Scaling Generative AI for Safer, Smarter Care with Dr. Mark Mabus

0:00The most important healthcare decisions don't happen in isolation. They happen when leaders come together. Becker's 16th annual meeting brings together more than 3500 hospital and health system executives this April in Chicago. With 800 speakers from Ascension, Cleveland Clinic, Common Spirit, and more, the conversations get real. Leaders will share how their scenario planning for policy shifts, breaking through value-based care barriers, and building clinical teams that translate new ideas into real-world care. Join top decision makers in the room April 13th through the 16th. For the agenda and event details, visit Becker's HospitalReview.com and click on the Events tab in the upper right. This is Laura Deerda with the Becker's Healthcare Podcast. I'm thrilled today to be joined by Dr. Mark Maybus, Chief Medical Informatics Officer at Parkview Health. Dr. Maybus is a pleasure to have him in the podcast today. Thank you so much, Laura. It's a great opportunity to be here, and I'm excited

1:01to share with the community. Absolutely, and this is going to be a fun conversation because I know there's so much you're doing right now at Parkview. Some really cool things in generative AI and more, and so I know that it is some of that, but before we do, can you introduce yourself? Just tell us a little bit more about yourself and Parkview Health. Sure, sure. Like you said, I'm Dr. Mark Maybus. I'm a practicing family physician. Still see patients a day a week. CMIO at Parkview Parkview is a large not-for-profit health system in Northeast Indiana and Northwest Ohio with multiple hospitals, a large employed medical group, a strong focus on community health. We have we have urban centers, suburban centers, and rural all within our area there too. My role, you know, I sit at the intersection of clinicians and technology and then governance to the application analyst for the EHR report to me,

2:01but I'm also in charge of provider efficiencies there too. So making sure that EHR AI, which we'll get into, and all these digital tools, you know, create efficient workflows, safer care for patients and quality care for patients. You know, it's a fun job to be in between all of that and be able to see all the different areas of care throughout our health system. That's, you know, so cool and definitely an interesting place to be having that clinical background, the technology and bringing that all together. You know, it's truly a view of health care that not many others have. Now, I'm curious. When you think about the last year or so, could you talk about the most important initiative that you led? What do you do and what were the results? I think our most important initiative last year was getting assistive AI, you know, generative AI deployed throughout our health system. We did that in a couple different ways. You know, the ambient documentation was

3:05a big one for our providers. You know, we partnered with a vendor to provide that service through Epic. And we use a lot of Epic's generative AI tools that they have available as well. And in fact, we just got our recent class research arch collaborative survey that said that 88% of our providers also said that the best improvement in 2025 to our EHR was the deployment of generative AI in some form or another like that assistive AI. We moved from limited pilots to broad access, you know, and made sure that we were able to deploy these tools across all areas of care delivery, you know, ambulatory emergency department and inpatient. And at the same time, we were very intentional about how we messaged this to providers. You know, this was the advent of assistive AI. We're not getting into automation and, you know, computers making diagnoses for you, nothing like that, nothing going to replace

4:08clinical judgment. It's kind of, you know, building that trust with your providers, hey, we've got these new tools that can help you be more efficient. And then, you know, deploying that out, you know, with piloting first, we're able to make sure it works for our workflows. Then we have the mass adoption. So very early in 2025, we had let our ambient documentation be available to all of our providers. And then in October of 2025, we made over 18 generative AI features available in Epic to all of our providers as well. So looking at, you know, different areas where we can deploy AI and do so at mass scale. You know, we prepped it with all these little pilots, especially using some of our super users are what Epic calls physician builders, you know, to help us with testing and feedback. And then getting it, you know, out to the masses,

5:10that I, you know, specific points in time and bang, big bang. I'll give you another statistic with the generative AI features, like I mentioned, that we deployed in October. We were running at about, you know, 60,000 tokens per month or something along those lines. In October, since we turned everything on, that jumped to 2 million tokens in that month. And just this past January 2026, we hit 3 million tokens. So it was like, yeah, just a little bit here and there, you know, getting our pilots under our ground and then throughout the whole health system, these are available. People are using them, you know, not just our providers. You know, there's, there's obviously things that help our providers, but, you know, we also deployed in coding and revenue cycle side of our non-clitical folks can benefit from AI. But yeah, planning those things out took a lot of work and then you can go at scale and get these in the hands of everybody.

