While the application of AI for drug discovery has been well documented and a source of considerable investment for the pharmaceutical sector, we are only just starting to see its impact for therapy developers. A 2024 report from McKinsey Global Institute estimates that generative AI could generate up to $100 billion in economic value for pharma and medical product companies, and therapy developers are eager to see these benefits sooner rather than later, no more so than in the complex and challenging field of cell and gene therapies.
With the help of two expert guests, Stuart will uncover how AI is being used in the industry today, and where it could go in the future.
Guests:
- Ken Harris, Chief Strategy Officer and Head of AI at OmniaBio
- Sean Bedingfield, PhD, Senior Advisor at Eli Lilly Genetic Medicine
.avif)
Setting the scene: AI in pharma and cell therapy
[00:00:04] Stuart Lowe Over the past two years generative AI and AI more broadly, have taken center stage in popular culture. But the pharmaceutical industry has already spent billions on acquisitions and AI enabled drug discovery. And in a 2024 report, McKinsey Global Institute has estimated that generative AI could generate between 60 and 100 billion dollars in economic value for pharma and medical product companies. So understandably, therapy developers are keen to see these benefits sooner rather than later. And I want to find out what this might mean for the cell and gene therapy industry. In this episode, I'm joined by two experts to help decode the role of AI in selling gene therapy. Who's taking the lead in implementing it? What operational impact could it have? And which areas of innovation might it help unlock? First, I'll speak with Ken Harris, recently appointed Chief Strategy Officer and head of AI at OmniaBio. Ken joins OmniaBio from Amazon, where he led strategic initiatives centered around artificial intelligence, big data analytics and multimodal computation for precision medicine. Then I'll be joined by Sean Bedingfield, senior Advisor at Eli Lilly Genetic Medicine, who's seen first hand the real world applications of AI and gene therapy. Well, welcome. It's nice to see you. Would you mind introducing yourself and telling us a bit more about what you do at OmniaBio?
Introducing Ken Harris
[00:01:37] Ken Harris Yes. Thanks, Stuart. So I'm Ken Harris and I'm the Chief Strategy and AI Officer for OmniaBio. Probably as a little bit of a mouthful, but the reason why they're combined at OmniaBio is that we're really driving the cellular therapy market towards intelligence and intelligence meaning using technology both for discovery but also for production. And so the combination of AI and strategy makes sense to put those together at the moment.
[00:02:09] Stuart Lowe I suppose you're saying that a lot of your strategy is around AI and digitization then?
AI and strategy at OmniaBio
[00:02:15] Ken Harris Correct. We do have our target markets and that's been a part of the the evolution of strategy because as you think about using technology, it's a broad basket and you need to be very use case focused for success. So in order to be use case focus, we also have to focus them in specific target markets. And so they do come together.
[00:02:40] Stuart Lowe Thinking about AI and maybe just setting the scene a little bit. Why do you think we are on the continuum of AI being implemented for sale and gene therapy? Are we at the beginning? Is it nearly complete? Why would you say we are?
The early days of AI in cell and gene therapy
[00:02:56] Ken Harris I would say we are at the very, very, very beginning. Artificial intelligence has been around a really long time. The evolution that happened with generative AI and foundation based models, let's say two years ago, really with the launch of chatGPT, which has only been a couple of years. Not only revolutionized the state of artificial intelligence, but I think it was a paradigm shift in people's thinking, and it really brought everybody into the mix of understanding how AI can help them and potentially some of the risks around AI and how society looks at AI and how industry looks at AI and how different types of regulators are going to look at AI. So I would say we really, in the last 24 months, started that maturation process that was not underway before then.
[00:03:53] Stuart Lowe But there's been some deployment of artificial intelligence, probably more on the machine learning side for health care before before this point. Right?
