Human-Centered Innovation Driving Better Nurse Experiences - with Umesh Rustogi of Microsoft

Welcome, everyone, to the AI in Business podcast. I'm Matthew DeMello, Editorial Director here at Emerge AI Research. Today's guest is Umesh Rostogi, General Manager of Dragon for Nursing at Microsoft Health and Life Sciences. Umesh joins us on today's show to explore how nursing workflows are straining under documentation burden and how ambient AI is being built, not repurposed, to fit the realities of frontline care. Our conversation also examines practical workflow changes already emerging in the field, from automated flow sheet capture, faster access to clinical sources, and streamlined note generation. The transformation also includes early ROI signals ranging from reduced documentation latency and overtime to measurable gains in patient satisfaction and nurse confidence. But first, are you driving AI transformation at your organization? Or maybe you're guiding critical decisions on AI investments, strategy, or deployment. If so, the AI in Business podcast wants to hear from you. Each year, Emerge AI Research features hundreds of executive thought leaders, everyone from the CIO of Goldman Sachs to the head of AI at Raytheon and AI pioneers like Yoshua Bengio. With nearly a million annual listeners, AI in business is the go-to destination for enterprise leaders navigating real-world AI adoption. You don't need to be an engineer or a technical expert to be on the program. If you're involved in AI implementation, decision-making, or strategy within your company, this is your opportunity to share your insights with a global audience of your peers. If you believe you can help other leaders move the needle on AI ROI, visit Emerge.com and fill out our Thought Leader submission form. That's Emerge.com and click on Be an Expert. You can also click the link in the description of today's show on your preferred podcast platform. That's Emerge.com slash ExpertOne. Again, that's Emerge.com slash ExpertOne. Without further ado, here's our conversation with Umesh. Umesh, welcome to the program. It's a great pleasure having you. Thank you, Matthew. I'm excited to be here. Absolutely. I think healthcare is a huge frontier, especially for big tech, a lot of hyperscaling interests. I think healthcare systems are, at least from what we hear on the show from professionals, are under tremendous pressure as nurses face historic burnout, high documentation demands, and growing administrative complexity. There's the promise of AI relief, but success depends on how well technology reflects the realities of frontline care. And we're focusing squarely on nurses today. I know we've talked a lot about doctors and other professionals in the past, but I think that you have a huge opportunity here just with patient experience and just the proximity to end users, as we say, on the data science side. But really, just to kind of get a lay of the land here, what are you seeing as the biggest workflow and documentation challenges facing nurses today? and how did Microsoft identify these as priorities for innovation? Absolutely, Matthew. I think we all know that nurses are the backbone of healthcare, right? And we have about four to five nurses for every physician. But I think what most people don't realize is the profound challenge that the nursing workforce faces today. So we have about two-thirds of the nurses, about 65%, who report high stress and burnout, which is causing a high turnover and essentially staffing shortage. In the U.S. alone, we have a shortage of about 78,000 registered nurses, and the gap is only going to grow over a period of time. It's projected to grow to about half a million by 2030, right? Driven both in part by the increase in healthcare demand, as well as this attrition thing that I just talked about. And a big driver of that stress and cognitive burden is the documentation and the administrative tasks that the nurse has to do, right? So it's, I mean, based on a recent survey, it said about 25 to 40% of the nurses' shifts time actually goes into these documentation and administrative tasks. And all that time is essentially the time taken away from taking care of the patient, right? One classic example of this documentation is what is called the flow sheet documentation, where the nurses constantly move around between patients, right? I mean, within a nursing unit, right, to take care of the patient by their bedside. And they are doing various kinds of assessments, pain assessments, vital assessment, head-to-toe assessments and all that. And what they have to do then is then manually document these row by row in the EHR system, right, as structured observations. And all that takes a lot of time, right, in addition to all the cognitive burden, essentially taking time away from patient care. The other part of the question about how did Microsoft identify these priorities, we had Dragon Copilot, which is our AI clinical assistant, which was initially built for physicians. And it has been reducing documentation burden for physicians for years. Those health systems actually told us the challenges that their nurses faced. So they prompted us to explore how the same ambient AI technologies can also be brought to the nursing workflows. So that feedback kind of shaped our innovation priorities. So our goal is to basically empower nurses with drag and co-pilot AI capabilities that reduces their cognitive burden and gives them time back so that they can focus on what matters most, which is taking care of the patients. Absolutely. And I think healthcare especially is a really great example for a lot of what you're talking about. And I know what immediately comes to the mind of our audience is documentation processing. That's a very deterministic kind of first generation AI use case. And I know we're in this moment of agentic AI and physical AI kind of coming down the coming over the horizon for many industries. healthcare is hardly spared in that respect, even if it might happen a little bit later, rather than most industries, but interested very much in how AI is being deployed to reduce documentation burden and return time to patient care, especially where they deal in those deterministic use cases that even our friends in financial services, manufacturing elsewhere, will say, you know, we've had this stuff for 10, 15, 20 years, but we're not even using enough of that where it could really be