6:11Yeah, generative AI is the thing. It's what everybody needs to be doing across their health care system to, to enable these tools for, for efficiency or ROI, all those things, you know, it's, it's here. Use it. I love that, you know, such a clear call for having this technology and doing what you can do in order to make the most of it. And it seems like you've got a strong process, not only for, you know, having some of those tests and those pilots, but then bringing things to scale. And I think that's the place where a lot of health systems and organizations really struggled to, you know, do that effectively. So what have you done or what have you found, you know, worked well when you're taking some things from, you know, those smaller pilots into a broader and scalable space and seeing this kind of like adoption, which is wild. A lot of it is figuring out workflows and where this fits properly. You know, you can have your vendors give you the instruction manuals or, or,

7:12you know, how this actually works in technically, you know, but when you dive into the workflows of an actual user and start using these things, you get to figure out that maybe it's not so cut and dry out of the instruction manual. You've got to create those best practices. That's what pilots were able to do for us on these in these scales. And, you know, it does take you adopting the right people for your pilots. I mentioned that we used physician builders, but, you know, those physician builders are also in tune with non-savvy users workflows too. There's kind of that spectrum. You got your early adopters and people who do great with anything that you throw at them technology or change management wise. Got your middle of the road folks and then you got your, you know, we'll just call them slower folks, you know, that little more resistant to

8:13change or, you know, you know, the ones that take some extra education. So you've got to keep all those different areas in mind. And so when we are developing best practices, we don't just look at those high technology adopters, we look at everyone across across the board. How can a workflow fit everyone at the same time? And that's what becomes our best practice. We use our informatics teams to write those best practices up, create videos, you know, document in educational emails or in, you know, in workflow education as well. We utilize in basket messaging for a lot of our informatics type and workflow education because people actually look at their in baskets in the EHR rather than having to go out to, you know, their email program to look at it that way. So connecting with providers, getting boots on the ground or informatics team knew that these were high initiatives for the year.

9:13So they actually scaled back some of their just kind of regular scheduled clinic supports and instead focused on these high level deployments, whether it was ambient documentation or, you know, October when all 18 of these features went live, they made sure they scheduled time in each of the departments to go over the features that applied to the users in those departments. So it's possible to get this out to all those users, but it is a lot of coordination, a lot of effort and, you know, kind of all hands on deck when these things do go live. Think of it like a mini EHR go live. You want to be as prepared as possible because these are high impact tools and kind of a different type of tool that they've had available to them than in the past. So we made these, you know, kind of our two big key points, you know, ambient documentation and then the big bang of generative AI.

10:16High points for the year, big focuses for those months that that happened. And I think that targeted approach targeted education and targeted feedback, you know, being able to take that back to our vendors and have improvements made from there. That all led to very successful deployment of these tools. And that makes a lot of sense and is, you know, really helpful to understand that mindset and how you go into understanding where different people are at along the adoption scale. And then, you know, looking at those workflows and figuring out how to communicate with them and having AI really be supportive to the things that they're doing on a daily basis. Now, something else you mentioned there too is thinking about the value to the health system and the ROI. How do you think about that? How are you measuring that? I think, you know, it's one thing to look at some of the software, ROI, time saved and everything else. But, you know, when you're thinking about the big investments that you have to make in order to get this right, what makes the most sense for you

11:17when you're trying to figure out, hey, as it's actually doing what it needs to do and, you know, what is that return on investment? For sure, with today's financial climate and uncertainty in healthcare, you do want to be a good steward of the resources that you are provided. The kind of initial approach that we took, you know, with ambient documentation, we knew there could potentially be a financial ROI with that. Yes, there are soft ROI's with, like you mentioned, decrease time and notes, decrease time in the EHR, more time with patience, decrease burnout. Those are all kind of soft and hard to quantify financially there too. But if you have, you know, say, in extra 15 minutes, were you actually, you know, using that 15 minutes a day to throw another patient on? You know, those questions were, you know, of course, big on our finance side. And actually, with some of our pilots, we did require that our pilot users,

12:17if they were going to use the technology to help cover the cost, you know, that they would have an additional patient per clinic day there. But as we went on and saw some of the other benefits of this, a little bit lower provider turnout rate and definitely decreased burnout metrics that we were seeing, you know, we did become a little more a lax on that requirement. And I know a lot of other health systems have done the same thing there. With some of the generative AI functionalities, I mentioned the non-clitical side of things in coding revenue cycle. Those are ones that are kind of, I would say, no brainers to implement. You know, that is something that can absolutely potentially get you a financial ROI. We even have one for our providers that helps them calculate their standard level of service ENM codes based on medical decision making. Were they coding things properly with mild moderate or high levels of complexity there?

