Pharma Sessions
Pharma Sessions
Why AI Pilots Fail in Pharma, And Why Yours Won’t with Jeremy Zhang
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
Pharma Sessions is hosted by Jonathan Kaskey
Follow along on LinkedIn: https://www.linkedin.com/in/jonkaskey/
Or join the Pharma Sessions Substack:
www.pharmasessions.substack.com
I have rolled out AI tools that failed. The number one reason for it was that the people in the business just didn't align before rollout that the expected failures were acceptable. And then we did it anyway because there was conviction within the data science team. Like, these predictions are great. These models perform. Like, look at the accuracy we get. We know that this is great, and we'll just roll that out. And that to me is like the number one determinant. Obviously, like we could talk about the technology itself. That's super, super critical. But to me, like engaging with the business functions and having those champions with you is the number one determinant for failure or success.
SPEAKER_00Hello, hello, and welcome to Pharma Sessions, a place for pharmaceutical leaders to come and learn from each other. I'm your host, Jonathan Kaskey. Technology and market trends are bringing change at an ever-accelerating rate, and no person, team, or company can afford to be left behind. Here, we dive into the strategies and tactics that our guests use to tackle these challenges and create new opportunities and how you can do the same in your own organization. On today's episode of Pharma Sessions, I'm thrilled to welcome Jeremy Zhang, Senior Director and Head of Data Science Solutions at Atsuka. Jeremy leads enterprise-wide AI, advanced analytics initiatives across RD clinical development, corporate functions, really helping to translate complex data science into practical decision driving value. Previously held a similar position at Gilead, so just brings a tremendous amount of real world experience to the AI conversation, which is of course something that everybody's interested in. So with that, thanks for joining me, Jeremy. Hi, pleasure, Jonathan.
SPEAKER_01It's really great to be on her podcast and excited to discuss the topic. I think a lot of people are interested in AI. And specifically, does it actually bring value?
SPEAKER_00So I'd say that's a million-dollar question. I think it's like the uh hundreds of billion-dollar question, the trillion-dollar question, maybe. I don't know how open AI is doing on fundraising these days. But before we get into all of that, I always like to start out to try to get to know people a little bit better. Um, so we're recording this mid-January, I guess. Last year, did you listen to any music that stuck with you? Any anything resonate? Albums of the year, songs of the year. What's the Jeremy Zang Spotify playlist look like?
SPEAKER_01My Spotify playlist is no longer in my control. So I have an almost two-year-old son. My Spotify completely consists of like Bluey and like even Miss Rachel has like a thing on Spotify now. So if that's not on, there's a problem. Yeah, if I had my own control, like I generally listen to podcasts. So I listen to like the journal Today Explained. I love acquired. Um, so if you've listened to that podcast before, it's really long-form business discussions. It's fantastic. But lately, the only music I listen to is like albums of Bluey.
SPEAKER_00So it's sad when my kids got a hold of my phone. My Spotify rapped changed dramatically. And now they're older, they're 10 and 12, but I'll still get uh the last one I had is like you're in the top 1% of Rihanna listeners of the country. I'm like, I don't know that I am, but somebody in my house is. So anyway, and I'll forgive you for not including pharma sessions on your list of really good podcasts. But all right, let's jump in. So you worked across biotech, big pharma, big-ish pharma. I guess Suka is certainly a top 50 pharma, although I don't know if they're considered big, big pharma reading AI initiatives. I guess let's start at the beginning. Like, what drew you there? What in your career led you to this place where this is your focus?
SPEAKER_01That's a really great question. And if you look across pharmaceutical companies and even biotech companies, what you'll notice is that when you look at people that are like mid-career in data science or AI, when they were in college or when they or people like me were in university or in grad school, data science as a topic didn't exist. So then, like, how are all these mid-career data scientists, data scientists? So you really have like two types of uh backgrounds, computer science backgrounds and hard sciences backgrounds, statistical backgrounds. So for me, like I come from a biomedical engineering background, both bachelor's and PhD. And I fell in love with data science because that was what was required for the first job that I had in industry, which was at Abbott Labs. So we were building a molecular diagnostics platform, essentially a core lab instrument for a hospital to do like PCR testing for like infectious disease. And as part of that, like for systems testing as well as some of the core algorithms, a lot of machine learning was used. And it was more of like a learn on the job, but because most scientists had training in statistics, by then, like we had done a lot of R and SaaS programming. Like that I fell naturally into it, although I did have to learn a lot of Python. But I really grew to love like the data analytics as well as the modeling aspect of that work. And I knew from then that that's what I actually wanted to do in my career and not necessarily more of like a biologist or chemist or or even a biomedical engineer. That's how it really started for me.
