In this episode of Enterprising Insights, The Futurum Group’s Enterprise Applications Research Director Keith Kirkpatrick discusses the use of generative AI to improve efficiency within the flow of work in the healthcare space. He also provides examples from other industries, and discusses an interesting announcement about an application that can be used to track the spending of generative AI. He then closes out the show with the Rant or Rave segment, where he picks one item in the enterprise software market, and either champions or criticizes it.
You can grab the video here and subscribe to our YouTube channel if you’ve not yet done so.
Listen to the audio below:
Disclaimer: The Enterprising Insights podcast is for information and entertainment purposes only. Over the course of this podcast, we may talk about companies that are publicly traded and we may even reference that fact and their equity share price, but please do not take anything that we say as a recommendation about what you should do with your investment dollars. We are not investment advisors and we do not ask that you treat us as such.
Transcript:
Keith Kirkpatrick: Hello, everyone. I’m Keith Kirkpatrick, Research Director with The Futurum Group. And I’d like to welcome you to Enterprising Insights. It’s our weekly podcast that explores the latest developments in the enterprise software market and the technologies that underpin these platforms, applications, and tools. This week, I’d like to delve into the use of generative AI to improve efficiency within the flow of work. Now, this is a topic that’s particularly timely given some of the announcements that came out last week at the annual HIMSS Conference in Orlando, where generative AI is being deployed in applications that are focused on the healthcare space. I’m also going to talk a little bit about how generative AI is being used in other verticals to also improve efficiency and accuracy, as well as, of course, deliver business ROI. Then as always, I’ll close out our show with the rant or rave segment, where I’ll pick one item in the market, and I will either champion it or criticize it.
So, let’s get right into it. Generative AI, we keep talking about it. Why? Because it is the flavor of the month, or year, or perhaps even the decade, but let’s talk about generative AI and its actual usability. Really, right now, the most value is still derived around addressing things like repetitive tasks, things that workers have to do every day, every month, every week, basically on a schedule. It’s the same process time and time again. This really, for the most part, provides the least risk for deploying generative AI, provided that the model is only acting upon vetted data, meaning that the model is not going out to the internet as a source of information. It is only grounded in company data that has been properly cleaned, vetted, verified, and honestly, it’s data that is constantly checked to make sure that there is no additional data being added to that, that might cause bias or other sort of unwanted content to come in and be susceptible to being used by those generative AI models.
Now, of course, if it’s done right, generative AI can deliver real benefits in terms of improving efficiency, productivity, and accuracy when it’s applied to repetitive tasks. And one of these areas, if we’re thinking about great use cases or even really interesting verticals, is the healthcare space. And everyone is familiar with this process. They go to see their doctor. And they’re speaking with their doctor. And they’re telling her, “Okay. I’m having this issue. And I’ve already taken these steps, so on, and so forth. What is the doctor doing?” Well, they’re taking down, hopefully, some notes. Now, one of the challenges is when you are taking notes within a system like an electronic health record system, like Epic, is you’re partially focused or the physician is partially focused on what you’re saying, I hope, but also they’re focused on how they’re taking that information and putting it into the system. Right now, the interface, it’s still a traditional application interface where you have to put information in the right fields. What does this do? It makes it much more difficult to have a natural conversation and really engage with that patient.
So, there has been a rise in the use of what they call ambient clinical documentation within healthcare. And essentially, what this does is it allows physicians to just dictate and basically carry on conversations and have that converted into written documentation. What does this mean? Well, it really means that the physicians are going to be able to reduce their administrative burden and improve their interactions with their patients. So, instead of focusing on inputting information, they can be having a conversation. The conversation is being captured. And then after the fact, all of that information is captured through this technology, summarized, and basically bundled up in such a way that the physician can make sense of it later on.
If you think about it, it’s almost like having a court stenographer who’s taking down everything verbatim, and then having an analyst come in and then pulling out, here are the most pertinent parts, here are the key terms that are being used, so on, and so forth, but having that done automatically in such a way that the doctor doesn’t have to spend a lot of time after the fact going in and manually compiling notes or going through whatever records that we’re keeping. I think as part of this whole discussion, I just want to throw out a few stats. I believe there’s a survey published by Athenahealth back in February that found that 90, actually more than 90% of physicians, have reported feeling burned out on a regular basis because the amount of paperwork that they need to complete.
