Making Heads or Tails of Gen AI Platforms

The advent of AI advancements have created quite a stir in healthcare. This is why we wanted to sit down with one of the experts of all things data, Fawad Butt. Fawad has built core data and analytics capabilities at 5 Fortune 500 companies and at two of the largest healthcare organizations in the world - Optum and Kaiser Permanente. As an early CDO - before the role became ubiquitous - he helped lead the data trend, and is recognized across the industry as an innovator in the data management, data ops, and analytics enablement technologies. He now takes on his next challenge in launching a company of his own.

‍

Polina Hanin: First and foremost, congratulations on deciding on starting a new stealth company in the GenAI space! You've been in healthcare for quite some time, but this is your first foray as a startup CEO. What led you to this decision?

Fawad Butt: The timing is right, and it goes back a little bit to my last role at, where I was the Chief Data Officer of Optum, and before that I was Chief Data & Analytics Officer for UnitedHealthCare. I had a lot of people working for me, which is interesting and satisfying foremost people, but wasn't really for me. For me, it was more about what we are doing and achieving with this team and technology that we're spending hundreds of millions of dollars on in a year. My team was in a cost center, so we ran aP&L, and we were very profitable. So I'd effectively been in a leadership role before where I was responsible for technology and teams that were delivering a lot of the backend processing that tends to happen in payers. The way Optum is set up, it supports the processing that UHC does, but it has a division called OptumInsights, which does a lot of these tasks for other payers across the industry.

What I learned was that we were really doing a lot of “break-fix.” What that meant was we're just trying to keep systems up and running and making sure that the industry didn't fall down if something went down. But if they went down, it wasn't just one customer that would be impacted, it would probably be an industry. You saw that what happened withChange Healthcare.  

As I was going through that journey, I built a team that was really focused more on thinking and looking at what comes next in terms of technology to solve these kinds of problems. And their job was to do about 70% research and 30% project work, file patents, use the latest technology, figure out what could be done. We were able to bring some new ideas and new technology to the table, but it was, in all honesty, incremental. So the challenge for me was: I had the team and the money, but the technology just wasn't there at that point to be able to solve the most complex problems. 

Technology has really evolved over the last 24months. And since the advent of GPT, since OpenAI introduced us to large language models and everybody got to play with them, we realized that this was the level of technology and intelligence that we needed to be able to solve these kinds of problems. It was time to take another shot at it.

PH: In what way is your new stealth startup addressing large challenges in healthcare? 

FB: Our purpose is to make sure that if and when any of us need healthcare, that you can get care as quickly as possible.Today, that's not how it happens - a doctor tells you what care you need, and then that gets pushed to an insurance company to get that approved, which can take days or weeks. The reality is, if you're in pain and suffering, an hour could be too much. So for me, the purpose really is to be able to take the clog out of the system. And the way we could do that is by optimizing the backend processes that create it in the first place - things like improving prior authorization, claims adjudication, risk adjustment, payment integrity or the things that are workflow and administrative in nature.

‍

PH: There's a lot of companies that are leveraging AI to various degrees and everyone is trying to figure out how to do it. What sets your approach apart from the others who've been trying to do interoperability for many years, or are trying to do this maximizing AI? 

FB: My team and I used to work at Optum and 11of the 13 of us were the people trying to solve these problems at the largest institution in the country. We have the expertise and the experience in trying to do it in the classical sense that the way everybody had been doing it. This is not one of those businesses where you can come up with an idea and come and disrupt an industry because you have to understand the process that underpins everything.

The second thing is that we're looking at it from an enterprise point of view. We're not trying to put an AI wrapper around a current workflow. All this does is create sprawl in a world where you’re used to having a rules-based intelligence, meaning that you define the rules and you knew how the warehouses and data marts were going to perform because you wrote the rules. This is very different than in a world where you have generative AI, where it sort of thinks and processes information in a way that can be unexpected.

So to have that level of sprawl would require a complete different level of governance, oversight, compliance. We're not just looking at where the puck is today. We're really trying to see where the puck is going to be when enterprises start to adopt Gen AI. Therefore, we're thinking about an end-to-end platform versus wrappers around Open AI or other language models. We have language models as part of our platform, but in many ways we're language model agnostic. We've built our own language models, which we think are better than anything else that's out there. But that could change in six to twelve months - another version of OpenAI or Lama could supersede that. 

The idea is to have a platform that thinks about all the things that enterprises need before they can deploy Gen AI, which means having connectivity to all the systems that they need to connect, having the ability to refine their own data, and having a governance layer that allows them to understand what data is being fed to these models.

You need all of these other layers in place, and that's what we think is differentiator as we move forward.

PH: InJuly, Rock Health published their quarterly funding report where they said that $1 in $3 in the first half of 2024 went to the companies that are leveragingAI. As the noise starts to pick up for enterprises, as somebody who has been within those types of organizations before, how would you guide a payer or provider on how to evaluate these companies?

FB: I think the buyers are all over the place right now. You could go talk to ten payers and you would see ten different levels of maturity in their readiness to be able to adopt AI. Readiness means that you have looked at the use cases that are most meaningful to you. Then you have to look at the data that is going to be necessary to leverage those use cases. Then the third thing, is you have to look for plasticity in the processes that support different areas.

So if you don't look at all three of these things together, it's going be challenging because you might come up with the technology that you bought, but you won't be able to adopt it into the processes. Or you may be able to have a process that's ready, but you may not have the data that's ready. It's a three-dimensional puzzle that sort of has to be solved collectively and together. My advice is to think about a crawl / walk/ run strategy here.

PH: You've worked with a lot of your team before and we all know successful companies are built by really strong teams. What's your personal philosophy in building up a team to tackle that charge?

FB: You have an understanding and a vision of where you want to go. Then you have to balance it between expertise that is foundational and necessary. You can start to do the work with the level of understanding that's necessary, and then you also sprinkle in ideas that perhaps weren't available in that industry from others.

I'll give you an example. I was on a plane ride today and I was doing an analysis on where I would hire sales talent from.Somebody would say, “well, why not go hire from Optum given your network there?” And my thinking was if you're building a platform, a better place to go hire sales talent might be from Databricks or Snowflake because those are the last two platforms that were built that are leveraged at the industry level. So they have the ability to think about it in a very different way.

My philosophy is that you have to have a foundational core of people who understand the problem, who have the ability to solve it and know how it's solved today. But then you have to bring in new ideas by bringing in folks who may need a little bit of support, but they have the ideas that are new to this industry. That’s the team that I believe gets you to a different outcome.

PH: Outside of building the team, what have you been proud of achieving thus far?

FB: I think we’ve done well on the capital front. At this point it's about finding the right customers who have done a little bit of the homework already. They've done some analysis and figured out at least what their use cases are, and at least have some sensitivity and understanding of the plasticity of their processes. Because AI is going to be different and it's not just plugging it into your existing process. I’m hoping we can find the right customers, which are likely mid-to-large payers, where we're offering what they need.

PH: Over the course of the next 18 months, what are you most looking forward to as it relates to your new stealth company?

FB: My goal is to be able to figure out what the right pulse and appetite is for Gen AI in healthcare today from buyers, and then to be able to deliver the solutions that our customers need. I'm excited about technology and what it can solve, but it's only meaningful if you have the customer appetite that aligns with the technology you have.  

I started my career in deep tech at GraceResearch, a supercomputer company at Silicon Graphics, that was like the Nvidia of the nineties and early two thousands. I've seen technology being way ahead of the customer - you have to always be a little bit ahead, but you can't leave them behind. You have to have the ability to slow down so you can go faster on the longer term.

PH: Thank you for sharing your perspective.