[00:04:02] Ken Harris So I think it depends upon how you define it. So if you want to go to the level of of algorithms and binary processing of algorithms and statistics, of course, that's been around a pretty long time. As you start to think about actually allowing in silico thinking, so to speak, that is pretty new. The tools were was not available for organizations to spin up affordably, to have access and also to be able to have the transparency to them that really we need and especially in the regulated space until recently.
[00:04:39] Stuart Lowe And do you think part of the power is the democratization of access to AI tools then?
Democratisation of access to AI tools
[00:04:45] Ken Harris I think it's absolutely essential because the way it can impact the economy and society as a whole, it has to be democratized. It has to be a part of our daily life. And we really have to reskill the workforce and society as a whole and in social media actually plays a significant role in that upskilling because most people are involved in social media. Even my 90 year old mother is involved in social media and they may not understand the level of AI that's happening in the background, but they know when they log in to certain online commercial organizations and something pops up to for them to buy that they were thinking about it and like, how did they know that? How do they know?
[00:05:31] Stuart Lowe Yeah, yeah.
[00:05:33] Ken Harris That's AI. operating in the background.
[00:05:37] Stuart Lowe It's clear that part of the optimism that Ken's feels about AI is that organizations can now access it much more affordably than before. And I wanted to know what sort of an impact that could have on a field like cell and gene. And where he expected to see early adoption. We were just starting to talk a little bit about how AI is is making an impact now and what types of cell therapies do you think this strategy of cost reduction and robotics ation would have a greatest impact?
Early adoption in CAR-T and autologous therapies
[00:06:12] Ken Harris Our opinion and our focus is that the cell therapies that are working today and proven and the market is adopting first and foremost are around immunotherapy and CAR T. It's also more than 50 percent of the total number of projects in the cell and gene therapy space. So it is where the market is. So we are highly focused there and in particular in the autologous space, which are the products that are are commercially available and the ones that are more advanced in that in the clinical pipelines as well. And they're very expensive to produce. So we are looking at AI there kind of, as I was mentioning before, both on incoming donor material. So in the verification that the patient's donor cells are really quality based and we don't want put them into production if they're not going to produce a good therapeutic product.
[00:07:19] Stuart Lowe That's something which are quite laborious to to achieve right now.
[00:07:23] Ken Harris Or actually not even being pursued. The current industry standard is to immunophenotype, the patient's T cells before putting them into production. And my team and there was a nature publication on this last year as very focused on transitioning the market to genotyping and transcriptome analysis of the donor cells. And preliminarily we can show much better correlation in genotyping of the donor cell to clinical response than immuno phenotyping does.
[00:08:00] Stuart Lowe Really? Okay.
Stratifying patients with AI
[00:08:01] Ken Harris Yes. So we think that's directionally quite interesting. And part of my team is going to be further advancing the machine, learning around that and hopefully getting the market to adopt that approach. Even thinking about it for clinical trial inclusion and exclusion criteria.
[00:08:21] Stuart Lowe I see. Yes. Yeah. So you've got this patient stratification angle in other parts of the former industry.
[00:08:27] Ken Harris Correct. That's right. So this is first and foremost for us and a high priority AI project and very focused. The second area is, as you think about coming into production and where costs are. The second area is really going to be post-production in automating quality control. And our goal there is to be able to basically get a 4 to 5 x output in the same infrastructure and therefore get the cost down and actually take humans out, which should improve even the quality of the analysis.
[00:09:04] Stuart Lowe Yeah. Yeah. So you know, one person a having results on on day one and possibly having a different result if they don't want a different day, correct?
Automation and scaling in manufacturing
[00:09:14] Ken Harris That's right. And actually being able to going back to our data discussion, harness much more data that can be useful even for development for organizations. Our pharma clients may want that data. It may be very rich for them just from a process development or research standpoint. So that becomes a value add. And then the third step is really around automating the Bioprocessing units. Because if you can imagine the size of the GMP suites that are needed as you grow in dose output, it's it's not scalable. And the only way we can make it scalable is to put go both vertically and horizontally and scale and service those units with robots.
