, but health systems, as you well know, are on the hook to report on various quality metrics. One of the quality metrics is around surgical site infections—identifying what percentage of patients have an infection at the site of surgery within 30 or 90 days after the procedure. That’s a marker of surgical quality, infection control, and lots of really important things for health systems. We have teams within our hospital whose job it is to both track these metrics and then identify ways that we can intervene and identify bundles. That we can introduce to help improve these metrics. But there is a core function of tracking and identifying whether a surgical site infection occurred. The way that works is that there is a team of folks who essentially read the patient’s chart after a procedure to identify whether or not a surgical site infection occurred. Reading patient notes about how the patient came back to the ER with what looks like a surgical site infection, where they were prescribed antibiotics by their PCP—those types of things would flag as an identified infection. 

It’s a ton of work to have a human read every chart following every surgery for 30 to 90 days. So, we developed a process by which a large language model is used to review the notes in that relevant time window. We ask the model a relatively straightforward question: “Do you think this note contains evidence of a surgical site infection?” Very simply, that. We provide a few examples to help it understand what are some key indicators. We do that for every single note the patient has in that post-surgical time window and come up with a score, basically, what percentage of the notes the large language model thinks refer to a surgical site infection. When we run this on our historical data, we see that the large language model thinks somewhere between zero and 60% of the notes refer to a surgical site infection. What we can do then is draw a threshold, say 5%. What we do is have the human review everything above 5%. We still need a human to understand the nuances of what counts as a surgical site infection—reading the lab results, reading the imaging results. 

We need a human to identify the true positives, but there are many true negatives. Somebody who never has an infection, does great, doesn’t come back to the health system—those probably aren’t worth a person’s time to review. If we can identify the true negatives, the reviewer could spend more time on the true positives. What’s fun about this is that the numbers are exceptional. For our preliminary data, looking at 2023 and 2024, if we set the threshold at 5%, the reviewer would be reviewing 70 to 80% fewer cases. We would miss somewhere between two and five cases. It’s not perfect, but we are nudging toward a world where we can spend our time reviewing the cases that are problematic and positive, and less time filtering through the vast majority of surgeries that don’t have an infection or an issue, moving those to the back of the queue. We certainly aren’t replacing a person. We still need the person to be involved, but we are helping that person focus their energies on the cases where their expertise and their interpretation of what is actually happening is maximized.

Q: Yes, that’s great to know. As we all know, Gen AI and LLMs are becoming all-pervasive. You have a large workforce at your health system, people at various skill levels, and now they have to either use some of these systems that are going to be deployed or are already participating in what you described as a use case process. If there is more appreciation of Gen AI, LLMs, and AI, they would be in a better position to do their job and embrace it. So how do you go about learning and development and upskilling the workforce? What are your thoughts about that?

Keith: That’s a great question, and I think it is foremost on the minds of health system leaders who are hoping to use AI in any meaningful way. The reason it’s so important is that it is totally unrealistic to expect somebody to use a tool that they are unfamiliar with, particularly when that tool is something as amorphous and multifaceted as a large language model. If you think about it, if we use the analogy, large language models distilled down are basically like office productivity tools. Think Microsoft Excel—Microsoft Excel has been around for 40 years. I used it for middle school projects. My mom used it when she was working at a bank in the eighties. 

Most people, by the time they enter the workforce, have experience with using Excel. It’s not unreasonable to have that be a requirement of the job, or when Excel gets augmented in new ways, for people to be able to jump on it. It’s relatively straightforward. We have none of that with large language models. Expecting people to use these things without a ton of training is really unfair and unrealistic. That applies up and down the organizational chain. What I mean is that leaders, just because you are an executive VP, you have exactly the same two and a half years of awareness of these large language models as everyone else. Historically, people could turn to coworkers or get extra training, but that doesn’t exist here because it is so new throughout the organization. You make an excellent point about needing to upskill our workforce and attaching an experiential component to that training. 

Sometimes we hear about organizations using online modules for training, but I don’t think this is the sort of thing where watching some videos will really give you the insight you need to understand how these tools work. The reason it’s trickier in healthcare is that it’s one thing to tell somebody, “Hey, go play around with ChatGPT. See what you learn.” But we don’t want people to use ChatGPT to come up with recipes or poems in pirate. We want people to use large language models for their job because we suspect there are major productivity benefits that can come out of this. In a health system, most people’s jobs involve PHI—patient health information. You can’t put PHI into public models. So we need to make things available. We need to make a PHI-compliant large language model available to our employees. 

At Stanford Children’s and our adult hospital, we have made a large language model that is PHI-compliant available to our entire workforce. We’ve had it available for about a year now. Our IT department is called Digital Information Solutions, DigiIS, and our chatbot is called Ask Digi. The idea is to encourage people to start experimenting with an appropriately provisioned large language model to figure out how it can make their life better. I have no idea how somebody in a revenue cycle role could use large language models to make their life better. Most folks in that role don’t know yet either. We’re going to figure it out in a couple of years, but we have to let people experiment to learn that. 

We have three broad approaches for training: One is online modules and training. We have a couple of training tools, both generally about what large language models are and about Ask Digi and the local tools we have available. The second is prompt engineering workshops. Two of our clinical informatics fellows have developed a two-hour hands-on, in-person workshop on how to develop prompts and how to know that your prompts are doing what you want. Engaging with a large language model through a prompt is a totally new skill. Starting to get people to understand what is required in a prompt, lowering the fear factor, and giving people the confidence that you don’t have to be a computer scientist to prompt these models—it’s relatively accessible—and that comes through education sessions. The third is having a local champion or expert who gets involved in a pilot project and then brings that insight back to their teams. For example, in our work with surgical site infections, we are working with the folks in infection prevention and control. 

They are seeing the ways that large language models are helpful and not helpful. They are learning alongside us. Those folks will now become the experts in their team for how these tools can be used. Hopefully, that will propagate into more training. Engaging folks in these pilot use cases is helpful not only because you learn about the use case, but also because you train the person in that area, creating ripple effects across the organization.

