Building an AI Strategy: From Idea to Implementation

Jonathan  0:03  

Welcome back, everybody to another episode of the strategy gap. We're in for a great conversation today on something that is probably on the top of everybody's mind and that is AI and how to think about AI and your strategy. Joining us for the conversation is Elisa Esch. Lisa is a leading healthcare strategy leader with extensive experience in strategy, innovation, transformation, generative AI, health equity, digital health and more. She is currently Senior Vice President and Chief of strategy, innovation and Industry Solutions at NTT DATA. She actually began her career on the healthcare side as a registered dietician. before migrating into the strategy world. She spent time in senior executive roles across various healthcare and technology organization, and is a frequent speaker on topics like general AI and healthcare transformation. Lisa, welcome to the show.

Lisa  0:52  

Thank you. It's great to be here today.

Jonathan  0:57  

Awesome. Well, before we dive into, really the the meat of the conversation around AI, I do want to spend a little bit of time on your career journey because it is unique and not something that you see all the time in the strategy space. So I love just briefly walk us through how you began your career more on the clinic clinical side as dietitian, and eventually you know, what landed you where you are today.

Lisa  1:17  

Yeah, I've had a really fun career starting off as a clinical dietitian, I love nutrition, biochemistry, those things about how the body works. And it was fascinating to start my career in that my first job as an RD was actually helping to launch an eating disorder program back in Omaha, Nebraska. And so launching ideas, launching programs working in in health care was where I really started, I worked with diabetes programs, I worked as a clinical dietitian for a while in my career and then pivoted to more of a program development role where I led launching a pediatric service line and a health care system. And from there really just got the bug for looking at opportunities, what is out there, and how do we grow and do those things. So building ideas, and putting them into place, I pivoted from traditional health care in a health care system, which I've worked for multiple health care systems, and went to a really interesting organization Healthgrades. At the time was a big transition for me, we focused on quality outcomes, patient engagement online, and they were really far ahead early on when they launched in patient engagement and the impact that like quality and ratings has on health care. And I spent almost eight years there doing that, and really learning about the digital health world and how consumers make decisions about health care and what to do. In between that I also did consulting and advisory. And so I had this mix of background that really helped give me broad experience to lead into strategy, but from Healthgrades I went to a large healthcare IT business help launch call. Sorry, just dropped. We're gonna have to repeat that. No,

Jonathan  3:17  

okay. Yeah, well, started where you want to edit it? Yeah. So

Lisa  3:20  

from health grades, I went to a large IT service business and help launch population health initiatives. And that was a big change. For me, that's really what I got into the IT world, the technology world. And it was still though tied to clinical clinical outcomes, helping patients and it was a fun transition for me. And through all of that, I've landed where I'm at now, with, with NTT DATA where I lead innovation, our strategy, our offerings, and it's a really interesting place to be today. Because throughout my career, and all the different roles that I've had, every time I've made a change, it's it's been towards building something new, bringing new ideas, how do we do that? And so honing those skills, and really learning the how to develop strategy and being mentored by great people. And working in really awesome organizations has helped me do that. And so today, I spent a lot of time on really looking at the market, I'm What do our clients need? How do we help them develop strategy? How do we put our own business strategy together?

Joe  4:30  

So it sounds like the key component to all that because of all the different diversity of experience, the key thru line would be curiosity. I think, I think, would you agree like I mean, basically, if you're into biology in any sort of like, how does the body work and all that I think it lends itself well, and when I'm hiring people, I always ask them, Are you like a naturally curious person because you're like, there's so many things we don't know. And there's two types of people people that are okay with not knowing and people that will at least make an attempt. Would you say that that has played a major role is that a key come? On end of your journey, that curiosity leading the way,

Lisa  5:03  

Joe, you, I would have thought you talked to my dad. So I'm curious person in the world asking questions, always asking questions and trying to figure things out. And I'm very, very curious. And and I think that really is a core characteristic of someone who, who wants to be forward thinking wants to not just do the routine and really thrive in that world of going forward. Very curious, I ask questions all the time. But one of my favorite things to do is actually like tackle those big problems that people are stuck with. And they may have more of a narrow point of view, I don't bring necessarily a technology point of view. I'm not a an engineer. And so when I go to tackle problems to figure things out, I'm bringing this different point of view into the technology world today, which I think helps us in thinking beyond where we might be focused. But curious is a very good word to describe me.