[00:10:01] Stuart Lowe That's what a few companies have latched on to as well. There's numerous spin outs and TTPs associated with with one of them.
Logistics and supply chain optimisation
[00:10:09] Ken Harris And then the fourth theory for us is really around being [00:10:12]the vein [0.0s] logistics, which is absolutely critical. And this goes back to a little bit of of my Amazon training. If you can imagine how Amazon gets your package to your door, rain or shine next day. That's all AI driven. And that's that is the same for us in getting your patients donor cells in producing them and then getting them back out in a short period of time for for infusion. And we use AI to do that.
[00:10:44] Stuart Lowe And I suppose all of these things need to work together, right? So you've named four different areas, but actually there's not much point in in working just on the logistics if you haven't got the automation to handle. The scheduling is is much more powerful if you're if you're working on all of them together.
Integration of AI across the value chain
[00:11:02] Ken Harris They go hand in hand. If we take our capacity from 1500 doses a year in CAR T to 5000 doses in a GMP suite, it just cascades in scalability across all of it. So, yes, they do have to go hand in hand.
[00:11:18] Stuart Lowe That's really exciting. What are your first readouts that you that you are excited about that you think? Yes, this is working really well.
Challenges in bioprocessing automation
[00:11:27] Ken Harris Yeah, I think for us it's probably most of these are pretty straightforward AI implementations. Where I think we have the biggest challenge is actually in the processing piece, which AI is going to play the smallest part at the moment, just because their Bioprocessing units don't have great data collection or sensors yet, and that's an industry blocker. But from our side, how we see the blockers there is I actually have to have better robots that have better dexterity. And if you could think about elbow and risk joints and so forth, move around and do the stuff that needs to evolve before better Bioprocessing automation is really within reach and that's part of our other area of development that we will focus on. So right now we're doing traditional based robotics and I do think that will limit us of where we ultimately need to get to and throughput. I don't think we can achieve that until the robots mature.
[00:12:31] Stuart Lowe Okay. Well, you're partnering probably with with the robotics providers who are working on that side of that side of things.
[00:12:39] Ken Harris That's correct. That's right.
[00:12:42] Stuart Lowe As Ken explains, there are several areas that right now would benefit from implementation of AI to provide operational efficiencies. I was also interested to hear from Ken how he thought AI could promote more innovation in the cell and gene therapy industry. You mentioned about the Big Pharma because they can benefit in quite, quite a few ways from the from the automation, from the from the data side. But also our other guest on this podcast from Eli Lilly has talked about how it's helped him to navigate large organizations as well using LLMs.
Federated data and secure collaboration
[00:13:15] Ken Harris And when you think about the collaboration capability that you can do securely, one of the things that we were working on and Amazon and I just this is just going to change the whole state of health care data and collaboration is the fact that technology is now evolved to the point you no longer need to send your data anywhere or give anyone visibility to it. You can put it in what's called your own clean room and you can federate data clean room. So I could have one, my competitor could have one, and we can actually compute against all of the data without anyone knowing or seeing what my data is or looks like. And that's just revolutionary.
[00:14:02] Stuart Lowe What do you think that will enable then?
Patient benefits of data sharing
[00:14:04] Ken Harris Let's actually start at the patient benefit of that. The fact now, because health care systems were really resistant to in the in the US, the regulations forbid it. Right. Actually providing medical record data. Now that can be done where especially in rare diseases where you need collaboration, you can drop it in and it never leaves your system. It's de-identified. And now 30 academic medical centers can quickly get to a statistically significant dataset and move on and not violate or take any risks legally. And that's just so powerful. And I think that's been and is being broadly embraced now. Certain organizations are trying to corner that market with data, which I think the regulators are going to have to step step in and say no. But technology probably helps counter as well.