SPEAKER_00That's really interesting. So you got into it very early in your career. I think one of the things that's interesting when people are designing these is so often it comes down to understanding the business requirements for how you're going to apply these. And I wasn't sure how much work you'd done where you could truly represent the voice of the customer, right? If you're considering your internal customers as you're working on these AI things versus just understanding the science and understanding the processes that need assistance with AI or with data science.
SPEAKER_01That's a very interesting question, right? Like one of the reasons why you see a lot of data scientists be very successful, despite not coming from a computer science background, is because when you're working on models or systems or AI systems in a pharma company or a life science company or biotech, the reality is CS folks will treat it more of like a computerized system where you have requirements and then you just work towards those requirements. Whereas scientists are more exploratory and you know that, especially with AI, because most of the outputs are probabilistic, you treat it more as a statistical product and you have to take into account like what business success looks like versus just fulfilling software requirements. So for me, I think the best approach is to treat AI systems not necessarily as, hey, let's get a bunch of requirements and just fulfill them. Rather, we have an ultimate business goal and we have to do exploration, experimentation, and then a lot of testing, validation, and statistical validation to make sure that the system functions within what the business needs. So it's very interesting to me because when I see, because we have like a lot of teams, especially vendors that we work with. Pharma is a very liberal outsourcer of uh computer system build. I'll see like teams that treat AI projects more as just gather requirements and fulfill them like a software project, they fail a lot more often than teams that are run by statisticians or data scientists.
SPEAKER_00So could you give just to kind of bring this to light, an example of what this from your career, right? Without disclosing anything, you should be like, what does this look like when it's going well? What does it look like if it's going poorly? Like, where did you kind of see things going off the road? If there's an example you can point to, I think that'll be quite helpful.
SPEAKER_01Without revealing too much, although I mean these days, every single company is doing this. I'll talk about content generation. So recently, you know, I've been involved in a number of content generation projects, and I'll be presenting on at least one of them at an upcoming uh biotech conference, Scope in Florida. So in content generation, you use the AI model, you essentially prompt it with context data from your internal systems. Like I have systems where I wear how some like my clinical protocol or artifacts for medical writers. And then you use that to essentially build new documents or new content. We working together with a team of software engineers, they essentially decided, okay, we just need to build the app. The first version of the prompt, oh, it looks good enough. Let's just run with it. Farther down the road, once you get into user testing with like your clinical teams or medical writers, what you realize is, oh no, that first version of the prompt, like it worked for the one document that the software engineers got their hands on, but it cannot generalize across like multiple different clinical trials or protocols. So that's where I now tend to think that people that rush into implementing AI systems without doing the experimentation to make sure will this system work across different therapeutic areas, different indications, different protocols, they tend to really struggle when they actually have to roll out the product. Whereas if you started in the beginning on doing more exploration and more actual testing of the AI systems and making sure that the actual AI functions correctly before building the app, that tends to go a lot better.
SPEAKER_00And I think that that's some of it comes back to what you're talking about before, where a lot of times pharmaceutical relying on external vendor partners for doing this. I have some experience in the space, literally around working with medical writers and protocols and you know, create uh informed consent form from a protocol and all that type of stuff. And it's much easier to train something on a single document and it you can provide a demo and it works in the demo, but there's a lot of variations out there, especially if it's a global company, right? Where there's all sorts of things that happen and a lot to work around AI is around the LLM space, and you're not always trying to come up with de novo content, right? Like you want approved content, it's a regulated space. So there's a lot that goes into these things that are really complex. And I was very surprised with, you know, we would try to be providing these demonstrations and good faith in all of that, but I think that a lot of people at Pharma just don't have the expertise to understand how you should be pressure testing these things with your vendors before you're moving forward. Jonathan, you're spot on.