successful even where agentic or generative or physical is coming down the pipe Absolutely So I mean within Dragon Copilot we have built a new set of AI capabilities which are specifically for nurses, right? So we focused on a variety of use cases, and I'll go through a couple of these. So first and foremost is the ambient flow sheet capture, right? I just talked about what these nurses have to do on a day-to-day basis, right, which is manually key in all these structured observations into the EHR system. What we have done with Drive and Copilot is basically integrate Drive and Copilot directly into the Epic Rover mobile app. And all the nurses have to do now is as they are going by the patient bedside, they basically just pull out the mobile phone, select the patient, and then start the recording. And then, and this is all this without being tethered to a workstation, right? So as they are having a conversational dialogue with the patient, all that recording is being captured. And then AI does the smart magic behind the scene and then extracts the relevant observations from that conversation between the patient and the nurse and automatically populates into the EHR system where the nurse can go in and then it's already populated. They can review and edit if needed and then file it in the EHR system, right? So this, in addition to their conversational mode, the nurses can also use the same thing when they are on the go. When they are moving between patient and so on, they quickly wanted to make an observation, just take out the phone, start the recording, and then here you go, right? And the AI at the back does the same thing in terms of extracting the flow sheet and so on. So that's one, the ambient flow sheet capture. I would say the second use case that we kind of focused on was providing access to trusted clinical sources. So nurses don't do the same thing like every single day, right? And they sometimes have to switch across nursing units. So often they need access to medical information from organization approved sources. And these are trusted, credible sources. For example, the CDC, which is the U.S. Center for Disease Control and Prevention, or the FDA, or Moke Manuals, or Medline Plus, and so on. So what the technology we have built basically gives them access to these trusted sources with proper citations, right? And with the proper safeguards to eliminate or avoid those hallucinations. I would say the third thing we have done is around automating routine tasks, right? So from the same patient and nurse conversation that I alluded to earlier, Dragon Copilot basically captured the conversation, has the transcript, and the nurse could now easily, at click of a button, can say, okay, hey, draft a nurse note out of this, right? Which could be either a narrative note, it could be a physician notification, it could be an incident note, right? And then take this and then, okay, here you go, right? So on click of a button, they basically get all this without having to manually type it in some system and so on. And all this basically means less overtime, less mental fatigue, and then gives them more time and energy for their time with the patient. And in addition to these, we also have, we know that AI, there are significant differences that can be there across health systems. people's or healthcare organizations' flow sheet schemas can be different, right? So we have provided capabilities for AI customizations, AI tuning, where to a certain degree, these organizations can do a self-service of those things, tweak things where needed, configure it the way they want it set up, and then also get access to various kinds of analytics and insights about how their nurse users are actually using the system, which is very important in driving change management and adoption within their organization. Absolutely. Really interested, I was at a manufacturing AI conference yesterday, and really what everybody seemed to be interested in, no matter what the presentation was, the extraction from the SMEs, really what is the process? Not even how, but even how long and the methodology and trying to get information around setting expectations, especially as they're trying to manage up in the situation, really keep leadership up to date. Take us through that process for Microsoft's work here, especially hand in hand with nurses. No, absolutely. And I think you identified it very well. So co-creating and co-innovating with nurses was actually at the heart of how we built Drive and Copilot. AI capabilities, right, for nurses. So we realized very early on, I mean, once we figured out this is a problem domain that we wanted to go after, we realized that the nursing workflows are unique, and they're quite different from the physician workflows, right? I mean, just to call out some of the differences, it's a very mobile environment where the nurses are always on the move in the nursing unit or the hospital, it's very fast paced, the documentation they do is much more structured observations document versus just the notes that the doctors typically do, right? And they had very different workflows. So we had to think about, we had to purpose-built our AI solution in Dragon Copilot for them, for nurses, versus just repurposing the current technology that was being used for physicians. So what we did was, over the course of, I would say, last 12 to 15 months, we basically did co-innovation with some of these leading healthcare organizations like Mercy, like Advocate, like Stanford Healthcare, among others. And we had our product management teams, engineering, design, research, customer team engage very closely with the nurses at these customers. And we got tons of valuable feedback from not only the frontline nurses, but also the nursing informatics team, the nursing leadership team. And based on that feedback, we designed the product, we built the product, we rolled out, We got feedback. We iterated fast, right, to get to a point that we now believe that, okay, here, they are robust enough, that we are going to launch it as a generally available product in early December. And we couldn't have done this without all this valuable real-world feedback from these nurses who used it in their day-to-day workflow and gave us feedback, what is working, what is not working, and helped us improve. Yes. And as I kind of say, there's a bit of a cursive lesson. The audience is used to me bringing this up, but a