[00:15:07] Stuart Lowe That's really exciting because it feels like everybody's incentives are aligned there. It's good for the patient, it's good for the researchers. Maybe you can monetize some of the data from let's say it's generated your manufacturing runs or generate too doing during negative run, during development or something like that. But if it feels like there could be some nice use cases in process and process development as well.
[00:15:34] Ken Harris I think there could be and I think you could even take it to the point of a pharmaceutical company could look at their patient post-treatment, which they have no visibility to now. They're yeah, they're blind. So basically they could pay the health care systems to federate the data of their drug post-market approval.
[00:15:55] Stuart Lowe Yes. Okay.
[00:15:55] Ken Harris And say, hey, can I now start to look at what's happening with this group and further segment patients, but also drive it back to my data and correlate that maybe to oxygen absorption levels in production to clinical outcomes. That's so powerful.
[00:16:17] Stuart Lowe I find it really motivating that people like Ken are returning to cell and gene therapy and looking to deploy AI to benefit patients. The decisions that he and others make could end up having enormous benefits for the way we develop and manufacture these therapies. And here are my three key takeaways from our conversation. Number one, what's really exciting about the emergence of generative AI is the ability for almost anyone to access it and to explore its benefits in their own part of the industry. Number two, the complexity and cost of cell and gene therapy, manufacture and distribution is a prime target for operational efficiencies by deployment of AI. Number three, AI could be used to break down barriers between organizations and allow collaborators to further optimize process development. Thinking about some of those organizations, I was intrigued to see how AI could be leveraged to drive efficiencies across big pharma companies. So next, I spoke to Sean Bedingfield, who works in Eli Lilly's Genetic Medicine group, to find out more about how he and his team have been using AI. Thanks for joining us on the Life Sciences Podcast. Could you start by introducing yourself and telling us a bit more about what you do with Eli Lilly?
Introducing Sean Bedingfield
[00:17:37] Sean Bedingfield Thank you. Yes, I'm Sean Bedingfield. I work at Eli Lilly. I'm trained as a drug delivery scientist and I work in Lilly's Genetic Medicine Group. So we match with experts in therapeutic areas like neuroscience, oncology, immunology, metabolic disease to deliver different therapies to different cell types. So I have a small team. I've been at Lilly for four years.
[00:18:00] Stuart Lowe What aspects of genetic medicine are you studying in your team?
Applications of AI in genetic medicine
[00:18:03] Sean Bedingfield Our team ends up getting pulled into a lot of different genetic medicine modalities. So we have the small molecule style modalities like any sense, like a nucleotides, small interfering RNAs for applications in RNA interference, but also things with our public deal with with proQR, where we're working on things like RNA editing. We have the larger cargo applications like gene editing, gene therapy. Those those play a big part in what we're doing. And then there's there's the stuff in between where you're using mRNA to translate specific proteins to address certain diseases.
[00:18:41] Stuart Lowe That sounds really, really interesting to be able to cover all that broad spectrum.
[00:18:46] Sean Bedingfield Yeah, I know. It's, it's exciting. It's a lot of variety, which is, which is fun.
[00:18:50] Stuart Lowe And the topic we're going to cover today is about artificial intelligence. Are there any kind of particular areas where AI has been useful for you?
[00:19:00] Sean Bedingfield Yeah, AI is popping up everywhere. And the area where we're getting the most traction is where data are naturally more ordered and more complicated. That can be areas as simple as flow cytometry where you have tens of thousands of cells. You have cell specific, often ever more high resolution data and metadata. And it's hard to keep not just the cool insights that are, you know, and the trends that you might miss, but also the simple quality control and trends where you might suddenly see a drop off in a certain cell population that you're expecting. And when a human being is preparing the next sample for the next experiment, already it's easy to breeze over those things if everything looks generally okay. Another area where AI has really been connecting the dots for me personally is in stringing together disparate experiments that have happened over time and that some of that natural data structuring happens when a when an experiment gets completed and summarized. And AI is surprisingly good at finding other times when other separate teams in our massive company have worked on something related. And so we get to capture some of the value back from that kind of lost to time experimental insight, which is useful.