Jonathan  6:06  

Yeah, and I think that perfectly explains how you got to where you are today with tackling one of the biggest problems that a lot of organizations are thinking about, and that is, what in the world do we do with AI? It's all the rage today, whether it's referred to as AI or generative AI or machine learning, most organizations are at some sort of crossroads. They're trying to understand it, you know, what do we need to do? Do we build it? Do we buy it? Just pretend we're using AI? when really it's not AI? Or do we just kind of hold our breath and hope it goes away, and we don't have to address it? So I guess to start, I'd love to get your perspective just on the general landscape and perception of AI in your mind what truly is and isn't AI or generative AI? Yeah.

Lisa  6:49  

So Jonathan, that's a great framework for this conversation. So, you know, Gen AI  was not born in January, it's been happening for quite a while organizations, large organizations have investing in it, but most of the world is tossing around terms. Chat. GPT LLM is foundational models, Gen AI, conversational ai, ai, ml, you know, all of that, like, what is it really? And I like to start with the perspective of what is generative AI? And it really is this, this process of taking information? Putting it together data from other kinds of data sources? And really focusing on creating answers is, is how I would say it. So how does it generate new artifacts, so it takes data and, and information, it trains itself, it learns it has fast sources of data and information. But the key to generative AI is it is generating new things, new ideas, new pictures, new, you know, new thoughts, new concepts, those kinds of things. Most of us are very familiar with conversational AI, the scripted AI, the Chatbot, we're very comfortable with that. And, and the best way to think about that is when you ask a chatbot a question, and it says, I can't answer that I don't have enough information. It's not in the script. So you know, that is we're all very comfortable, very used to interacting with with conversational AI. And there's a really good place for that, because it can take out a lot of the just the, the human piece and the and the barriers to actually get into answers quickly. So there's a really great role for that. But beyond that, then is where that those datasets that information is trained to create new output. And from there, what you have is really the foundational model. So there's multiple foundational models you have for Google, you have the AWS has won, Microsoft has won, you know, there's lots of foundational models in there that are really taking disparate information, again, that data and their their training and learning and creating content. The foundational models create more than just language content. They're broad, they can do the pictures that you know, the imaging and things like that. And But what a lot of organizations are focused on right now, though, are the LLM the large language models. And this is where AI is trained on text. It's trained on you no words and information. And it really is focusing on how to create this human like interaction via words. And so that's what most of us are, when we're exposed to Gen AI right now, it probably is tied to an LLM. Because that's where a lot of the the training and the models have come from. And there's a place for that. There's a lot of value that comes in that and there's lots and lots of organizations out there right now. Free Getting single point solutions tied to MLMs, solving very specific problems across the ecosystem of business of human interaction of consumer interaction and patient interaction. And then the big one is obviously chat GBT. It's got a foundational model, it has MLMs in it. And it really is trying and it has conversational chatbot in it as well. So it has multiple sources of AI coming together, and creating content and language. So that's the framework for what you know what the AI noise is all about. That can be overwhelming. And I think you were talking about, you know, where to get started, what do you think about what do you do? And I think it's important that we really are thoughtful about how we use Gen AI in particular, in our business in our life, and how we interact because it's so easy to just like, go to go to an open an open, I go to chat GBT and start interacting. From a business standpoint, there's some risks associated with that unless you've got a closed foundational model. And if you're using the open one, there's some risk associated with that with IP, etc. But for the most part, you know, that's the framework about what it is.