And this paperwork certainly includes things like clinical notes. More than 60% of the doctors say they feel overwhelmed by these clerical requirements. And they work an average of 15 extra hours a week outside their normal hours to keep up. And this is really a challenge because it’s adding or contributing to burnout, early retirement for some doctors. It basically takes doctors away from doing what they want to be doing, which is treating patients rather than doing busy work or clerical work. So, what are the solutions here? Well, obviously, by using generative AI, they’re able to, again, just within the flow of work, reduce the amount of time it takes to document what’s going on with patients. And really, there are a few interesting companies out there that have been on this for the last couple of years here. One of them, Microsoft who acquired Nuance a few years ago, I believe it was in 2019, but don’t hold me to that date. They actually have a solution that’s called DAX Copilot. This is now generally available. I believe that there are about 200 different organizations that are using the technology. And really, what this does is this allows a doctor to basically have a conversation with the patient, capture all the information, have it summarized, and then it is automatically input into Epic, which is the electronic health record system.
This really reduces the amount of time that it takes to capture all the pertinent data, and then, of course, makes it available within the system of record that is not only used by that particular physician, but throughout his or her own hospital system. And then eventually, if a patient is being treated at another organization that is also on Epic, they can also access all of that information. So, it’s a really great solution for reducing the amount of repetitive work, and again, not only reducing the amount of work, but also creating a way to capture that information very accurately, which is not always the case. If you think of doctors who are overstressed, who are doing so much of this extra busy work, sometimes, mistakes happen. They’re human. Now, certainly, Microsoft is not the only company involved in the space. There’s another one called Abridge, which also integrates its ambient clinical documentation solution technology directly within Epic. They talked, I believe, at the conference that one of their big customers is a California-based UCI Health. I believe they are rolling out the solution system-wide, which is pretty interesting because clearly, they must feel it works well for it to be utilized across a wide range of physicians. Now, they were saying that their technology can save doctors up to about three hours per day. And it’s automating more than 90% of the clerical work that it’s used for. I don’t know if that’s true or not. I don’t have any verification information on that, but I can say that if you just think about it logically, it makes sense that it would reduce a significant amount of time it takes to take on these processes.
There’s another company that actually made some news there, Suki. They also are an ambient clinical documentation provider. This company was founded about six years ago. And now, I think the company claims that it is being used across more than 30 different specialties in more than 250 health organizations around the world. They also said that six large health systems have gone live with the implementation in the past couple of… excuse me, past couple of weeks. They also published some stats around ROI for the solution. They’re saying that it can reduce the amount of time a physician spends on documentation by an average of 72%. So, again, really, really strong metrics there in terms of using the technology versus more traditional way of capturing that information live and then inputting it into an EHR. So, let’s see what else. Oracle Health, they also made an announcement to their Oracle… sorry, Oracle Health Data Intelligence. They have a new generative AI service for care management that will summarize their patient history for care managers, again, to reduce the amount of time it takes to review a manual chart. It allows them to serve more patients each day. This service is now available, I believe, on a limited basis. So, it’s these types of tools that are leveraging generative AI in what I would consider to be a fairly safe way. And what do I mean safe? Well, because, again, the models are only operating or only interacting with very specific data, so you don’t have to worry as much about hallucination when it comes to things like if you’re trying to summarize what’s going on in an interaction. It’s only acting on that individual corpus of data.
Now, the other obvious question around this in terms of data and safety is making sure that there are appropriate data security safeguards in place. Obviously, HIPAA has a lot to do with that complying with those regulations, but in terms of the actual models, they’re not acting on data that might be or that should not be data that’s out in the wild. So, you’re not going to see, hopefully, any crazy hallucination. Now, that does not mean that you don’t need to go into this data and make sure that there isn’t any bias creeping in or any artifacts for whatever reason. You do need to keep checking that model to make sure that there hasn’t been any drift or that it’s still performing the way it should. So, if we look at the future of this, obviously, the technology is going to get better and better as more organizations use it and vendors continue to refine the functionality. I can see in the future that these applications go beyond just physicians, but even nurses in terms of their interactions with patients. I can see it being used cross-department or within teams of physicians that might be treating a patient to make sure that everyone is on the same page. Again, saving the time of physically or manually having to share information. I think the ultimate goal, of course, is really to transform the way healthcare is delivered, increasing the speed as well as the accuracy of the information, because, really, ultimately, that’s the big challenge that healthcare has is you have so many patients that have critical needs.
