Phil Fersht — HFS Research[00:12]
So we’re very excited to have these amazing conversations with industry leaders, and somebody who I’ve personally known for probably over 10 years now, in his roles with Infosys and Mars, and now with UnitedHealth. One of the first people I know who has been all over this, has been very, very deep in this for a long time, is Sandeep Dadlani. So Sandy, everyone’s talking about the shiny new object of GenAI and trying to drive a narrative into the industry. But how does this factor into your thinking? How do you see this changing your business and the healthcare industry?
Sandeep Dadlani — Chief Digital and Technology Officer, UnitedHealth Group[00:41]
Yeah, no, I think I was about three months into the role where this whole trend exploded. And so we’ve actually taken a pretty interesting view on it. We have — I call it a pragmatic, exciting, and responsible approach towards generative AI. And so we want to be clear about what are the superpowers of generative AI, and basically it does three things. It does great synthesis, extraction from documents, structured, unstructured data, fantastically well. It does content generation, which means that it can generate letters, essays, chat prompts, etc. very well. And then it automates code writing. So it does automate software development, which I think is a fantastic superpower. But that’s it — those are three things it can do, and it has its kryptonite. So the pragmatism here is we aligned with all my business leader peers on what are the top 14 to 15 problems we’re going to solve for the business. Defining the problem has always been the first step for any exciting technology. And so we picked up some great problems. Bottoms up, we have 500 use cases on generative AI being piloted, but top down, those 13 to 14 use cases, we made it a fun, celebratory event. We called it Tech Tank across the enterprise. We have tens of thousands of people engaged on it, and we’re driving a hard business case view of scaling those use cases very successfully. Most of them are in the administrative area of healthcare, where I think there’s so much low-hanging fruit. Right now I have thousands of call center advocates using auto-call summarization. It’s a very simple thing that has eluded the industry for a while, but it just really eases the work for our call center advocates and has them focus on caring for the person who’s calling. Or even simpler things like appeals letter decisions. We get thousands of appeals for claims and other things from our members, from our customers. And the letter that communicates the decision to the appeal has to be written in plain, simple English language. Well, obviously we have tools that can do that very effectively. Other areas: how we look at unstructured clinical notes, doctor notes, and interpret data from there and present it to other clinicians to take decisions. All throughout, we are not replacing the clinician or the pharmacist or the doctor, but we are helping make their work easier so that they can operate at the top of their license. I mean, frankly, this whole idea of GenAI and healthcare needs a huge dose of responsible AI — the extra focus on fairness and bias, explainability of how the AI came out with the decision, and always having the human in the loop. I think at our scale, we feel very good about how we are scaling these use cases.
Phil Fersht — HFS Research[04:11]
So let’s shift gears a little bit to the IT services business. You know, I was talking — after a quarter of a century, we’re finally going to see, I think, a flat growth period for the traditional services industry. Do you think GenAI is going to act as a catalyst to drive it forward, or do you think this is going to completely change how we acquire and consume these services in the future?
Sandeep Dadlani — Chief Digital and Technology Officer, UnitedHealth Group[04:11]
So I see two types of behavior here. One is the threat that generative AI, especially with code automation — all the early pilots we have done show that there’s, I don’t know, 20, 30, 40% productivity to be gained, and that is apparent to clients as well as IT services companies. So then, who takes that 20, 30, 40% productivity? Should clients demand it? Is this an iPhone moment where all you’re going to do is take 36 Kodak film pictures and get them cheaper on the iPhone, or are you now going to explore it to 2 million pictures — i.e., you can deliver more technology much, much faster than ever before? And that’s the interesting paradigm. Saying this is an iPhone moment, which means all these tech services firms should be saying to their clients, hey, we can deliver double, we can deliver faster, and who doesn’t want more technology faster? There’s enough technology problems to be solved. The other part, which I think is interesting, is for the first time, this gives us superpowers to interrogate data very, very differently, and not just wait for the standard dashboard report to provide the standard answers. Now we can ask questions of our large swaths of structured and unstructured data, images, clinical notes, and really get deeper and deeper insights. I think the sky is the limit in terms of the possibilities there. So there’s an opportunity and a threat. If you look at this as an iPhone moment, then I think the opportunity is tremendous, but it will need all tech shops, including my own, to really, really get on the learning journey and have thousands and thousands of people trained on responsibly tuning LLMs better.
Phil Fersht — HFS Research[06:15]
So LLMs are clearly the flavor of the year, but as we look at maybe a couple of years’ time, where is this leading? What do you think we’ll be talking about next? Where do you think this exploration on GPT-4 and advancing technology is gonna take us?
Sandeep Dadlani — Chief Digital and Technology Officer, UnitedHealth Group[06:37]
On every job, in every task we perform, whether there is something aiding us, making us faster, more productive, more mindful, is perhaps a future that I’m looking forward to. That means elevation of work, elevation of everything that I do, really elevating our focus on the business of our business. My business is healthcare, which means that if all employees of UnitedHealth Group are focused on care, the caring part of health, then that’s better, then all the busy work goes away. That’s what I hope to expect in this journey. Also, the pace at which this is improving is dramatic. I mean, Phil, if you and I go into our Twitter feed right now, today, there’s been some amazing discovery of a cheaper, better LLM that just does more magic than what they did last week. So the learning challenge is even higher. So, much faster learning, which means you don’t have to take the classic 300 AI courses — you learn every week a new skill on an ongoing basis, and elevation of work so that we focus on the business of our business. So I’m excited about how generative AI makes us focus on our mission more than ever.
Phil Fersht — HFS Research[08:08]
This has been fantastic. Really enjoyed getting some time with you again.
Sandeep Dadlani — Chief Digital and Technology Officer, UnitedHealth Group[08:08]
Always looking forward to it, Phil.