Q: That’s great to know that there are so many initiatives for upskilling and learning, and development in the organization. One other topic that I thought I would touch upon—we talked about it a little—is PEDSnet. You mentioned it early on in the conversation today. I would love for you to share with the audience how this helps in terms of interoperability and data sharing, and any new aspects it might bring as well.

Keith: Definitely. Happy to. Just for a quick background, PEDSnet is a pediatric EHR data-sharing consortium that has been around since 2014–2015. It’s primarily run out of CHOP—Children’s Hospital of Philadelphia—which serves as the coordinating center.We have a relationship with them, as all members of the network do, and we send them a harmonized version of our EHR data four times a year. The harmonization is based on an OMOP model. The Odyssey community is an open-source EHR standardization initiative, and PEDSnet has adopted their common data model as the bedrock of what we use. 

We have some minor modifications that are specific for pediatrics and for care in the U.S., but it enable us to share data to a central location and conduct studies with a volume of patients that is unmatched elsewhere. About 15% of the children in this country are represented in our PEDSnet database. It’s a chance for us to do large-scale studies at a fraction of the cost it would otherwise take to develop these types of tools. As we move into the world of large language models, it’s not hard to envision a future where we are able to have a large language model help us process unstructured information from these different sites, extract relevant insights from notes, and conduct large-scale studies using unstructured data. That’s not here yet, but it’s in the near future. It’s really exciting to think about the potential.

Q: That’s awesome. Talking about the near-term future and potential, if you were to look in the crystal ball, what are some of the things that you see coming down the pipe? Agentic AI that people are talking about, virtual nursing, and ambient listening. There’s so much going on. What do you think are some of the big things coming our way?

Keith: Definitely. We are rolling out a pilot of ambient listening and have a similar cue as many organizations for things that we’re going to be adopting in the near future. Taking a slight step back from a regulatory and oversight perspective, it’s important to remember that those types of issues aren’t going away, regardless of how excited we are about the technology. Any investment in technology requires an understanding of the upsides and the downsides. I think we’re currently a little over-indexed on the upsides of this technology. We’re very excited. 

From a regulatory and oversight perspective, it’s important to remember that those types of issues aren’t going away, regardless of how excited we are about the technology. Any investment in technology requires an understanding of the upsides and the downsides. I think we are a little over-indexed right now in terms of what the upsides of this technology are. We’re very excited about what it can do now and what it can do in five years. We’re starting to see some efficiency gains. The feedback, particularly around DAX and other ambient listening, is generally very positive. We are less concrete about the downsides, and what I mean by that is there’s lots of talk about potential issues that come with AI. 

In the past, we have relied on federal or state agencies to provide oversight to make sure that those downsides aren’t present or are appropriately mitigated and recognized. AI, and particularly large language models, is proving very difficult to regulate because it’s such an amorphous entity. I think it is unrealistic that we’re going to see a robust regulatory system in the near future. What that means is that the burden for making sure this technology works is falling on the provider, and that’s great—until it’s not. What I mean by that is the burden is on providers to use these tools appropriately. At some point, we’re going to see a lawsuit from someone who claims they were harmed because of a large language model’s involvement in their care. Once that happens, we are going to get a very concrete piece of evidence about the types of downsides inherent in using these tools. We haven’t reached that point yet. 

Right now, the downsides are so amorphous that they’re easy to ignore. Once there is a price tag attached to the cost of mistakes, then things become different. If that price tag of a mistake is enormous, the overall value of these tools could change substantially. We know that, particularly early on, we are potentially introducing risks and mistakes with the use of these tools. Even if the output of a large language model is 99% accurate and you have a human in the loop who is reviewing it, and their review is 99% accurate, there are still errors that are present. Part of the reason I bring that up is that at Stanford, we take it very seriously. We are ultimately responsible for the use of these tools, how they impact our providers, and how they impact our patients. Nobody is going to take that responsibility from us. That is appropriate, but we are building systems and processes with that worst-case scenario in mind in order to prevent it from happening.

Q: That is very cautious. So with that, I think this is a great session. Any closing remarks?

Keith: Thank you so much for providing this opportunity. It’s exciting to be able to talk about the work that we’re doing here. I think there’s lot to discuss regarding pediatric implications, so maybe we’ll find some time in the future to talk again.

We hope you enjoyed this podcast. Subscribe to our podcast series at www.thebigunlock.com and write to us at info@thebigunlock.com   

Disclaimer: This Q&A has been derived from the podcast transcript and has been edited for readability and clarity.

Rohit: Hi Angelo, welcome to The Big Unlock podcast. We’re so happy to host you, and thank you for taking the time to join us today.

Angelo: Thank you, Rohit and Ritu. It’s an absolute pleasure. I was delighted to be invited. I’ve listened to some of your episodes—you have a fantastic program. It’s an honor to be part of what you’re doing here.

Rohit: Thank you! That’s awesome. I’m Rohit Mahajan, Managing Partner and CEO at Damo Consulting and BigRio, and also the co-host of The Big Unlock podcast.

Ritu: Angelo, welcome. Absolute pleasure having you here. I’m Ritu Roy, also a Managing Partner at BigRio and Damo Consulting, and co-host of the podcast. We’re looking forward to an invigorating discussion today.

Angelo:  Thank you again both for the invitation and the opportunity to speak with you today. I really like what you’re doing with this show. I think it is speaking to issues that are so important right now in healthcare, in innovation, in healthcare management.

So, it’s a tremendous pleasure. I can tell you a little bit about who I am and what I do, and then we can jump deeper into the conversation. But my name is Angelo Zo. I am trained as a pediatric cardiologist and still spend a little less than half of my time in clinical practice. And my clinical practice includes patients from prenatal life all the way through to adulthood who have primarily congenital forms of cardiac disease, but also acquired forms of cardiac disease.