bit of a cursive lesson that I think goes hand in hand with manual. When you're at your AI white belt, you know, bringing this into the organization, the enterprise, all manual processes are the enemy and you need to eliminate them like, you know, space invaders. When you get to your kind of your AI black belt it not that manual processes are the enemy It that you want those manual processes to be meaningful Is that reinforcing how you remembering something Is that bringing you a closer relationship to the patient It sort of the kind of tap dance we do in education around cursive because it really grabs headlines. This was sort of like, you know, 12, 13 years ago when there was kind of a war on cursive and everybody was kind of saying, especially on the tech side, oh, we should be teaching our kids coding. But the thing is about cursive is that It actually builds deep neurological bonds in your memory to what you're writing down. And I think the same kind of goes for manual processes, especially documentation, not too far departed from nursing or cursive rather. Just very interested in whether you saw the nurses engage in manual processes for the benefit of the patient, even after they had the solution, or if there was kind of a more give and take, not that they had to do it, but this would bring them, you know, some greater advantage, some greater proximity to the patient itself. No, so I mean, that's an excellent observation. So I would say they were based on the rollout that we did, and the feedback that we got, it's clear that all these nurses, as well as the organization basically had tangible outcome. And I'll answer two different perspectives, right? So one was the positive impact this technology had in the day-to-day life of nurses, right? And then second is the more quantifiable organizational benefits like KPIs and metrics that tend to improve as a result of use of these technologies, right? So for the first one, we have seen like many real-life stories, nurses who now feel more confident, more connected, being able to spend more quality time with patients and deliver like more compassionate care, right? And that's what we were striving for. So I'll just take a couple of examples. So we had, for example, this nurse at Stanford Healthcare, who told us that she loves how this diagram copalette now makes her life easier. It cuts down on the charting time by extracting automatically the flow sheet data, and then gives her more time to spend with the patient. We also had another nurse at Mercy Health, I mean, who is actually hard of hearing. She told us that she knew she was missing things, right, before she was using this technology. And now she feels a more confident because she can go back, refer back to the transcript of a dialogue with the patient, right? And then in addition to all the flow sheet data being automatically extracted, she can now refer back, right? And this is how it has basically changed the day-to-day life of these nurses. So from the organizational perspective now, we have actually seen like Mercy recently reported they have seen about 21 21 percent reduction in documentation latency and this documentation latency I mean maybe for your listeners it is the reason you do documentation is that everybody involved in the care team different nurses different physicians and so on who are attending to the patient can have access to the same data right so it is important that this documentation is actually done as near real time as possible. But in a traditional way, without the use of this technology, nurses would actually make paper notes or maybe sometimes scribble on their hand and then bring to their workstation. It could be half an hour later, could be 45 minutes, could be one hour later that they are actually entering the information. And now with the use of this technology, I mean, they're basically recording it as the conversation is happening, right? As the nurse is actually making the observation and it goes back into the EHR system and all the nurse has to do is kind of quickly review and then, as I said, edit if needed and then file it. So the documentation latency has actually reduced, right? Mercy said, okay, hey, by 20 plus percent, we had another customer who said, in their case, it reduced by almost 70%. Then that's one big organizational benefit. We also had nurses or these organizations who are reporting that, okay, in a given shift, We have almost anywhere from eight to 24 minutes saved per shift, right? And which is huge because all this time saved, I mean, in addition to reducing the cognitive burden, okay, is the time given back to the patient. We also had these organizations kind of report about 29% reduction in incremental over time, right? Again, because of the efficiencies that have been. And the last one, and this is actually one of my favorite, is we, like Mercy actually did a survey of the patients to contrast patients which were being served by nurses who were using this technology versus nurses who were not using this technology. And what they saw was almost a 4.5% increase in patient satisfaction of nurses who were using this technology because those patients felt as if they were better taken care of or the nurses were more attentive to them as they were taking care, right, as compared to the other patient. And this, I think, is huge, right? So all in all, I mean, as I said, we have seen both these kind of tangible outcome or advantages coming out of this technology, both from how it changed and improved the day-to-day life of nurses, as well as the more macro-level organizational outcomes and benefits. Absolutely. Patient experience improving in any way is its own benefit, of course. But I think especially in this beginning era of AI coming into the healthcare space, I think the first filter is patient experience, but that's tangibly different than patient benefits. I'm wondering just in terms of your outcomes, what are you measuring in that latter category outside of maybe how the patient feels about it? No, so very, very valid question. So, yes, in the end, what matters is the patient outcomes, the health outcomes, right? So now one of the things I said is the documentation latency improves, right? Things like documentation latency, when the patient information is available to the care team at the right time, I mean, it essentially leads to better health outcomes, right? Another additional thing that I wanted to