Speaker 1:

From the New York Stock Exchange at the corner of Wall and Broad Streets in New York City, welcome Inside the ICE House. Our podcast from Intercontinental Exchange is your go-to for the latest on markets, leadership, vision, and business. For over 230 years, the NYSE has been the beating heart of global growth. Each week, we bring you inspiring stories of innovators, job creators, and the movers and shakers of capitalism here at the NYSE and ISIS exchanges around the world. Now, let's go Inside the ICE House. Here's your host, Lance Glinn.

Lance Glinn:

AI is no longer a side experiment in big Pharma. It's becoming the engine behind a shift in how medicine is made. By analyzing patterns invisible to the human eye, AI is uncovering hidden drug targets and optimizing treatment pathways with breathtaking speed. It's transforming every stage of the pharmaceutical pipeline from discovery and development to distribution and post-market monitoring. Eli Lilly, that's NYSE ticker symbol LLY, made waves last October by naming our first guest of the biotech series, Thomas Fuchs, its first ever chief AI officer. With a background that spans NASA's Jet Propulsion Lab, Mount Sinai, and co-founding the AI Driven Cancer Diagnostics firm, Paige, Thomas is pushing the company to embed AI across its organization. Under his leadership, Lilly is turning its massive troves of data into a competitive edge. Thomas, thanks so much for joining us Inside the ICE House.

Thomas Fuchs:

Thank you for having me. It's great to be here.

Lance Glinn:

In October, you were appointed Eli Lilly's first AI officer. Now, being the first in a role comes with the challenge and of course, the opportunity of setting the tone early on. How do you think about building not just AI capabilities, but setting the tone to build the right culture, the right mindset, and internal trust about using AI responsibly at Eli Lilly?

Thomas Fuchs:

Yeah, that's actually a very good point. First of all, it's a great privilege to be the first one to be in that role. There's a lot of trust the leadership team put into me in doing so, and I'm very appreciative of that. There's, of course, enormous opportunity. Lilly is at a very unique point in time where we can really leap forward in multiple areas, and AI is for sure one of these areas where we absolutely want to do that. One of the most important things, actually, maybe let's start really deep in the trenches, and I tell my team all the time is, machine learning and the AI, it's really the art of the failure. You try to model something, you fail, you try to model something, you fail, and you do that over and over and over again. Very often after months of work, you have nothing except a better understanding of the problem. It's really the scientific loop in practice. That's also why machine learning and AI is so successful. That's now part of that culture.

We have to do things that are very risky in terms of technological risk, always safe for patients, but technologically we are really at the forefront of what we can do with AI in building medicines, finding drugs, and that, of course, includes a lot of failure and that's part of the culture. Then, you have, of course, the usual staples with deep integration into the business, deep integration into the science efforts. We set up joint teams from discovery to manufacturing where we are all together in the trenches. You really have to understand the problems in depth to make progress with AI in these areas.

Lance Glinn:

You have a company like Lilly that is so steeped in history and innovation and success and discovery. What was the process like in inspiring executives and colleagues at such a legacy brand to buy into what AI is now, but also, the long-term potential of what AI could be?

Thomas Fuchs:

I had the enormous luxury to get actually into a place where that was already the case mostly. Diogo Rao, who did a fabulous job at Chief Information Officer to get the company up and running in terms of AI. Lilly was already using language model throughout the company in hundreds of examples where they're used for workforce improvement, workflow enhancement and so forth. Fabulous. Dave and Dan are enormously supportive and they saw the future already before they brought me on board.

What I can bring to the table is to really push the envelope even further. That, of course, means also bringing many more within Lilly on board. Again, to make progress, you really need, for example, discovery, the chemists on board, and of course, many of them in a very traditional company have done it one way very, very successfully. To be successful in the future, you might also want to do some things differently and take advantage of the AI tools we have. That means, of course, consensus building, a lot of explaining what AI can do and what it cannot do

Lance Glinn:

I want to begin really with AI's impact on pharmaceuticals. I'm curious how you think about just the magnitude and the pace of transformation as a whole. How would you describe the role AI is playing in just shaping the industry? Do you think we're at a point where AI is making fundamental changes to drug discovery and drug development, or do you think we're still a little bit away from that really taking shape?