13:17They have a little calculator that AI can read their note and tell them, you know, which level of service to code. So that brought a little bit of financial ROI there as well. The inpatient folks, you know, being able to get their discharge summary has done faster too, you know, so they could move on to see their patients. So finding those key AI functionalities where you absolutely can see a financial ROI is kind of key to getting things started. And then those help to pay for, you know, some of the other functionalities that may have more of that soft ROI based more on efficiency and time savings. I can use some of these to pay for those that was kind of the mentality that we used. It's also nice to have, you know, for example, with Epic systems,

14:20you can have a monthly fee for many, many different generative AI functionalities. So you know that you are, you know, getting your money's worth by having the revenue cycle side to help pay for the clinician side of the AI functionalities there too. So focusing on, I guess I can mention that too, you know, focusing on your key strategic vendors that you already have, there are some vendors that health systems, you know, pay seven figures or more too, that offer embedded generative AI now. Why not use those as opposed to searching out a third party, you know, that is a niche area, you know. Get your money's worth for what you're doing. I think with financial headwinds this year, there's going to be a lot of health systems that have to think that way. So if they do have to prioritize things, prioritize what is already included or included with a minimal or a bump in, you know, licensing fees as opposed to, you know,

15:32a huge investment in something else, you know, we have to weigh the options of each of those and what works best for our system at that time, what works best for that service line or service area. At that time, there's, there's a lot of different ways to go about it while still being, you know, financially responsible. That's helpful to understand. Thank you so much for providing a bit of extra context there and you know, it seems like a very important thing to be thinking about as you're looking into the future and speaking of that, what are some of your big priorities and headwinds that you're focused on for 2026? Sure, it's, it's the scale of AI, you know, not just getting all, you know, things available to all our providers because it's really, you know, even though it's available to all of our providers, not all of them are using it. I'll be honest, we have about a third of our providers who are in that high user category that we will, you know, see broad adoption of, we have another third that a middle of the road,

16:37where although you some, but not the others and another third that seem to be a little bit more disengaged. So our focus is going to be this year on the education of that middle and lower thirds. And maybe we can hit, you know, 50, 50 or 60, 40, you know, this year, I think education and, you know, perhaps even one on one training on some of these things may help us move that needle for clinicians and how they are utilizing this AI. Another big area is to be able to, you know, properly inventory and document. Now that all these AI functionalities are live, you know, you got to make sure there's ongoing review, make sure there's not bias or drift that happens with some of the models. You know, making sure that these tools are still functioning in a safe and high quality manner. We're doing some of these things through our what we call AI steering committee, you know, every health system should have some sort of AI governance.

17:43We call it a steering committee and that is overseeing and directing some of these tools. As we're waiting on a different, you know, regulatory bodies or legislation to, you know, specifically say things, be proactive and, you know, start audits yourself and keep track of all the functionalities that you already have live. So it's this year is the age of continuing education for users. It's the age of inventory's libraries and feedback mechanisms and and upkeep, you know, all while weighing additional options with not just the generative AI solutions, but agentic and even perhaps autonomous in the coding arena, nonclinical, definitely best to go nonclinical for automation there first as we're waiting on some of the other things to happen with regulatory there too. But getting getting that groundwork so you're prepared to, you know, take on additional functionalities or more advanced functionalities while keeping in the back of your head, you know, how do I educate my users, how do I ensure safety quality and regulatory compliance as all these tools are there.

19:01And that makes a lot of sense and I know it's always that fine line between wanting to be innovative, continuing to push the boundaries, but also not taking on too much risk. Yes, absolutely. Well, what do you think the hardest thing you're going to have to do in the coming year will be? I think it's saying no, actually, you know, because of that pacing, you know, are there things that need to need to wait in, you know, my health system in particular, you know, we're one of one of the leading medium size healthcare organizations in the country, you know, we don't have unlimited resources. So what makes sense for us, you know, may be something that is provided through an existing vendor, like I was saying, rather than some new latest and greatest things. What can we take on throughout all of this? We may have to say no to focus on instead upscaling what we already have educating who we already support and things like that.