And my clinical practice keeps me very busy. It’s a very interesting field. It’s a field where we have many applications of technology, which I think are really interesting. It’s a—pardon me—a field that includes some procedural medicine, some diagnostic medicine, and we work very closely with other specialists, including surgeons and anesthesiologists and clinical care physicians.

So, it really is a fantastic type of clinical practice, and it’s one that many people may not be familiar with, but is a really important part of what we do in the realm of children’s healthcare.

The other part of my work is as an administrator. So I am the Chief Medical Officer for what we call the Duke Health Integrated Practice. So I am here at Duke University Health System, which is based in Durham, North Carolina.

The Duke Health Integrated Practice—you can think of it as the faculty practice of our health system. It’s actually a bit more than that. We have about 4,200 physicians and advanced practice providers—so our nurse practitioners and physician assistants are counted within the faculty practice as an important part of what we do.

And we really are the specialty practice of the health system. We have another organization that manages primary care as their primary book of business—no pun intended. Primary care, primary book of business. So that is not part of my purview, but I work very closely with that organization. That organization has its own Chief Medical Officer, but I’m the Chief Medical Officer that interacts with our 18 clinical departments—everything from pediatrics to internal medicine to OB-GYN to radiology to surgery to dermatology, et cetera.

And I help to develop strategy—primarily for our ambulatory platform—and also implement and operationalize clinical innovation procedures that help us improve access, that help us open up our door, so to speak, to new patient populations, to new areas of practice—whether it’s medical or surgical, or diagnostic or procedural. So I have a broad responsibility here and a broad engagement.

And I think the final thing I would say is that our health system is associated with—obviously—Duke School of Medicine, which is a fantastic medical school, always considered a top-tier medical school. So we have a very significant mission around education, around training, around research, and around advocacy. And all those functions are directly connected to the clinical mission. So everywhere we practice medicine, we also want to innovate. We want to do research. We want to teach. We want to educate. And we want to advocate as well. So even though my role is primarily a clinical leadership role, I get to work with other leaders in all those other areas, which makes my job really multifaceted and really exciting.

Rohit:  That’s awesome. Thank you for that introduction, Angelo. You know, it’s been quite a journey for you. You’ve been with this health system for a long time, so would you like to share your journey with the audience? Like how did you get started? What interested you, and what are some of the new things that you are working on? And also, talk to us a little bit about the patient population in the local area.

Angelo:  Sure. No, I’d be happy to. So in terms of my journey into healthcare and healthcare administration, it actually started quite early when I was finishing up my fellowship training. I was offered the opportunity to build a branch of our academic pediatric cardiology practice away from the mothership, so to speak—to build essentially a large satellite office.

So very soon in my career, I got involved in practice building, practice management, and healthcare administration. Within about a year, I took over the medical directorship of this practice, and over the course of about 10 years, grew the practice from a very small operation with just a couple of physicians seeing a few dozen patients a week, to what is now a very large, multi-specialty practice that has about three dozen medical and surgical specialists coming in and out of our doors, seeing thousands of patients a week.

Very proud to have been part of this project from its initiation, from its conception, and through its phases of growth. Along the way, I had very significant hands-on training with executive leadership, with management, and also with day-to-day clinical operations, which became an area of interest to me.

I was eventually offered the opportunity to take on a new role, and that role was as the Vice Chairman for Clinical Practice for the Department of Pediatrics. And in that role, I took what I had learned at the local practice level and brought it to an entire department, and became the operational and strategic leader for everything we do in our Department of Pediatrics, which includes about 250 or so clinical faculty, another 50 or so research faculty—primary care, specialty care, diagnostic care, urgent care, emergent care, inpatient, outpatient—all different aspects of our care.

I did that work for a little less than a decade, and I’m actually moving away from that role now, because last year I stepped into yet another role as the Chief Medical Officer for the Duke Health Integrated Practice. So I’ve been in that role now for a few months, and I hit the ground running when I took on the role—immediately, as in my first day on the job—had to help with some challenges we were going through as an organization.

It’s been very exciting, in just these first few months, to begin to see the opportunities that I will have. In terms of—you know, you asked about some of the things that we’re working on—and we’re going to talk today about some clinical innovation, and that’s been extremely exciting to work on.

Rohit:  That’s awesome. Yeah, so that leads us right into our next topic. What are some of the clinical innovations or challenges you’ve faced that you were able to overcome? And what are some of the digital health programs you’re getting involved with? Anything you’d like to dive into and share more about?

Angelo: Yeah, fantastic question. So, my first day on the job, I had to address the audience of our entire practice—all the clinical leaders from across the health system—and I told them that I saw my role as Chief Medical Officer as one who is there to support clinicians in their day-to-day work.

My platform rested on the idea that I want to bring aid, tools, and tactics into the practice that actually help people do their work. In the last few years, we’ve been facing workforce challenges—people feeling burned out, disillusioned, like cogs in a machine. I want to bring the joy back into the practice. And I think the way to do that is to allow physicians and other clinical providers to spend as much time treating patients as they can.

So we’ve identified areas where we can innovate—where we can bring technology to bear to improve the experience of delivering care for our doctors, nurse practitioners, physician assistants, and other clinical personnel.

One thing we’ve done—something very concrete and in partnership with a very innovative healthcare technology organization—is apply what’s called ambient listening technology. And we’ve done this in a way that’s been thoughtful, in response to a specific problem: how can we alleviate our clinical providers of some of the burden? In this case, the burden of documentation.

We know that physicians and other clinical providers today spend a lot of time in front of a computer—often while they’re in the room with a patient. This can make people feel like their job is data entry. We don’t want the job to be data entry—we want the job to be managing patients. We want it to be engaging. Spending that face-to-face time matters, because we know so much of diagnosis comes from the history, from having an in-depth conversation and hearing the patient’s story.