Thomas Fuchs:

We are at the tipping point. Historically, I understand that people are skeptical. For example, in the discovery site, there are very few examples out there where you would have a really they never designed the AI drug yet, but that's changing drastically. It's changing because of, first of all, progress in, of course maths and machine learning. We have better models to approach these problems, but we also have drastically more compute. That, of course, allows us to do so much more. Supercomputers at the scale that weren't present just a few years back. That allows us to push that envelope. We have internal models that already propose molecules for some of our targets with new motifs the chemists haven't thought about before, which are very successful. I'm very, very bullish on that front.

Then, of course, there's much more to producing medicines than the discovery and development phase. We also do a lot of AI in development, in manufacturing, development as well, but also manufacturing. That's, of course, an area that's very, very important for Lilly. There are many areas from optimizing all kinds of processes from API production to digital twins of manufacturing lines to robotics where AI can have very, very concrete impact and where we already see enormous returns of investment now. Then, or course, in development for trial enrollment for executing trials more efficiently and so forth. Yes, the future is coming, it's changing very quickly, which is very positive because, of course, we want to get out more drugs for more patients as soon as possible.

Lance Glinn:

How do you infuse the trust factor in it? When you talk to the common person, there's so much information being thrown at them around AI, across all industries, not just pharmaceuticals, but how at Lilly do you infuse that trust factor into AI to make sure that the patients, the ones ultimately, really, the most important stakeholder, they trust that the AI at Lilly is using is 100%?

Thomas Fuchs:

Yeah, very good point. That means also trust, first of all, internally and then also trust externally with patients. You're also right, so that the dark side of the hype is, of course, that there is enormous fluff out there. Million companies who sell snake oil, a lot of confusion in that space. Looking back, I actually got my Ph.D. in machine learning at the time where it was not cool, where the largest conference was 400 people and it's, of course, nicely changed. Now, you have to actually fight that hype, and you can do that with a very sober outlook and a very clear vision of what we can do and what we cannot do.

We do a lot of education internally and training and giving talks and explain where the limits of the current models are. The current models don't have any will or volition wouldn't even know how to get there. These language models are next token predictors, which are enormously powerful, but will not get us to AGI. A lot of the fears are overblown. These models will not replace roles, they will replace tasks. For these, they are very, very useful and we have to guarantee that they're safe and effective and equitable for all patients. We do that from the beginning within development of these models, testing these models, tracking these models, having data provenance and then running trials or in the future, using them as endpoints. For example, in trials, you have to make sure, you have to do all these steps and you can do that.

Lance Glinn:

We talked a little bit about it before, most of the headlines tend to focus on AI and R&D when it comes to pharmaceuticals. I'm sure as you even mentioned, there's a much bigger story to tell across the pharma value chain of how AI is being used. It's not just in drug discovery, it's not just in developments. It's in things like manufacturing, as you mentioned. What else or where else are you seeing AI make a meaningful impact across the broad industry? You mentioned manufacturing, but as in things like supply chain, strategy elsewhere, et cetera?

Thomas Fuchs:

Yes, of course. First of all, of course, the current models are very good in language. Anything that touches language, we try to use AI. That means, for example, improvement in call centers, contact centers, internally, externally in preparing documents for regulatory bodies, helping us to answer, for example, questions to the FDA. Then, for example, internally for managing trials, trial enrollment at scale. Anything with language, these are no brain minors to touch. Of course, there's a lot of AI outside of language, and that's for pharma, that's actually often more interesting. One thing is, of course, the biochemical space we talked about that is discovered. The other one in manufacturing is, for example, so we use computer vision for quality control, quality assurance on the manufacturing lines. We use a lot of digital twinning at different granularity that help us to optimize processes drastically.

Then, of course, there's also a lot of time-dependent data. We have an AI team that, for example, looks at financial forecasting, all kinds of forecasting of more than a thousand profit centers within the company, Forex forecasting or time-dependent data, again, from machinery, so all the telemetry that comes out of robotics and these machines. There's really a plethora you can do with AI. We have a lot of teams in these different areas. It also makes it so exciting to be in a place at Lilly because it's not the one-trick pony. It's really everything. You can't drop any of these balls. If you would just focus only on discovery, you would immediately get blocked in development. If you only do development and not manufacturing, and you get to the bottlenecks in manufacturing. Then, at the end, of course, finance, but also effects of commercial. We do traveling salesman problems, help to contact HCPs and so forth.