[00:20:25] Stuart Lowe That's interest as useful sort of operationally as it is inspirationally.
[00:20:32] Sean Bedingfield That's right. That's right. I think it's easy for me to sit down and spend time trying to pull out the magical cool trend that gives you the new angle for a better therapy. I have a hard time sitting down for the same amount of time to just make sure that everything's in order and that all the all the right boxes are checked.
[00:20:53] Stuart Lowe And what type of AI is it that you use most of all? Do you have a working definition of your AI? Is it more chatGPT side or is it more machine learning? Have you found the best use case?
[00:21:05] Sean Bedingfield The best use case is the language learning models are great, right? And they're great at sifting through PowerPoint slides and [00:21:14]the and [0.7s] that's great. The area where I think machine learning takes more of a front seat is in prospective molecule generation. So whether that's a RNA sequence, a ionized lipid for a little bit nanoparticle, there's there's a lot of options where you start getting into chemo informatics or sequence optimization. Those are areas where that self training dataset gets more valuable. I don't think either has been outmoded yet. You know, we're still I can't think of any company that has so much data that they don't have a need for rational design, right? I think we're all starving for data in the cell and gene therapy space. I think the next 15 years will still have plenty of people in Python trying to pull out chemoinformatic trends on top of machine learning.
[00:22:07] Stuart Lowe It's obvious that Sean has seen the benefits of AI across many aspects of his work. And one of the applications we discussed earlier with Ken was the ability to foster collaboration. So I asked Sean about how AI was breaking down barriers at Eli Lilly. Is there anything that's particularly surprised you with the advent of LLMs?
AI in operational efficiency and collaboration
[00:22:29] Sean Bedingfield Yeah. Yeah. So this is fairly specific to my situation, but we work in this genetic medicine team that's not espoused to a specific therapeutic area. And so we're constantly faced with a brand new set of disease state conditions, considerations that we've never had the time to appreciate the nuance of and and we do rely heavily on our therapeutic area experts, but it is really helpful just to peel off the first layer of what experiments have been run within that therapeutic area in the last six months to help inform what we're going to do next. What are the questions that I've already been answered by other experiments I haven't considered? What are the best practices within the disease state? And of course you set to check your sources and whatnot. But it has been an easy way to navigate a brand new white field of unexplored things that you need to suddenly become expert enough in.
[00:23:26] Stuart Lowe Yeah, So. So that instant expert. Yeah. And must have shaved months off the cycle time and getting up to speed on a new therapeutic area.
[00:23:37] Sean Bedingfield Well certainly. And often when genetic medicine, your work is quite exploratory. Right? I sit in the early discovery phase. Everything we do is starting on a on a paper napkin and then working its way towards IND filing. And and it's a funnel where you're starting with lots of ideas and then you work down to fewer ideas that really have have legs. And when you have more things at the top of that funnel, naturally it can be hard to get some of the attention that the projects need to get off the ground. And so these therapeutic area experts, they're stretched across portfolio projects, they're stretched across late stage projects. And so you can't just sit down and have them hold your hand the whole time. And so just in terms of access and man hours, it's shaved off time and in advancement. It's established a faster shorthand with therapeutic areas.
[00:24:25] Stuart Lowe Yeah. Okay. Okay. Did you see that happening? Do you think that was going to be a likely outcome?
[00:24:31] Sean Bedingfield Of artificial intelligence getting incorporated into work? Definitely not. When I imagined artificial intelligence and what that would look like in my field, I thought it would be mostly on the front end in the theoretical world. Then what we're really finding is that the endpoints and the results are getting a lot more connected and they're getting a lot easier to review and analyze. And I think that's a step away from seeing nuanced trends. But it's a time to find trends in in the archeology of the data that already exists.