point out was the quality of the documentation improves, right? So we have had organizations who say that, okay, with this, actually the quality or the thoroughness of the documentation has improved as well because, and there is a term used called invisible care. So there are many things that we would actually not bother to document just because it takes more time to document it. But by the use of this technology, it automatically picks it up and then documents it, right? So the quality of the documentation has improved as well all those quality of the documentation the documentation latency the fact that the nurse has better access to trusted information the fact that a nurse has lesser cognitive burden and can spend more time with the patient is more attentive All these things contribute to better health outcomes for the patient. Absolutely. Just a final question. And especially as we start to see these tools really come across the horizon into the healthcare space, I think there's a more defined view of what nursing is going to look like in the next five, the next 10 years, drastically different than being beleaguered by paperwork documentation as it is right now. Just any advice for nursing leaders, nurses themselves about what that future is going to look like and how they can better prepare to have a stronger, more close relationship with patients going forward. No, no, absolutely. So I think one of the things, and this may happen much sooner than five years. So a few things that will happen is I think AI will become more and more pervasive in the nursing workflows. Something that impacts obviously the nurses, but also the health systems and the patients. What we will see more and more of is things like agentic AI, right? So when I talked about the administrative task, I mean, there are so many different use cases where, for example, the nurse is spending time on, let's say, at the time of discharge of the patient, just calling an Uber or a Lyft for the patient. Right. It takes time. Right. And these are things where we will very quickly and potentially see agentic AI kind of takeover where, hey, just ask the agent to actually book a taxi for this patient. Right. You will have other problems like scheduling, like a nurse manager typically spends about four to six hours in a given week just planning the schedule of nursing in their units. And they have to call up various nurses, coordinate who's available when, making exception. Hey, that nurse has to go on vacation because she's sick, can't show up. Can you do an overtime here and so on? And this is where, again, agentic AI can actually pick up. So we will see those kind of use cases where things that can be offloaded to an agent will be offloaded to an agent. We will also see a lot more hands-free workflow where what I described still required somebody to actually carry a mobile phone. Right. And then open the app and then select something or start the recording. Organized health systems or hospitals are actually investing in smart hospital room technologies. We'll see these technologies get integrated with a Dragon Copilot technology for nurses, right? So that we can deliver a hands-free experience to nurses where system is just observing, right? No need to take out the application. It's all friction-free, right? You can focus on the patient. You can focus on the work itself. Exactly. That's a key, right? So a nurse, I mean, most of nurses' time should go towards taking care of the patient and nothing should come in the way, right? And they should go about doing their business as naturally as they can. And it's basically the audio and the video and the sensors in the room are basically picking things up and then doing what needs to be done at the back end, right? So the system is actually now working for you. So these are some of the things that we will see come up more and more in the coming years. I believe it will come much sooner than the five years, 10 years that we were talking. And ordering an Uber, especially for patient experience, is not nothing. I mean, we've had other healthcare professionals on the show come on and talk about how when they were doing the data gathering process for their AI adoption, that a huge factor in patient experience was parking spaces. parking spaces. And it seems like such the tangential thing or so silly to think about of like, oh, we got to think about parking spaces in terms of the patient experience. But I'll tell you from, you know, my wife is expecting right now, that is a huge concern for when we show up to the hospital, just making sure that everybody's got parking. But very, very encouraged to see that we're going to see a world, maybe we're not, we might not have infinite parking spaces, but at least the Uber is ordered and taken care of. Umesh, thank you so much for being with us this week and giving us a peek into this future. And thank you, Matthew, for the opportunity to talk to you and your listeners. Wrapping up today's episode, I think there were three critical takeaways for life sciences leaders from our conversation today with Umesh Rostogi, general manager of Dragon for Nursing at Microsoft Health and Life Sciences. Here are three we'd like to summarize before wrapping things up today. First, frontline workflows require purpose-built AI, not physician-centric repurposing to meaningfully reduce documentation burden and cognitive load. Second, ambient capture paired with trusted clinical sources can shift manual structured observations into automated, review-ready inputs, improving accuracy and freeing time for patient care. Finally, early deployments show measurable ROI, shorter documentation latency, lower overtime, and higher patient satisfaction when AI augments, not replaces, nursing judgment. Interested in putting your AI product in front of household names in the Fortune 500? connect directly with enterprise leaders at market-leading companies. Emerge can position your brand where enterprise decision makers turn for insight, research, and guidance. Visit Emerge.com slash sponsor for more information. Again, that's E-M-E-R-J.com slash S-P-O-N-S-O-R. I'm your host, at least for today, Matthew DeMello, Editorial Director here at Emerge AI Research. On behalf of Daniel Fagella, our CEO and head of research, as well as the rest of the team here at Emerge. Thanks so much for joining us today, and we'll catch you next time on the AI in Business Podcast. Bye.