SPEAKER_01I tend to see happen is because of the huge degree of outsourcing for a good reason. Um, although it's very interesting now that AI is getting a lot more advanced, like that is actually, especially consultants, isn't like one of the first dominoes to fall, but that's a separate discussion. What I see is teams that spend more time up front understanding what the limitations of AI are and how much they have to put into the system to guardrail it and make sure that it's generalizable. They tend to succeed. I've seen teams see the one demo from the vendor and be absolutely convinced that this is going to work for you brought up the ICF as an example. It will work for pregnancy ICF. It'll work for like different types of molecules with a similar phase, or it'll also work for different regional ICFs instead of a global one. It'll just lead to that assumption and the vendor will absolutely nod their head and go, oh yeah, yeah, that'll work. But then what you quickly realize is these vendors, their biggest interest is not to have too much differences between their clients, first of all. So they won't customize for you. But second, they really, really don't want to put too much variation into their systems because that just creates more overhead and more things for them to manage and it impacts their revenue and profits as well. So, not to say like I've worked with really stellar like software vendors, so it just totally depends on the team. And that's why I think it's really important for pharma companies to have people internally. You don't necessarily need them to like build everything, but they need to have the expertise to be able to sniff out like what's BS.
SPEAKER_00And I can share, because I've always been on the vendor percent, the vendor side, I think that that line of questioning is something that people rarely get into. A typical SaaS company, if you want customization or feature improvement, you basically make a request with the product manager. And then depending on how, exactly as you said, how uniform they think it is, how much money you're spending with them, whatever the case may be, you end up somewhere on the priority list. And I think that when you look at my personal belief from being around this for a long time, a lot of the times programmers might be hired from other industries where that's okay, right? Because it doesn't so much matter. But pharma is its own unique beast where it's like, even though there's one set of rules, each compliance department has its own interpretation of those rules. So, like, just because something works for Pfizer doesn't mean it will it might have no value for Amgen because they have a different legal perspective. And that, as much as companies rely on vendors, it's usually an after-the-fact conversation about like, hey, how does your product roadmap? That might be like a whole nother, I'm gonna have you back on for a conversation about you know vendor interviews, vendor how to make a successful text file. Exactly. But so with this, because then the other the other piece is defining requirements. So let's keep using the medical writing example. So let's say another thing that medical writers might do is medical information. And you might come up with a project where your goal is you're saying, you're like, hey, I looked at my stuff, 80% of my requests are copy-paste, like pretty boring, but like standardized responses that are needed. Trying to solve for that is really different than saying, hey, I want to get something that's gonna do 100%, give me a first draft of 100% of problems. I guess it's probably a judgment call, but like, where do you fit on that spectrum of trying to be super ambitious versus trying to be more workable? Like what's possible and what has a high likelihood of succeeding right now.
SPEAKER_01You are touching on something that I think is an art form, really, right now. Because on the one hand, you have executives um having been at two companies and seeing the perception of AI projects. I think in 2026, executives are definitely leaning in the direction of I want to see if this has value. You can no longer be successful just having interesting data science projects. So you're under that pressure to show value, but then you're also under the pressure of like, can I actually spend the time and energy to make sure that this product works for that team? So it's always to me about working with the functional area. And I work across like a broad range of different types of functions in pharma, like medical writers, clinicians, clinical operations, people in clinical data management, people in quality. So there's a variety of different people with different experiences. It's been in the industry shorter or longer, many companies, one company. But you got to set the right expectations with them and help them understand that AI is not perfect. I think that to me, like getting the point across that no matter what we do, no matter how much time we spend, we could spend 10,000 hours on this application or this model, it will not be perfect. You will always find some reason or some way that this thing is gonna fail. So then it's about setting the expectation. Okay, so risk adjusted for what you are doing, how much failure, hypothetically, can you stomach or can this process handle? And for my team right now, we human in the loop is a requirement. Like we're not ever, I can't say never, but currently we have no plans to deploy something prescriptively just because it's so regulated, you know? Like, so it's a hybrid, it's a balance. Like, okay, so you pull back on the performance of your AI, but then you have to include more human in the loop. Um, yet how much of it, how much time savings the team wants, is an art form to determine where that sweet spot may be for them. And then as the team rolling out or partnering to deliver that technology, then you actually have to deliver the goods. Like at the end of the day, you actually have to have the AI system that can perform to that level. So it's about like having that negotiation, getting them to understand. And once they have that understanding, isn't it?