Yes, we can do a lot with diagnostic testing, but at the end of the day, it’s about engagement with the patient. That’s how we build a differential diagnosis. That’s how we begin to form our assessment and plan.

So what we’ve done is bring this ambient listening technology into the clinical encounter. The tool records the conversation between the clinical provider and the patient, and from that back-and-forth, it creates the encounter documentation.

It may seem like a simple evolution of the standard tape recorder-to-human-scribe model, or even having a human scribe in the room. But this is truly the modern version. The conversation is filtered through a natural language processing algorithm, and the finished product is something that’s usable as clinical documentation.

This has been a very exciting opportunity for us to leverage technology at the point of care and remove some of that documentation burden.

Rohit:  That’s a great use case—one of the hottest we’re seeing in many health systems, Angelo. Tell us a bit more.
You mentioned some metrics before we started. How do you measure the success of this wide-scale implementation? And you also said it’s integrated with the EMR/EHR system. These are critical for those still thinking about this.
Many have already taken the plunge with varying degrees of success, but you’ve really embraced it and rolled it out broadly. Any key takeaways from your experience that were instrumental?

Angelo:  Yeah, fantastic questions—thank you.
I’d start by saying our pre-implementation data was both qualitative and quantitative.

From a qualitative perspective, we were literally hearing from our doctors, nurse practitioners, and clinical providers that they were losing the joy of practice. These are people who’ve invested so much time and energy into becoming clinical providers. They want to engage with patients—that interaction is a special opportunity.

But over the last several years, we’ve created systems that make them feel disconnected. These electronic health records require tons of digital input—typing, checking boxes, clicking through screens.
At the same time, we’ve opened the door to patient interactions outside the visit. They’re now answering after-hours messages, doing telemedicine and telehealth visits, which are great but can also add stress.

So it starts with qualitative data. Are our people enjoying their practice? If not, we have a problem. We saw workforce challenges—many physicians and nurses left during the pandemic. We need countermeasures.

On the quantitative side, we look at “signal” data—engagement with the EHR: login times, time spent in the system, responsiveness to messages, how quickly notes are closed, and whether they meet patient expectations.

We also track provider ratings. We ask questions like: Did your provider listen? Know your history? Speak clearly? Explain things well? These bridge qualitative and quantitative—they’re numeric but reflect the experience.

If a provider has low ratings, we drill down. Often they say, “I just don’t have enough time with my patients.” And it’s not just visit length—it’s quality.
Those were some of our pre-implementation metrics.

Another key part of our approach: bringing our performance engineers to the practices. We call them performance excellence engineers. They’re incredibly skilled at turning data into actionable information.

But in recent years, they’ve been working remotely or from centralized offices—not their fault, it’s how we structured things.
We’ve found that when you put them at the point of care—whether it’s the OR, ED, primary care, or endoscopy suite—they really begin to understand the work. That’s when they can help translate practice data into something actionable.

This is classic business school 101—Toyota production system. Bring the people doing the work into the improvement process. We had drifted from that. Now we’re returning to it and seeing the benefits

Ritu:  That’s awesome. Dr. Angelo, I was at a conference recently where Abridge—one of the main players in the ambient space—shared that 81% of notes didn’t need to be edited by doctors. That led to a huge jump in satisfaction. Are you seeing similar numbers?

Angelo: Ritu, this is a very interesting question, and it’s interesting for a number of reasons. First of all, full disclosure — we are using Abridge, so that is our partner in this space. One thing that’s been wonderful about that relationship is they’ve craved our feedback about the product.

They want to iterate, and that’s why they were such a natural partner for us — we share that DNA around innovation. They want the product to keep getting better and better as we use it.

To your comment about 80% of the notes not requiring editing — this is a really interesting data point that we are looking at. We’re actually tracking the number of times notes do or do not get edited, and there’s a signal in both directions that we need to be careful about.

We need to be careful about the people who never edit their notes — because that’s one kind of story — and also ask about the people who are always editing their notes.

I can tell you that in my practice — and I have more than 20 years of clinical experience — and as someone who likes his notes done a very specific way, I’ve actually been quite happy with the quality of the content.

And when I have edited, it’s been primarily not to correct factual errors, but to change the language to better match my personal style. So it’s really been more about the aesthetic of the note rather than the content itself.

I’ve found the content to be quite good. And when it hasn’t been good — this is the other nice thing about it, Ritu, you mentioned earlier that it’s embedded in our EHR — we can provide feedback immediately.

So if I see a note that didn’t come out the way I expected, we have a chat box — we can go right into it and talk directly to the development team and say, “This is why I don’t like this note.”

We can also see the stems — the original stems of our conversation — and how the natural language processing model created the note from those stems. And again, having that direct pipeline to the developer has been critical.

Just a few weeks ago, the Abridge team came to us. They did an extensive two-day site visit — and again, they wanted to hear directly from us, from our doctors, from our clinical providers — how they were engaging with the technology, what was working, what wasn’t.

With very rare exception, most of our people have said — and I hesitate to use the term “game changer” because everyone uses it and it’s almost lost meaning — but this has been disruptive in the best possible way.

This has enhanced our practice. I’m able to close my notes. I’m able to leave the office at the end of the day not having to go home and work after dinner, or work in my pajamas, or while I’m trying to do something else — wanting to spend time with my kids, or being taken away from my hobbies and interests.

This is allowing me to close the books at the end of the day, feeling good about the quality of the content.

And again, so much of the upside is that ability to spend time — and we’re really interested to see how we improve on that one key metric: Did your provider spend enough time with you? I’m going to be really interested to see how that improves with the implementation of this technology.

Ritu: Yeah, so in the same conversation they also mentioned that some of their physicians are seeing Russia traffic for the first time, which can be, you know, good or bad depending on how you look at it.
So to your point about being able to close the notes and go home — that’s really good to know.

Angelo: I think it is. One thing we really like about this technology is the modularity. You can use it for certain elements of your note and not others. You could use it for the entire note if you like, and that flexibility has helped adoption.