Lance Glinn:

As someone who grew up in machine learning, now obviously serving as Chief AI officer at Eli Lilly, you mentioned there's so much excitement of what AI is doing right now, but obviously, what AI could potentially do in the future. With that excitement, obviously, comes a lot of skepticism too about the potential dangers of what AI could bring. How do you balance those and really try to appease or suppress, suppress might not be the right word, but try to calm some of those nerves, some of those worries and tell people that yes, obviously AI's future, a lot of it is unknown, but we here at Lilly know that where we're going to use AI in the long term is going to be beneficial for us as well as obviously for the patients that are getting our drugs.

Thomas Fuchs:

A lot is, of course, again, communication, education, constantly pointing people to the right information sources to have a really sober look at the whole scenario. I mean, we saw that before. When I was, for example, building the AI department at Mount Sinai or even before at Sloan Kettering, these were times when physicians were extremely skeptical about AI. You had to do that constantly to explain that yes, we can contribute something to cancer diagnostics, for example. Now, these days, every MD wants to be an AI expert, so the world completely changed, of course. That's positive, but, of course, you have to differentiate the wheat from the chaff in the information space to do so, and that's how we help everyone.

Then, one way to actually also reduce the anxiety is, AI is not magic. Everyone internally, when we engage in these models, these fabulous large language models, for example, constantly fail in very specific use cases because there are, of course, models that statistically are very good in what they're trained of. As soon as they're out of their normal distribution, then they start to fail. In these areas, of course, it becomes very interesting to develop your own, to fine tune your own, or we also build foundation models from scratch. That also shows everybody within the company that their expertise in their business area or their therapeutic area is absolutely key to make that work. You cannot just throw these problems into LLM and expect an answer that solves your problem.

Lance Glinn:

One of the most, I think, anticipated frontiers is how AI might empower patients and providers directly through, things like predictive tools, personalized treatment, even digital therapeutics potentially. How is Lilly enhancing the patient journey and supporting physicians with better data informed decision making and AI?

Thomas Fuchs:

Yeah, I completely agree. That's a huge, huge area where Lilly was one of the first ones to push that forward with Lilly Direct. That means direct patient engagement with all kinds of means of communications and language models and chatbots are one of these. Then, of course, also telehealth, combining telehealth in that setting, and then hopefully also getting the medicines to the patients much more quickly and in a much easier way. It is, of course, a challenging task because it depends also in which country, in which jurisdiction you are. We do that differently, for example, in Brazil or in Australia, then we do it in the U.S. or in Europe. You have to be very specific. For a global company like Lilly, you have to consider all of that. It's education of patients and then really giving them the means to get to the medicines they need as quickly and as efficiently as possible.

Lance Glinn:

You mentioned a global company like Lilly. Unfortunately, there aren't a global set of rules around AI. Different countries have different sets of rules. How do you manage all that? How do you balance all that? When you're going to a country like Brazil, compared to a country like the United States, compared to England, compared to somewhere in Asia, you have to balance all these different rules, regulations. What's that process like?

Thomas Fuchs:

We do have to stay on top of all of this because again, the overarching goal is to produce not only medicines, but AI that's safe, effective, and equitable across the world. You have to stay on top of it. That's an area where you cannot cut back. The way we do that, first of all, internally, we have an AI registry which actually tracks all our AI projects, so the hundreds and hundreds and hundreds of projects within that for all these different areas. Then, depending on the risk, it also goes through our risk function evaluation, and that's compliance, legal, cyber security, et cetera. That helps us, first of all, to really have a great overview of what goes on and then to track everything. Then, depending on where you go, you, of course, have to go deep into the regulatory setting, engage our legal teams to do so, and then prove out, of course, first and foremost that these things do work. Digital or computational diagnostics, we know they already work. Digital therapies in some areas could be very interesting. That's certainly early days, but in some areas, that it's also a promising field.

Lance Glinn:

I want to quickly just touch on the biotech and big pharma relationship just for a second. We're seeing, I think, a wave of AI native biotech startups that approach drug discovery like a data science problem, using machine learning as really the core engine of innovation. Do you view startups more as partners, competitors, or something in between? How does this influx change the way pharmaceuticals like Eli Lilly operate?