They need to take the time to spend with each patient. And it should be spent delivering healthcare advice or actually doing the delivery in terms of examining folks instead of sitting around and doing clerical work in terms of data entry and that sort of thing. Now, we talked a little bit obviously about the healthcare industry, but it’s not just healthcare where we’re starting to see examples of generative AI applications and use cases that are specifically focused on reducing busy work and manual work. I did a podcast not too long ago where it was with a company called Flowcast. And they were talking about using generative AI to really help reduce the amount of busy work around accounting. So, if you think about the accounting profession and the month end close and that sort of thing, there’s a lot of manual data manipulation that needs to go on. And it is a frustrating situation for folks who work in that field because that’s not why they became accountants necessarily to do repetitive work. They wanted to actually get into looking at strategy, being able to actually take and look at numbers and actually see, hey, where can we change things? Where can we take all of this data and then make recommendations to business leaders to actually affect change, to make the business more profitable, to reduce costs, that sort of thing? And really, generative AI is a good tool to help them reduce the amount of manual work so they can focus on some of those other tasks.
And in addition to actually reducing the amount of time it takes to close a month and that sort of thing, it’s also really important in terms of making sure that the workers who are in the space stick around because if all they’re doing is this very, very boring, repetitive mind-numbing work, they’re not going to stay in the field. They’re going to go somewhere else and do something else. And that’s going to be a real challenge as we move forward in terms of making sure we have enough people to fill various roles because regardless of the power of generative AI, we still need to have some humans there, I believe, at least over the next several years, to actually handle this type of work, if anything, just to be a conduit between business leaders and the data itself. So, I think that’s a real opportunity for generative AI. And on this podcast a few weeks ago, I also spoke about field service applications increasingly incorporating generative AI to, again, handle the low-hanging fruit of summarizing the interaction when a field service worker is out at a customer’s location, or immediately serving up product, or cross-sell or upsell opportunities, all of those things that are not core to what the worker has been trained to do, but are absolutely integral in terms of making sure that the things that need to be done are completed as well as, hopefully, enhancing that in customer’s experience by making the process smooth, making it quick, and making it easy.
So, in the end, I think that we’ve been talking on this podcast and others a lot of that generative AI and some of the, I guess, pitfalls around hallucination, looking at things like text-to-image generation, all of that kind of stuff. And yes, there’s still obviously concerns about generative AI there, but I think the point to hammer home here is understand that when appropriately deployed on a very specific task and making sure that the data is grounded or the model is grounded in data that has been vetted, there could be a lot of benefits with generative AI in terms of really reducing that busy work that nobody wants to do and ultimately shouldn’t have to do because the technology is getting to the point where it is good enough to accomplish those tasks. Okay. Now, before I move on, I just want to call to everyone’s attention something really interesting that came across my desk in this past week. There was a press release that went out in the middle of the month from a company called Dataiku. I believe I’m saying that incorrectly. It is Dataiku. They’ve actually launched a generative AI cost-monitoring solution. And apparently, this solution is designed to give enterprises real-time oversight of their spending on generative AI. I think this is a really interesting area or segment in the market given that we’re starting to see organizations deploy generative AI both through SaaS providers where the AI is embedded in the product as well as on their own as they’re building their own models or using other models, but building their own applications. Because if we think about generative AI and the actual cost involved, there hasn’t been a lot of talk about it. And the reason is because a lot of the vendors, they’re almost like ice cream trucks that are rolling around the neighborhood or candy vendors.
They want to get folks, users, hooked on the technology. And as a result, they haven’t been pricing it necessarily in a way that is fully transparent in terms of, “Okay. To do X, like generate an image, here’s exactly how much it costs.” They may set up an abstraction layer, like calling number of tokens, that sort of thing, which is great, and it is assigning some sort of a value to a generative AI task as opposed to just saying, “If you want generative AI functionality, here’s a seat license. It’s an extra $30 a month or an extra 50 a month, and so forth.” So, I think what we’re going to start seeing are companies like this who are starting to shine a little more of a light on how much it actually costs to use LLMs, and providing that transparency and visibility in terms of comparing different providers. I think it is really interesting to see how, as we get further on down the road here and starting to see enterprises really take a close look, they’re going to want to be able to do things like tag and track all of their expenditures around LLMs, and then assign the cost for each specific project because you want to make sure that you’re not spending $100 here on one, and 10 here, and your priorities are actually the other way around.
The other thing that a tool like this and what I think folks are going to be looking for is a way to distinguish between production expenses around generative AI and development expenses because really, you need to know how much it costs to set things up, but then on an ongoing and recurring basis, how much is generative AI really, really costing you? That is going to be really inform what ROI you’re actually able to generate. And then, of course, you want to have some a red-flag system set up or a warning system to look to see where you might have cost overruns, either due to governance failures, that sort of thing. You need to make sure that you have at least an assessment of or a system set up to understand where there might be risk. And I’m thinking specifically in terms of if you have a specific use case where you may project a certain amount of utilization, but then it goes way over because of whatever. I think that’s important to make sure that there are systems in place to do that. So, you’re catching that, and you can actually flag, and then put a pause on the system before it gets totally out of control. And then, of course, the most important thing is, of course, looking at getting as much data and insights from looking at usage as you can, and having a platform that is actually going to look not just at one model or one provider, but across all the ones that are being utilized because then you can identify, okay, which models are most appropriate for which use cases?