If you’re really good at documentation and only want to use it for your history of present illness, you can do that. We’ve encouraged everyone to try it, and almost universally, people have at least given it a shot.
Most who’ve tried it have stuck with it — and we’re tracking that with metrics. We know what percentage of users only tried it once, how often it’s being used across different note types, etc. We’re seeing very strong onboarding of the technology. We’re not at the plateau yet — we’re still ramping up — but it’s exciting to see adoption growing.

Rohit: Yeah, shifting gears — I know we haven’t touched on this yet, but Angelo, what are your thoughts on value-based care? Is that something your practice is actively thinking about?

Angelo:  It’s very important—and it’s part of what we’re trying to do. I’ll tell you, in this part of the country, the Southeast United States, I think the move toward value-based models—bundled care and similar approaches—has been slower.

When we look at our colleagues in the Northeast or on the West Coast, we see folks who’ve been very innovative and progressive in this space. We’re certainly trying, and I think there are a number of reasons why adoption has been slower here—probably more than we can cover in this conversation.

Nevertheless, we’re very excited about the opportunities to improve care efficiency and reduce costs. We’re looking at how to bundle services in a way that clearly shows value—taking that classic formula of quality divided by cost. We believe our quality—the numerator—is very, very high. Now it’s about shrinking the denominator and reducing costs.

There are ways to do this that may involve technology, but also some analog solutions too.

We work very closely with payers and are really interested in what they have to say. Over the past few years, we’ve built strong, bilateral relationships with payers to understand what’s on their radar in terms of quality and performance metrics. At the same time, we want to hear from our own clinicians—what’s important to them?

Sometimes that aligns with what payers are looking for, and sometimes it doesn’t. But when it doesn’t, we have strong enough relationships that we can go to payers and say, “Hey, this matters to us.”

For example, our pediatricians have raised concerns about adolescent immunization rates. A payer might be more focused on vaccines in the first year of life or among older adults—like meningococcal, pneumococcal, or RSV vaccines for seniors. But we’re hearing from our team that teens aren’t coming in for pre-college immunizations.

In most cases, payers are at least willing to have that conversation.

We’re also thinking about how to clearly demonstrate our value proposition—especially with so many large employers moving into Central North Carolina. With the research triangle and three major universities here, this region is growing fast.

We want to make sure that when employers include us in their health plans, we can go to them and show the value—whether it’s for hip replacements, organ transplants, or more common services like primary or preventive care.

Historically, large academic medical centers haven’t always been great at cost containment. But we’re learning. We’re identifying opportunities, especially around clinical operations. There’s a lot of potential for efficiency there—and when you focus on that, you can truly begin to lower the overall cost of care.

Ritu:  That’s awesome. I would like to ask you about AI education and the trustworthiness of AI. Did you face any barriers from your physicians on those regards? Sometimes we see that physicians can be reluctant to try new solutions. There was a New York Times article that showed when ChatGPT was diagnosing on its own, it was much better. But when the physician came in, they brought in their own biases, and it actually did worse because they didn’t trust the AI and took it in a different direction. How have you made sure everyone is educated and had the right context within your organization for such a wide rollout?

Angelo:  Again, fascinating question. We could spend a whole podcast on just this. We have certainly needed to address this issue, but I’ve been pleasantly surprised that most of our clinical providers—physicians, advanced practice providers, physician assistants—have asked questions in a very thoughtful way. They’re not necessarily showing hesitancy, but they are trying to keep pace with developments in this area. They want to ensure we’re good stewards of the technology.

They’ve asked questions like, “What are we doing with these recordings? Are we using patient data to improve the system? Are we training the language models with our information?” They’ve also wanted to know about the process for obtaining informed consent. And yes, we are certainly asking for permission every time we use this technology, in every instance.

One thing I’ll mention is that we have a very robust structure here for digital innovation. We have several teams, and one that focuses specifically on AI. They’ve done great work ensuring our faculty are educated on basic terminology around AI, natural language models, and generative AI. Most people feel they have amazing colleagues providing these educational materials.

Another key step was the pilot program. Before rolling out anything, we did a well-thought-out pilot. We tested a few technologies and opened it up to a large number of clinicians across the entire system. This was incredibly beneficial because during the pilot phase, which lasted several weeks, clinicians provided us with feedback. We conducted pre-, intra-, and post-pilot questionnaires to evaluate all the technologies and assess what worked and what didn’t.

During the pilot, we also did a lot of education. We had review modules, town halls, and went to all 18 of our clinical departments to answer questions. Our data and IT teams, including many doctors, were also part of those conversations. That was a huge benefit because when you have IT leaders who are also physicians, it builds a lot of credibility with the clinical audience. I think that was key to winning people over.

Rohit:  Yeah, I’ll chime in with another one, Angelo. We’ve all been hearing about agentic AI, and how agents are the next big wave in generative AI. We’re currently working with several clients to define use cases for implementing agents in the workplace, as they’re talking about creating a new digital workforce that includes both humans and agents. Any thoughts on that?

Angelo:  It’s an area that’s going to grow exponentially. Right now, we’re testing the ability to use AI to screen inbound messages. As I mentioned earlier, one of the outcomes of opening up digital access to our patients is that we’re receiving thousands of messages per day. That’s a good thing—we want our patients to communicate with us—but it’s a lot of work, real work.

We have an AI innovation research team working on models that can review these messages and sort them into categories. For example, distinguishing between a message that just says “thank you for the great care” and one that says “I need a prescription refill” or “I need an urgent appointment with my orthopedic surgeon.”

The next stage is not only sorting these messages but starting to build at least the skeleton of a response. This is a really complicated problem, even for humans to do. But what’s exciting is that we have people who recognize that this is another way to alleviate some of the clerical burden on our clinical teams. If we can reduce the number of messages that require human attention, we can focus on the urgent ones. The others are still important, but they’re not as urgent. If we can triage this, it will be incredibly helpful.