[00:25:07] Stuart Lowe Interesting. Yeah. So has got a better institutional memory than probably the people who work in institution.
[00:25:13] Sean Bedingfield We are we have tens of thousands of employees. I couldn't even tell you all the coworkers names, much less what they're working on. And often that's things that I need to know about to to make a project go forward.
[00:25:27] Stuart Lowe The operational efficiencies Sean describes are impressive. Imagine being an instant expert on thousands of topics just at the click of a button. But for those who are already experts in their own field, how does AI help them come up with the next innovation? I asked Sean for his thoughts. In the drug development process, which aspects have traditionally been the most difficult or most inefficient to to execute?
Overcoming inefficiencies in discovery
[00:25:54] Sean Bedingfield We have plenty of inefficiencies, especially in early discovery. Traditionally, the major challenge in genetic medicine delivery science is that you can't predict effective delivery in a flask. Let's take the random example of a CRISPR therapy that is administered intravenously to engage stem cells to correct a mutation. Sure, the particles could see activity in stem cells in culture, but that's really the bottom rung requirements. They need to circulate long enough in the right places to have favorable uptake into those stem cells and preferably avoid cells That might be problematic if they receive the same cargo. The the typical way to design for that is to run rodent studies. And lots of them do take careful, careful measurements of bio distribution. You run extensive flow cytometry and eventually also a lot of sequencing after cell sorting to confirm the extent and quality of the correction. So all of that is to say that you are throwing real resources and time and manpower at risk into every step as you advance your design process. Like in much of medicinal science, poor predictors of in vivo behavior are a primary driver of inefficiency in the research process.
[00:27:09] Stuart Lowe This is something time and again you got good data in rodents and then the first time you try it in humans, there's some reaction or some side effect that you weren't expecting and you hadn't had any final?
[00:27:21] Sean Bedingfield Just a rodent to higher model translation problem. But there's also there's a lot of small companies out there right now that have very limited in-vivo data. The genetic medicine is a hot area where if you have if you have a new mechanism that's been, you know, elucidated from your bio prospecting of cephalopods in the Gulf of Mexico, that all of a sudden people get really excited and you may have something that's only had some activity in cells a handful of times from there to just working in any sort of organism can be quite a jump, too.
[00:27:54] Stuart Lowe Definitely. It's an area ripe for innovation, a ripe for improvement.
[00:27:58] Sean Bedingfield That's right. And even let's say that it doesn't work well in higher models. And you go back to the drawing board, there's 20 branches, if not more, that you could take, and then you have to decide which ones you're going to synthesize, which ones are going to, you know, put in culture. And then of those that how do how they look and culture, well, a lot of them are going to look pretty great. You still have to limit what gets what gets in the rodents in the first place. So even as you even if something is looking somewhat promising and you're just iterating on a promising idea that iteration is is a heavy lift.
[00:28:29] Stuart Lowe This is something that you have to deal with on a on a pretty regular basis. Right? You're making those sorts of decisions for Lilly, right?
[00:28:37] Sean Bedingfield That's right. Yeah, we're right. We're right there. We are really trying to work on an improvement over what exists or what's in clinical trials right now. And that improvement requires really specific, really small changes. And we've all seen very subtle changes to genetic medicine profiles, whether that's a tweak on a capsid peptide sequence or one carbon on an invisible lipid has made a huge difference in and in how bio distribution plays out in an animal and what kind of potency and tolerability. Heck, as we go through that, there's a lot of resources burned. And I think the real the real important thing there is that you pick the right questions to ask and the right things to design for, to keep things in a space where you can have a high probability of technical success.
[00:29:23] Stuart Lowe And it's an area where you've seen AI then then start to move the needle for your outcomes.