Joe  11:30  

And that's helpful. I mean, when we're thinking about who's going to be first, who's going to be second, I mean, the thing that comes to mind, we love case studies here in the strategy guy, IBM Watson, right, because we're talking about healthcare. I remember I remember how excited I was about the promise of that, right when they tried it out on Jeopardy, and we're like, your doctor is going to be replaced, and you're not going to see a primary care physician anymore, because Watson will do it for you. And we all know how that went. Basically, that whole division has been mothballed. And I guess there are a lot of smart people out there going to help IBM, the IBM can't figure it out, then what chance do I have? So I guess the question is the idea of buy versus build, is it? Is it? Is it something that at this stage? If you haven't started it already? Are you kind of far behind? And it's better to partner? Or is there still plenty of blue, blue space there blue oceans for you to be able to carve out your own destiny? Because it seems like a lot of people have made a lot of progress in the last two years, and it might be too hard to catch up.

Lisa  12:25  

So buy versus build. This is the age old question. Really every organization asks all the time, particularly around IT services? Do we just, you know, do we do this ourselves? Or do we use partners? My advice in this space is pick a foundational model and partner, you this is way out of most organizations wheelhouse to really understand the complexity of this you need to be partnering with with organizations that have deep deep experience with security, compliance, law, you know, the rules, all of those things with, with the ethics, the bias that happens in AI models, like there's there's so much complexity here, that I that you've got to pick a partner, and you have to pick someone that can help you build a strategy and move around it. What's interesting, I was at JPMorgan this week and attended the future in health conference, pre JPMorgan Sunday, before JPMorgan started in one of the speakers. We were talking, of course, talking Gen AI. And we got we actually got almost to the afternoon before we actually had our first gen AI conversation. So I was impressed that we focused on health, the first part of it and then move to the the buzz of the world right now. He had her like there was 300 people in the room and help these are healthcare executives, senior executives all across the country. And he said, How many of you have Gen AI projects now at 90% of the people raise their hands. And then he said, How many of you have a strategy in like a strategy for your gen AI like for people raise their hands. And so the risk isn't necessarily to buy versus build. I think most organizations are just kind of going after shiny some of the shiny objects and trying to solve point solutions or very specific, narrow problems right now, which Gen AI can do, particularly the LLM's can do that because they're trained for specific things. But what's happening is you're going to have this fragmented application and you're going to have like, single tower apps, and you're going to have all these things happening in your organization that eventually are going to be hard to manage and hard to really understand and value. And so our recommendation is really you've got to get your governance going. First, you have to have a governance you have to have the ethics and compliance defined you have to you have to have new people on your team constantly engaged with what you're doing in this space because As a human interaction, you've got legal, you've got compliance, you've got your, you know, your –. And they're going to be more involved with this type of application in the ecosystem than maybe other types of applications in your system. Because there's this, you pick your problem you're going to solve. That's where you start, what is your problem you're trying to solve? And what does solving look like? What is the value that you come? How are you going to measure that. And then you want to pick your partner as part of that. So you'll probably pick a strategy partner. And then you want to pick your foundational model partner, and then figure out how to set it up, how to manage it, how to govern it, how to keep it safe, how to keep your organization safe, how to keep your customers safe, and then build out your, your, your prototype, What's your concept. And when it comes to that piece, the other thing that I really see a lot of is just this, like, again, point solution, point, solution point solution. And again, that's okay, you can get small impact with that sometimes you can take care of your most stressed, like physicians, like if you're in healthcare, it's like this whole physician piece is big. But the other thing we would challenge you to do is take a step back and look at Where's Jenny, I gotta have the biggest organizational impact. And it's not going to be from a single point solution somewhere, how do you really help transform your business to the future, using this technology in a way that makes your business more efficient, better, stronger, and really help you grow and move to the future. So there's two tracks, there's the big transformational play the Gen AI can take care of. And then there is the single point solution where you have significant pain points you need to solve and you can solve them immediately. With each of those use cases, you will want to, again, define the problem you're solving. What is that ROI? How will you measure it was actually successful? How do you build a prototype? How do you test it? How do you tweak it, make sure it's still compliant, your team has to go back and then execute it across your organization. So those are some of the guidelines and the lens that I would check take on Gen AI with and I know it's how we're tackling it as it NTT, our own company, as we start to use this type of technology in our own business.