And to your point, Rohit, I think the future will involve a synergy between human and technological capital. If we can create that synergy, perhaps people won’t lose their jobs but will take on new kinds of work. That’s the future I see.

I’m old enough to remember the early days of the internet, and the doom-and-gloom stories that came with it. What we’ve learned is that we need to approach these tools with an open mind and recognize that we can use them responsibly if we choose to. It’s about figuring out the most responsible and applicable use cases, and being good stewards of the technology.

I’m encouraged because many people in the space, especially those applying these solutions to healthcare, really understand the practice of healthcare. That’s why partnerships with institutions like Bridge, or technologies like Copilot, are so important. When you have a strong clinical-technological partnership, amazing things can happen.

Rohit: Angelo, would you like to share some closing remarks?

Ritu:  I think we need another podcast for more questions! But Angelo, thank you for being such a great guest and providing really insightful answers. We’d love to hear your closing thoughts.

Angelo: This was a fascinating conversation, and again, I love what you guys are doing with this show. I hope we can convene again at some point. I’d love to come back.

In closing, one thing I’d leave the audience with is that these feel like incredibly challenging times, but we can do incredibly challenging things in response. Sometimes, the most amazing road to success is when you’re pressed and facing obstacles. The tools we have today have amazing potential. We just need to be good stewards of that potential—thoughtful stewards of it.

We shouldn’t be afraid. Instead, we should continue to ask the fundamental questions and think of the future as a synergistic one, where we combine the best of our human capital with technological tools and innovation, figuring out the next step forward.

I believe in the idea of abundance. I think we have a future full of great potential, but it’s a matter of deciding how we engage with that abundance. How do we take the tremendous potential of natural language models, generative AI, and other technologies, and apply them thoughtfully, rationally, and in a way that speaks to the needs of those moving these things forward?

In our case, we listened to what our people were telling us about their work and realized that the only way to really move forward is to change the work itself. We can provide support, offer a shoulder to cry on, or an arm to hold in times of distress. But when we change the fundamental nature of the work, that’s when we begin to truly change people’s engagement with it and make them feel supported, with a clear way forward.

Doctors are resilient. We don’t lack resilience. What we need are tools to help us engage with the work by changing the terms of the engagement.

We hope you enjoyed this podcast. Subscribe to our podcast series at www.thebigunlock.com and write to us at info@thebigunlock.com   

Disclaimer: This Q&A has been derived from the podcast transcript and has been edited for readability and clarity.

 Q. Hi Dwight, welcome to The Big Unlock podcast. My name is Ritu Uberoy, and I’m one of the co-hosts. We’re very happy — this is season six, and we’re glad to have you back. Delighted to hear new perspectives, especially on AI and your initiatives at your new organization. Let’s get started with your introduction and then dive into a few topics.

Dwight: Great. Really great to meet you and be here, Ritu. I’m Dwight Raum, Chief Digital Information Officer for Rochester Regional Health. We are a health system with about nine hospitals in Western and Northern New York. We span the spectrum from large, urban acute care hospitals to small, rural regional hospitals.

We also have a large ambulatory practice throughout the region, and we really take care of patients from birth to death — and every aspect of healthcare in between. I’ve been with the organization just over a year.

Before joining Rochester Regional Health, I was in the startup world for a short time. Prior to that, I served at Johns Hopkins Medicine as the Chief Technology Officer and Interim CIO. I’ve been in healthcare for quite a long time. I really came up as a technologist, which is increasingly unusual in senior IT leadership roles. But I’ve truly enjoyed my time in healthcare — the mission really calls to me.

 Q. That’s a very interesting introduction, Dwight. It seems like you’ve seen all three aspects of technology and implementation — being at Johns Hopkins, which is a world-renowned teaching hospital, then moving to a startup, and now at Rochester. What was your role at the startup, and what did the startup do, before we get into your current role?

Dwight: Sure. It was technically a joint venture between Independence Blue Cross and Comcast. The startup was called Quil Health at the time. We were trying to build an in-home monitoring system that would allow seniors to age in place.

That consisted of various IoT sensors placed throughout the seniors’ homes. From those sensors, we ran AI models to anticipate and predict certain behaviors and patterns. Insights were drawn from the data and shared with caregivers.

It was a really interesting idea — just struggled to get it across the line, if you know what I mean. But I think it had tremendous potential. It actually speaks to the opportunities we have with signals and AI, and how that can really transform the way we provide care.

Q. Yeah, that’s very interesting. We had a webinar yesterday where we were presenting some case studies. One of the case studies we were going to present — but didn’t — was for Cherish Health. They do something very similar, radar-based. They have a device used in senior homes that can detect falls, go through walls, and anticipate events using AI.

Dwight: Yeah, I mean, the technology has evolved certainly over the last several years, and radar fall detection has really improved. But I still think there are a lot of challenges — in terms of privacy and getting over that technical hurdle for seniors to adopt and use the technology with trust.

Those are still barriers that aren’t easily overcome, even with the technology in place.

Q. Absolutely. Okay, great. So tell us a little more about your role at Rochester Regional Health — what the digital landscape is like there, your experience with generative AI, and some of the initiatives happening at your organization.

Dwight: Absolutely. As I mentioned, we’re a nine-hospital system. We’re mostly on one EMR — we’re an Epic shop, single instance. We do have some smaller EMRs, but we’re slowly but surely moving everything into a single instance.

In terms of initiatives, I’d group them into a couple of categories. The first is optimizing the system and the engine of care delivery we already have. That means ensuring throughput, doing so in a safe and high-quality manner, and providing access to our patients.

The second part is really transforming how we deliver care — shifting from a traditional fee-for-service model to a much more risk-based, value-based model. We’re focusing on treating patients for their outcomes instead of on a transactional basis.

That shift brings a lot of changes — not just in how technology works, but also in how our operators and clinicians provide care. It’s a transformational change for the system.