20:10We don't want AI to overwhelm all of our users, but we do want to be transparent with what's available, how we can support things now, things like that. It's not, it's, yeah, I guess it's, I guess you could call it pacing yourself, you know, yes, we deployed all this stuff, we got things going, can we maintain that pace? We'll see, we have to balance the yeses with the nose. Absolutely, and I know that can be certainly no small task, especially the volume of things that are coming at you and your colleagues on a daily, consistent basis. So, yeah, I know you mentioned being able to, you know, kind of like make sure you've got things covered with your existing vendors and partners along those lines. And so, you know, when you do have folks that are excited about something new or coming through, how do you communicate, you know, and identify the things that are going to actually make a difference and decide on taking a new partner when that situation arises.

21:12That's one of the ways that our AI governance helps out as well. We do have an update to our vendor intake forms, you know, everybody's got their own way of getting a new technology or new services into a health system. So, embedded within that, we now created our own AI review. The AI does focus on generative agentic or autonomous. We have other mechanisms in place for, you know, machine learning and some of those quote older forms of AI. And one of those questions, you know, that's on there, you know, first off, what's the need, you know, why are you looking at AI for this. And then another question is, does our, what we call trusted vendors, you know, those vendors that you spend a lot of money on and you better be getting your monies were from, do they offer a similar technology and have you explored that.

22:12So always that if not, please discuss with information services or informatics because they're, you know, up on this technology, they're in the know of what our vendors are bringing. They know the road maps of where our vendors are going. And so that is that point of entry, you know, for, for our vendor. If it hadn't happened before then, it gets caught in that process and it allows our informatics and IS teams to take that opportunity to work with the requesters. And, you know, show, show features that are available and our, our, our vendors have plenty of documentation on that or, you know, happy to jump on calls to explain some of those things. So embedding that in a review process or that formal intake to get your budget dollars approved, that is our way of making sure everybody knows what's going on.

23:13Got it. That's helpful to understand. Well, before we wrap up here, I wanted to ask you about growth. Where do you see some of the best opportunities for organizational growth in the future, especially considering all the different use cases for AI and how cool for the technology is evolving. I myself can bring so many efficiencies there, you know, to give time back to our providers, our nurses, our care teams, you know, one of the, one of the top reported benefits of using AI charting from our providers has been I get to spend more time with the patients. And that's what's medicine is all about and we've thrown so many other, you know, hoops to jump through just to be able to deliver care to patients. So I'm very excited and glad that AI charting has been able to provide that connection back with our providers. And as we take things live with our nurses this year as well, our nurses can spend more time with their patients to so giving time back that it that is precious and is so helpful for growth for our, you know, employees there to our users of this technology of AI.

24:31You know, the same thing can happen on the on the nonclinical side of things, getting administrative tasks done faster, getting through that coding work queue a little faster, that will, will help in their efficiencies to as a, as a health system, we are blessed with organizational growth as well. While we are expanding our reach throughout the state able to support, especially some of our rural hospitals in the state of Indiana with our technology to offering through our epic connect program as well. And also some managed services agreements, rural health is very important and is very much struggling. And, and so we're happy to be able to provide services there, you know, some of this, this AI technology, you got to think that there is some equity concerns going on too. How can a small

25:35hospital or maybe FQHC clinic, you know, afford some of these technology tools. And, you know, there are larger health system partners that can help out in that area too. There is some growing change, at least at Indiana's legislative level to also offer assistance to critical access hospitals and others that may need access to these tools too, because, you know, the technology will help promote safer and higher quality care. All areas of care delivery should have equitable access to these tools. And I'm happy to be part of a system that wants to share that, you know, with these other areas that may, you know, have a bigger bit of a struggle to get there. Absolutely. I think that makes a lot of sense. And certainly, you know, having that ability to support other facilities really keeps that care local and make sure everybody has access to what they need. And so I think that's very much a point while taking it and thinking through how quickly technology is changing. There's so much possibility and potential. It seems like a really exciting time in healthcare right now.

26:53100% agree. Absolutely. Well, Dr. Mavis, thank you so much for joining us on the podcast today. This has been a really fun conversation. I can tell you're very passionate about everything you do, which is so much fun. And, you know, look forward to seeing you as well at our annual meeting. I know you'll be speaking on a panel of things deeper into many of the things we talked about today. And so I look forward to seeing you there. Thank you so much. It's been a pleasure.