It may not be an LLM for every single use case. You might need a smaller language model for one thing, or even just you might want to have a trade-off in terms of capability versus cost in others. So, all of that is really interesting. I think it’s an interesting category of applications that goes along with the other what they call a digital adoption platforms out there, like your WalkMe’s of the world, that will actually also look and see how are you using or what is your utilization of… or your application stack as a whole, and see are you using all the licenses you have here? Is there overuse here? Do you have basically in the spot of your organization? All of this stuff is really important as we start to see tightening in terms of spending on technology. And I think we’re going to have that moving forward over the next couple of years because there’s still a bit of uncertainty around the economy. And tech is a great place to really start tightening the belt because you have the metrics. You can see what’s being used and what isn’t. Okay. So, with that, we finally come to our rant or rave section. And today, I actually have another rave. And this, again, comes out of the HIMSS conference. And this rave is around a new consortium of healthcare companies that are forming something called the Trustworthy & Responsible AI Network. I think the acronym they’re using is TRAIN, T-R-A-I-N. And this is designed to help make safe and fair AI accessible to all healthcare organizations.
So, what is this really about? Well, let’s see here. There’s a bunch of different companies that are involved in this, or healthcare organizations, everything from Admin Health to Boston Children’s Hospital, Cleveland Clinic, Duke Health, Johns Hopkins Medical, Vanderbilt University Medical Center, on and on, and on, and on. And I believe Microsoft is the enabling technology partner. Now, what are these guys? What are these companies trying to do? Well, they’re going to try to work together to basically do a few different things. They want to share best practices around using AI in healthcare settings. This is extremely important. They’re going to be looking at the safety, reliability, and monitoring of AI algorithms as well as the various skill sets to manage this AI responsibly. I don’t really have to even elaborate on this. This is so important, particularly when we’re talking about things like healthcare data, patient data, all of this kind of stuff. There has to be a concerted effort, industry-wide effort, to do this stuff because I could see a massive, massive problem if this is not managed early on in AI’s, in generative AI in particular’s early days. Now, of course, one thing that the consortium says is that data and algorithms are not going to be shared between member of organizations or third parties. And that’s important to note. What else are they going to do? They’re going to enable the registration of artificial intelligence that is used for clinical care or clinical operations through a secure online portal.
They are going to be providing tools to enable measurement of outcomes associated with the implementation of AI, including looking at best practices for studying the value of AI methods in healthcare and the leveraging of privacy-preserving environments. This is really, again, just trying to make sure that there is a multi-organization consortium dedicated to looking at these very, very important factors. And then finally, they say that they’re going to be facilitating the development of a federal national AI outcomes registry for organizations to share amongst themselves. This will capture real-world outcomes related to the efficacy, safety, and optimization of AI algorithms. Again, really important because when we’re talking about artificial intelligence and we’re talking about healthcare, we’re talking about safety, we’re talking about privacy, and security. And all of these things are… it’s tempting, if things go wrong, to close the door or sweep it under the rug. By setting up a consortium about this, it’s demonstrating that the industry does care about this stuff, it is important, and that they’re willing to work together to develop these best practices and essentially set up a system to shine a light on here’s what’s working, here’s what’s not working. So, with all of this, I would just say this is definitely a rave, a step in the right direction for utilizing AI in a safe and responsible way. So, that’s all the time I have for today. So, I want to thank everyone for joining me here on Enterprising Insights. I’ll be back again next week in another episode focused in on the happenings within the enterprise application market. Thanks to everyone for tuning in every week. And be sure to subscribe, rate, and review the podcast on your preferred platform. Thanks. And we’ll see you next time.
Author Information
Keith has over 25 years of experience in research, marketing, and consulting-based fields.
He has authored in-depth reports and market forecast studies covering artificial intelligence, biometrics, data analytics, robotics, high performance computing, and quantum computing, with a specific focus on the use of these technologies within large enterprise organizations and SMBs. He has also established strong working relationships with the international technology vendor community and is a frequent speaker at industry conferences and events.
In his career as a financial and technology journalist he has written for national and trade publications, including BusinessWeek, CNBC.com, Investment Dealers’ Digest, The Red Herring, The Communications of the ACM, and Mobile Computing & Communications, among others.
He is a member of the Association of Independent Information Professionals (AIIP).
Keith holds dual Bachelor of Arts degrees in Magazine Journalism and Sociology from Syracuse University.