A lot of credit goes to our CEO, Dr. Chip Davis, who’s really leading that charge and planning for the future. When 2030 comes around and CMS starts to pay more earnestly based on value, I think we’re going to be very ready here in Northern New York to take on that challenge.

Q. In terms of the Clinical Access Center — is it a unified center? We recently worked with Sentara on a huge multi-year project where they brought everything into one place. Now they’ve implemented AI on top of that to handle call routing.

We did a webinar on their case study, and they shared something interesting — because of the AI they implemented in the call center, they can now predict when call volumes will be lower. That allows nurses to take flexible time off, which has been a big plus for them.

Dwight: I think that’s a great comparison for us. We do have a unified call center, and we are absolutely using AI for call routing and predicting high call volumes, etc. We’ve really been able to drive efficiency using AI in the call center.

Our call center is located right next to our command center, so we’re constantly looking at overall system operations — understanding volume, where to shift resources, where to move patients, and so on. Combining that with the call center is incredibly powerful.

Our call center isn’t just receiving calls — we’re also driving outbound traffic. And the third leg of the stool is that we’ve just rolled out a digital front door on our website — a robust digital experience for patients. It offers capabilities for patients to serve themselves, schedule appointments, and find the closest care options. We’re really trying to cover the full experience and serve patients as consumers as best we can.

Q. For the digital front door, do you have AI as well? Like virtual chatbots? How does that work?

Dwight: We’re not deploying chatbots in a very robust way at this point. Honestly, I’m not sure I see them as a strong tool for care delivery. There’s probably some limited front-door use for chatbots, but I don’t think they’re a great solution overall.

That said, our digital front door does provide access to direct virtual care. Patients can usually get an appointment within an hour or less — they’re able to see a provider almost right away. So it’s really about combining access with technology to deliver care where patients are, in almost real time.

Q. So. Sentara mentioned they use something called Edia for the AI part of their call center. Do you know which AI implementation you’re using?

Dwight: Yeah, we’re a customer, and there are embedded AI tools within Genesys — that’s what we’re using.

Q. So, how big is the technology component at Rochester Regional? And do you see a role for a Chief AI Officer? That’s a hot topic these days — ethics, governance committees, Chief AI Officer roles. Would love to hear your thoughts.

Dwight: That’s a broad question — let me take a shot at it. In total, we’re about 400 people in our IT organization. I might have had a bit of a Freudian slip earlier — I’m very much trying to reorient the organization to think about AI, but more broadly, innovation.

So, how do we bring technology and new ways of thinking to transform the tools and experiences we deliver?

To your specific question — do I anticipate us having a Chief AI Officer? The simple answer is no. I don’t think that makes a lot of sense, quite frankly.

What we have done is create a Center of Excellence for AI. It’s a group of the right people who need to be at the table to do two things. First, there’s the “pull” — a lot of demand from our operators who see new tools and ask, “Hey, can I get this?” We need to be thoughtful about privacy, contracts, ethics — all the right teams are part of the COE to address those questions and respond responsibly to that pull.

But there’s also the “push.” AI has been evolving at an incredible pace — even in just the last 6 to 18 months. It’s impossible for any one person, or even our providers, to keep up with it all. So the second part of our AI COE’s responsibility is to articulate what’s possible — the art of the possible.

We’re maintaining an initiative list of opportunities we think are ripe for AI — places where we can deploy capabilities, whether it’s LLMs or machine learning, to have an immediate or significant impact.

Q. And for all your AI implementations, are you planning to stay within Epic, since they’re building their own AI tools? Or are you looking to integrate outside solutions too?

Dwight: I think we’ll always look at what we already own first. Epic is a chassis in many ways — we can plug in different Epic modules and activate them. So yes, we’ll definitely start there.

But that doesn’t mean we’ll only look there. As I mentioned earlier, Genesys offers incredible AI capabilities — that’s an example of something outside of Epic but still integrated into our environment.

So no, this isn’t going to be an Epic-only play. But there are definitely some low-hanging opportunities within Epic that we’ll evaluate, ensure they provide value, and then activate them for our providers.

Q. Okay. I’m just mulling over your answer because other CIOs and CTOs we’ve been talking to feel the need for a Chief AI Officer, for the same reasons you mentioned—trust, governance, and the pace of change. It’s difficult for one person to keep up. Some believe you need a dedicated person whose job is to look at that landscape, understand what’s happening, and decide what’s best for the company. Just playing devil’s advocate here, trying to look at both sides.

Dwight: My reaction is, if we were in the business of developing AI ourselves, I would agree. But we’re not. We’re a system focused on serving our patients. Our priority is patient care, and we’ll use AI to support that. I do think it requires a collection of experts. From legal, privacy, and algorithmic safety standpoints—maybe a Chief AI Officer could be the lightning rod that brings that all together.

But honestly, we’ve come up with a methodology that works well for us in the COE. It allows us to bring together those experts and, more importantly, align their interests. Even if you have a Chief AI Officer, they still have to orchestrate all the soft power and influence that drive real change. In my view, the AI COE approach forces that to happen from the start.

Q. Yeah, that’s a very astute observation. That’s exactly what we’ve learned in our webinars—it can’t be mandated; everyone has to buy into it. It’s a behavioral change, especially with these new tools coming in.

The other point of view I have, which came up in a webinar yesterday, is that the EU has passed a law mandating AI training for any company using AI. What do you think about the training landscape? Are all your employees getting AI training?

Dwight: No, we haven’t mandated broad-based AI training yet. That’s likely to happen depending on the tools we’re using. We talk about it a lot as leadership. As we approach deployment of smart ML models for diagnosis or therapies, there will be very targeted and intensive training around each one.

Q. So as of now, people just learn on their own? Both—like LLMs or ML tools.

Dwight: So, as I said, I’ve personally hosted several webinars across the organization to provide baseline information about how tools work. But we’re not yet offering formalized training. The challenge is that healthcare is very busy and, in many ways, understaffed. Capacity is a real concern. We have to be cautious and ensure training is impactful.