Joe  17:29  

It's interesting, you mentioned the fragmentation because I mean, you probably know better than anybody, the number one EHRs came out, and it was there was a no no, like close to 100 companies, I'd say it was in the pharmaceutical world at that point. And then over time, there's no two, maybe three, or three. And so the idea there the thing, though, is that, at least for me, and I would love for you to shed some light on this when it comes to, to AI. We all know that TPT is fantastic. It does some really interesting things. Mid journey is something pretty interesting with the images and things. But there hasn't been a singular, just killer application for business that you're like, yes, that is now the thing. And I guess that maybe just by my naivete, I haven't seen it. Have you seen some application out in all this wonderful work that you're doing, you're like, This is going to change this industry or change the world in a way that maybe the average listener maybe wouldn't know, from just knowing the name the general names like Jaggi potatoes, or someone that we don't know a sleeper that we should keep an eye on, I

Lisa  18:27  

would say that Gen AI as just a capability is going to be that. And you will have multiple organizations, each one has a strength. And so one of the part one of the things that's really important in the strategy is when we talk to an organization, where what are you really trying to accomplish? How do we get there? And then how do you pick them the foundational model that's going to get you there that in the best way. And so there isn't going to be a one size fits all, there's too much good in all of them and investment, that that you can that you can pick them you might have multiple models as well, depending on what it is you're trying to do. There isn't going to be one answer, this is too big to have just a single one that's going to take it all in, it's just too big. And I really challenge any organization that thinks they're going to build their own foundational model when the likes of AWS, you know, Microsoft, Google have invested billions and billions of dollars in this already. There's really no point in trying to do that. You can take those models, train, make them safe for your organization, do what you need to with the information and be very successful for that. So pick a model, pick a foundation model, pick a partner and build your strategy around that.

Jonathan  19:49  

Yeah, absolutely. I want to go back to the example you use earlier around kind of the question race or was how many people are using AI and how many people have a long-term strategy because I do think the natural inclination for organizations to say, we have to do AI. What's our biggest problem? Let's throw a point solution. And everybody has a consumer has experienced that you go into an application. And it's like, wait a second, this is an iframe of chat GBT embedded into their platform like, I don't know if that's really, really helpful with what I'm looking to do as a consumer and certainly on the organization side does not accomplish what they want it to. But taking that step, I think from a point solution to a long-term strategy is difficult for organizations. It's not just the framework that you laid out, but it's how are we going to think about our ultimate goals and how we want to grow up into this AI strategy and what we want to do to transform our business. So as either you're working within your organization, or as you're speaking with other leaders, how do you take that step from going from a point solution to actually thinking long-term about AI's impact on our business?

Lisa  20:50  

Yeah, that's a really great question, Jonathan. And I, the place to start, the conversations we're having is we've got a deeply look at like the business, what parts of your business are spent on just processing, you know, the coding, there's just the real basic transactional types of things. And that's, those are the things that you can really tackle with Gen AI to transform your business and really automate and, and do things. Now the one thing that I remember early on when the original bots came out, is that everybody bought bots. And they went in, and they automated all this, all these processes. And they did automation of things that didn't need bodies to do anyway, you could just automate it. And what happened is, they took very inefficient processes, automated them. And so then they could be inefficiently automating things at 10 times the speed, they did it before. So it's really important that what you when you look at what you're going to automate, that you you fix the process, if it needs to be fixed, and then apply the automation to it, you don't want to just like be automating things that are still not very efficient. So that was kind of a chuckle. That was what we saw a lot of in healthcare, by the way. So how you do that is really just looking under the business. And we run businesses in all industries. So we've got expertise. So we can say, here's the most likely places you're going to drive efficiency, this is where you should start. And we can make those recommendations based on just what we're seeing in the market and what we're experiencing. And then the things that are unknown, are a little bit more challenging. So part of what Gen AI is it's creating new content, it's creating new things it's creating, it's generating something. And so, in healthcare in particular, you want to be a little careful in some of what you're doing in that space, because you have some risk associated with what you're generating, if it's clinical, if it's patient and those types of things. So there's this balance, about business transformation, which is what every organization should be talking to their partners about, how do I do that, because if you're not doing it, your competition is and you need to get there. And the second thing really is that one of those other things that haven't been done before that maybe are is easy to do, and that are going to require a point solution specific thing built for you to really go ahead and do that. And in healthcare. The one thing I find really interesting right now, as we as we talk about, like the clinical workflow, clinical engagement of Gen AI, and some of the algorithms that we're seeing, for reading, you know, images for diagnosing, for treatment, those kinds of things for the clinical experience, there's this debate about how perfect we have to get the AI before we actually let it go do its job. And I think there needs to be an open discussion about how in medicine, people always practicing medicine. There's massive variation in how people practice and what the outcomes are. And I do want to explore those clinical things that we should be doing to hat to help patients to improve and impact and deal with capacity and those kinds of things. And I don't want to be limited by fear of having an algorithm that might have an adverse outcome when we have that every single day with human beings providing health care. So I don't want the bar to be so high for AI that we never get started in testing those things. That's my comment. Because anybody who's in clinical care knows that care varies. You know, you get a new a new grad, a new resident, a new a new physician varies, and even experienced ones clinical care. are various. And so we've got to have a reasonable bar for Gen AI to tackle these issues of access and care delivery in a way that allows us to still have an impact like we could. We can't be too scared.