Thomas Fuchs:

First of all, yes, it is a data science problem. At its core, it is data science, and the way we solve it is machine learning and AI. That ecosystem, these are our partners foremost. For us, of course, it's always a decision what you're going to buy and what you're going to build. I'm always arguing, we should buy what accelerates us and build what differentiates us. What differentiates us is sometimes very unique data. Lilly has data second to none decades of data from reaction data to toxicity data to molecules. All the molecules that didn't work are enormously valuable, of course.

Based off that, it makes often sense to build our own solutions for our own purposes. We were very successful already in doing so for some of the molecules, for example, and some of the targets. Often, of course, it does not make sense to reinvent the wheel. If you do something that can also be done outside, it's much, much better to partner. Lilly's building out also our own ecosystem of startups. We have the Gateway Labs as incubators, we have the Catalysts 360 program. We do invest in many, many startups and sometimes acquire them.

Lance Glinn:

I want to start wrapping up our conversation. If we look ahead five or 10 years, how close do you think we are from a world where AI can potentially generate drug candidates, validate them and then move them into development? Is that science fiction right now or are there signals that this could eventually become a reality?

Thomas Fuchs:

We are not there yet, and it will not be in the next two, three years. In 10 years, yes, maybe sooner, maybe later. Also, one thing to consider is actually why is it so difficult? That's also where people sometimes get thrown off. The reason we can capture language so nicely is that it's produced by our brain. It's not that old. It's a hundred thousand years old. It's structured, it's discreet. We have the whole internet to train on. Then, if you go to robotics and digit manipulation, that's already millions of years of evolution, much more difficult. Then, if you go down to a single cell, you have machinery. After billions of years of evolution that are so complex that our human language even fails to describe them. That's why you can't throw them in an LLM. They will never be able to solve that. That's why you need dedicated models. You can combine that with LLMs and agentic workflows to orchestrate everything, but you need these dedicated models.

By scaling up these models, we are much better in actually finding the novel molecules and make sure they bind properly. They're not toxic and so forth. In 10 years, yes, and if we envision that very positive world, then we might be in a world where we actually are very target-rich and there's an abundance of molecules in silico that theoretically would bind, and then, it's all development and manufacturing, and that's something Lilly is really, really good at. That's why we are pushing so hard to get to that point in AI at Lilly, because then, we can play all the manufacturing and development cards to really be the dominant driver in billing medicines in the future.

Lance Glinn:

Absolutely. Thomas, as we wrap up our conversation, speaking to that same five, 10 year period, what's your vision for how AI could fundamentally influence Eli Lilly? How are you hoping the technology helps transform the organization in the next decade?

Thomas Fuchs:

Within Lilly AI, we'll touch every job we have, all roles. That, of course, means efficiency in some areas and doing more, of course, with what we have now. That's table stakes. Every company is doing that. Every company should do that. What really, really drives me, what's much more exciting is how we can leapfrog the whole sector. How can we actually produce in discovery better molecules much, much faster? That means, of course, seamless lab in the loop to go from in silico to in vitro. We built that out at various places in the world to do very, very fast, very effectively, and then go into development to run trials much faster, much more effectively.

Then, of course, AI plays a huge role in all these other areas we discussed. In manufacturing, we see significant improvement in producing more drugs faster. Then, again, finance, communication, commercial marketing, and all the direct to consumer space. That's why it's important we have to push all these areas and not do one, but doing that well, Lilly will be the AI-driven medicine company of the future, and having the opportunity to be a little part of that is a huge privilege and very humbling.

Lance Glinn:

Well, Thomas, congratulations again on being named Lilly's first Chief AI Officer, and thank you so much for joining us Inside the ICE House.

Thomas Fuchs:

Thank you for your time.

Speaker 1:

That's our conversation for this week. Remember to rate, review, and subscribe wherever you listen and follow us on X at Icehouse Podcast. From the New York Stock Exchange, we'll talk to you again next week Inside the ICE House. Information contained in this podcast was obtained in part from publicly available sources and not independently verified. Neither ICE nor its affiliates make any representations or warranties, expressed or implied, as to the accuracy or completeness of the information, and do not sponsor, approve or endorse any of the content herein, all of which is presented solely for informational and educational purposes. Nothing herein constitutes an offer to sell, a solicitation of an offer to buy any security or a recommendation of any security or trading practice. Some portions of the preceding conversation may have been edited for the purpose of length or clarity.