My sense is that as opportunities become more acute, and we identify specific use cases, we’ll do more. We’ve also deployed an AI policy that specifies what is considered safe use for generative AI.

Q. In your hospital environment, do you have people using ChatGPT and possibly uploading data without realizing it could become public or used to train the model?

Dwight: That’s been part of our internal communications campaign—clarifying what’s appropriate. It’s hard to police, and I can’t say for certain it hasn’t happened. But we’ve provided guidance and education. A lot of people don’t realize that unless you opt out, anything you type can be used to train the model.

Q. Exactly. Okay, great. So, Dwight, talking to multiple CIOs and CTOs, the main generative AI success stories we’re hearing fall into two categories: scribing and ambient. What’s your experience with either, and do you have a success story?

Dwight: Honestly, we’re a little late to the game. We’ve just started our pilot for Ambient Digital Scribe. We’ve done early testing, and it’s clearly a win for providers. It helps reduce pajama time and after-hours documentation. But these tools are also incredibly expensive, so we’re evaluating ROI—how to unburden providers while sustaining the investment financially.

Q. Yeah, very interesting. At HIMSS last year, Nuance shared that ambient tools save about six minutes per patient. That doesn’t sound like a lot, but across multiple physicians and patients, it adds up quickly.

Dwight: Absolutely. Over the last 15 years, digitizing healthcare hasn’t been seen as a net plus by most providers. It’s negatively impacted their quality of work. COVID contributed to burnout, but so has technology fatigue. Any opportunity to unburden providers using tech—we should seize it.

Q. Good to know you’re in the early stages and testing. Any other digital programs you’d like to highlight?

Dwight: Yeah, I think, you know, there are several that I would highlight, and I would kind of back it up to a higher level for us. As I mentioned earlier, Chip, our CEO, has really been very much driving transformation. But I think a supporting function there is really innovation writ large, and AI fits into that innovation model as well.

So we’ve really begun to create an entire innovation program for all of RH, where we have two parts to that program. The first is focused on internal performance improvement—transformation and innovation. And then we’re coupling that with an accelerator program to actually take some of those innovations to market, and also to partner with other startups to bring them back into RH.

We’re doing this in a very prescriptive and deliberate way, ensuring it’s aligned with our strategy. So, where we have opportunities or challenges, we’re trying to line up those companies that may actually fit and help us solve some of those niche problems. I’ll give you a really good example of this.

Healthcare is currently experiencing a drop-off in employees, and there’s certainly a challenge in maintaining staffing. Nurses, I think, are the most acute area where we see that. But we’re also in this perfect storm where demand is increasing at a remarkable rate. So that gap between supply and demand—the only reasonable way to think about closing that gap is through technologies.

As you mentioned, AI—and we talked about Abridge as a digital scribe—as one way of doing that. But there are certainly other opportunities in that space. Just scheduling nurses, for instance, is incredibly burdensome right now and leads to tremendous inefficiency. Every little point of improvement we get there pays multiples in dividends.

So as we look at opportunities, a good example is nurse scheduling—can we apply AI to actually improve the speed and efficiency of that process? And the answer is yes. That’s a great example of a strategically aligned innovation opportunity that we’re actively pursuing at RH.

And I can stack up numerous examples like that. We talked about the digital front door and trying to improve access by casting a broad digital net, but also empowering our patients with tools to self-serve and gain access to care as quickly as possible.

It’s those types of activities that we see as the through line from the innovation continuum I was talking about earlier. So while an innovation may start by solving a very acute problem within RH, we believe we can develop those ideas into something that scales across the entire organization—and potentially even goes to market more broadly.

Q. That’s really interesting. The accelerator program—does it include in-house participation too?

Dwight: It does. A great example is a physician in emergency medicine who developed an app called COIST. It’s a heads-up display on an iPad that guides cardiac resuscitation based on ACLS guidelines and documents the process in real-time. We helped develop it, and it’s a great example of internal innovation with life-saving potential. It also eases the provider’s burden in emergency care—and it’s not a problem unique to RH. We believe it has wider potential.

Q. Yeah. So they’ve seen really good success with longitudinal care plans and generating care plans—still with the human in the loop to make the final call.

Dwight: I think especially with LLMs, that’s critical. I don’t see that changing anytime soon. One of my personal hobbies is coding, and I use LLMs all the time. It’s remarkable how good they are, but also remarkable how they hallucinate—just make things up. So I think there’s a definite need for caution and keeping a human in the loop.

That’s why I think RAG implementations have seen greater success—because you can bound them, put guardrails in place, and ground them in truth.

Q. Any challenges you’d like to talk about? What do you feel is your biggest challenge with technology?

Dwight: Yeah, I mean, I think, you know, the financial pressures in healthcare are certainly still very acute. And I think, in a lot of ways, that’s the oxygen you need to really drive wholesale innovation and transformation. So that’s certainly a challenge.

I think cybersecurity is also something that continues to be a top-of-mind concern for any CIO these days. And, for that matter, the accompanying investments you have to make to really tighten things up, so to speak. Mm-hmm.

Those are the challenges I see. And there are challenges around workforce, too—maintaining a team that can really drive and advance the ball is very critical as well.

By and large, I’m an optimist by nature—and a technologist by nature. I actually think there is real opportunity. If you take a step back and look at that gap I was talking about earlier—around supply and demand—and, for that matter, how much we spend as a country on healthcare, I do think there’s good reason to be optimistic about the future and how technology can play a really important role in helping us, as a society, meet the demand of healthcare and actually improve people’s lives.

We hope you enjoyed this podcast. Subscribe to our podcast series at www.thebigunlock.com and write to us at info@thebigunlock.com   

Disclaimer: This Q&A has been derived from the podcast transcript and has been edited for readability and clarity.