Joe  25:14  

Exactly. I mean, you mentioned health equity. I mean, some of my graduate studies were in that same realm. And the idea that this could be helping a certain subset of the population that doesn't have access to this now, in a way that would benefit them more than by holding it back. And so that's a fantastic point you're making. And I guess the overall through line here is the idea that we're generative AI is right now is it does a good job at automating certain types of tasks. I think we all can agree on that. And I guess every company then is going to look at all their processes and say, Well, which one is the biggest rock in our shoe? And as a strategy leader, half of what you do is what you say you're going to do? And the other half is telling people No. So what advice have you found to like cast the net wide enough, where you're getting the right voices in the room, but also gently telling somebody? That's a good idea, but we can't do it. And you were gonna do this one over here without hurting feelings and causing those problems? How have you navigated those, those pitfalls, because they happen every day. Oh,

Lisa  26:10  

and gosh, no, is really the job of a leader as well. And you got to focus. So my, it's going to be challenging, and people are going to be upset that they don't get to try that little widget that the Oh, it's just this little LLM, just solve my problem. For me, it's going to be a huge challenge for leaders to really balance that you'll have 100, like ideas for for MLMs. And for jet AI to go tackle issues. And you could solve each one of those. And so as a leader, you have to just decide what are the biggest challenges? What are the biggest ROIs and you have to have an idea of a problem to solve, if helping to balance all that really is the ROI, what is the return? How are you going to measure that and then measure it and prove it? So that's a big key to what what you're going to do and how you say no, when I came on board that in my current role, part of part of what my job was, was really to define what are we going to take to our customers, and define that and be really good at, which then takes all the other things that we could do with them off the table. So is just as important about what you do with your with, with your roadmap with your strategy, is what you're not doing. And so you really have to look at impact outcome, the financial ROI, the culture change that people change that might be required. There's a lot going on with these capabilities and tools. So saying no, is a is a big part of it.

Jonathan  27:39  

Yeah, and it's certainly been a ton of insight shared throughout this conversation. And I think we could talk for hours on this topic. But unfortunately, we do have to close out fairly shortly. Before we do one last question, really to summarize it. And what we ask all of our guests is thinking about everything you've learned in your career, generative AI strategy, healthcare, if you had to go back to your first day stepping into your strategy role, and give yourself a piece of advice, what do you think that advice would be?

Lisa  28:11  

That's a great question. I think it would be to put people around you, that will challenge your ideas to make sure you're not so focused and so narrow, that you miss something. So I think the advice I would have is surrounding yourself with people that know more than you that will are not afraid to go to give feedback and be willing to take that feedback. So you can become a better you and really get the picture out there that needs to happen to develop strategy. So put smart people around you and let them challenge you. And

Jonathan  28:54  

that smart person may just be generative AI in the future. So great. Well, perfectly. So we appreciate the time, tons of really thoughtful insights that I know our listeners will appreciate and we look forward to future conversations.

Lisa  29:09  

Thank you so much. Have a great day. Thank you

Transcribed by https://otter.ai