How AI supports design and endpoint analysis
[00:29:28] Sean Bedingfield I would say that we've seen most of the benefit from AI in endpoint analysis and in aiding in cargo design and chemo informatics. Those are the areas where I think you can actually see the AI influence as a as a potential benefit.
[00:29:47] Stuart Lowe Thinking about the resources Sean has at his disposal, it's clear that AI is able to augment his ability to pursue groundbreaking science. Given his experience, I wondered what advice he had for companies just starting out and how they should use AI most effectively. If somebody approached you who is trying to deploy AI in discovery, if they haven't already tried to, what what advice would you give them to get the most out of it?
Advice for small biotechs adopting AI
[00:30:15] Sean Bedingfield Let's just keep it to genetic medicine, right? Yeah, That's an area that I know. And in genetic medicine the drug product is so complicated. And so if you really want to build around AI, you're going to need to do more data structuring and more characterization upfront than you would do without it. And so that means that your lab notebooks not going to be as fun to fill out. It means that you're going to have to spend more time on more pieces of characterization. Every experiment will have more QC controls, because if you are training on bad data, that's a problem. You're going to have to invest more in QC, more in characterization, more in structuring your data. And at the end of the day, I think that will focus you on doing more of the right experiments and will push you towards getting the right thing.
[00:31:02] Stuart Lowe Lilly has just appointed AI officer. Should more pharma companies be looking to develop a strong focus in AI?
[00:31:10] Sean Bedingfield I think that whether that's outside or inside their company, if they're working on complicated drug products like those found in cell and gene therapies, they should probably be heavily considering how their competition is using artificial intelligence and how they can use artificial intelligence primarily as it's nothing else, a cost saving tool. The unromantic side of this is that everywhere that people are applying this, the main value proposition is that it's saving you on labor. And if you're a very small biotech with limited runway, saving on labor is great. I think the challenge is there's a lot of things to get easily excited about with artificial intelligence and really finding how is this actually going to save me experimental time? How is this going to save me? Man hours will probably make that a more useful conversation than if you did attracted to something shiny where you have to build a whole new workflow to do it. It would be exciting to jump to what things will look like in 20 years. Right? Where?
[00:32:09] Stuart Lowe What do you what do you what do you think? What do you think things will look like in 20 years?
Looking ahead: AI and the future of innovation
[00:32:12] Sean Bedingfield I think that more of what is a company now will be a free tool in Python in 20 years. And so as more things become a free tool in Python or other places or whatever, the next AI enhanced pseudocode version of Python is, that's going to save on on labor, which means that more people can try more crazy things. And that activation energy required to try something crazy will be lower. I think that's going to be part of the future too, is easier to integrate resources from central data generators and more tools to have labor elimination and lower energy activation for cool ideas to come forward.
[00:32:57] Stuart Lowe Sort of smoothes out the risk profile a little bit.
[00:33:01] Sean Bedingfield For everybody. Yeah, for everybody.
[00:33:03] Stuart Lowe Yeah. That can only that can only be a good thing. Who knows where the next big idea is going to come from? They'd be ashamed to to to limit that because of cost. I really enjoyed my conversation with Sean and finding out about the many and varied use cases for AI across Eli Lilly and in genetic medicine in particular. My key takeaways from our conversation are number one, AI is really good at sifting through datasets that might already exist within an organization to draw out new trends. Number two, you can use LLMs to shortcut getting up to speed on new concepts, which makes collaboration across departments easier. And it shows researchers and drug developers are all speaking the same language. Number three, we might see the democratization of AI as an opportunity to unlock innovation across companies that don't necessarily have the resources of Big Pharma. In the cell and gene therapy industry, the complexity of development to manufacture means that there's a good chance that we'll see tools like AI driving operational efficiencies. In coming years, we might also see innovation in the way that people work together, both within and between organizations. And if that leads to better outcomes and lower costs for patients, that can only be a good thing. Thank you to Ken and Sean for sharing their knowledge. I hope you found these conversations as insightful as I did.