SPEAKER_00I think that that almost goes back to kind of like management principles or something where at the end of the day, people want to accomplish their work, do a good job. And when you're rolling out something, there's this whole change management issue where if it fails in a way that they're not expecting, they might never want to go back to it again. Versus if something is identified ahead of time as a limitation, that's different. Then it's like, okay, well, this is actually working the way, the way that we're expecting, it still requires personal expertise, which is not a bad thing. So have you seen like on the change management, do you have any examples of rolling something out that worked or or rolling something out that that didn't work that you could point to?
SPEAKER_01Great question. So I've had the pleasure of rolling out content generation um specifically to you know medical writing team as well as clinical team that works. And the reason why I'll say like the top three reasons why those content generation projects have worked. So in one case, we're doing, you know, we're helping the clinical teams with database build, another case with like ICFs and plain language summaries and that kind of stuff. Top reasons for things working is engaging with a team early and having champions from those teams work together with you as you're developing. So the reason for that is they can exactly what you just mentioned, right? Failing in expected ways. So they have an understanding as they work with you, like how this system can fail. And they can become your champions during rollout. And we do just-in-time training. Like we never just roll something out, have like a luncheon learn and call it that. Like you have to, unfortunately, there has to be an air some hyper care for Gen AI because not everybody is educated on what Gen AI can or cannot do. And having those champions working with you as you're creating the tool or developing a tool or partnering to roll that tool out is critical. Like you have to have support from within the functional area. It's like mandatory. In instances where in the past, I'm not too arrogant to admit, just being humble, I have rolled out AI tools that failed. The number one reason for it was that the people in the business just didn't align before rollout that the expected failures were acceptable. And then we did it anyway because there was conviction within the data science team, like these predictions are great, like these models perform. Like, look at the accuracy we get. We know that this is great, and we'll just roll that out. And that to me is like the number one determinant. Obviously, like we could talk about the technology itself. That's super, super critical. But to me, like engaging with the business functions and having those champions with you is the number one determinant for failure or success.
SPEAKER_00If you've been enjoying the conversations here on Pharma Sessions, you should know they're made possible by the team at Xunt. XSunt helps life sciences companies turn complex data into clear, actionable insight. For years, Excent has made complicated data sets simple to help commercial, medical, and operations teams map what's happening, predict what's next, and make stronger decisions faster. And now there's an added AI layer that makes everything work so much better. I was actually pretty jaded about some of the AI approaches I'd seen, but when XSunt showed me theirs, I actually left my job to come work for them. It's really awesome. So if you want to understand your market, your customers, or your performance with more clarity than ever, check out excent.com. That's xs-unt.com. All right, let's jump back into the episode. You mentioned now where in 2026 people are really executives are saying, show me where this is delivering value. Whereas two years ago, it was, oh my God, I need an AI strategy. I need to be showing pilots, I need to be doing this. And it's like it was obviously off from the beginning, right? Where it's like, if you're in clinical, you don't need an AI strategy, you need to document whatever content creation strategy and AI might be a part of that, and something else might be part of it. It seems like what you're saying is you need to stay super focused on the business value and engage early with the business teams, your subject matter experts on the design, on testing, on communications, on training, on sort of all aspects of it.
SPEAKER_01I mean, I'll give you such a great example of this is for one of our projects where we are doing content generation in the clinical space, one of the people most engaged with us is literally the head of clinical standards at the company. When you have the support of the person who is developing the data standards for the clinical trials, like it's not necessarily a luxury. I would actually highly recommend other people in pharma companies who are working in content generation and clinical space. That person or persons, if they have multiple rules for it, is mandatory. Like you just cannot do content generation and clinical without the clinical standards person being part of the team. So I think it's there's a lot of business rules as well, in addition to like the technology roles that are going to be hypercritical for some of these AI initiatives that are very business focused.
SPEAKER_00It's funny, right? Because on the vendor side, like our my version of that is some of the people that I'm hiring over to the team, I'm hiring essentially some of our users from pharma because even as we're developing products, like if you have actually been in the field and been in analytics, right, you know better than somebody from the outside, no matter how much market research we do, what's needed, where are their gaps. And so you can't fake it on either side. You need that person, and even better if they're in charge of it, and it's like you sort of have both the carrot and the stick for coming on and participating. So that's really interesting. So I talked to some people, and this is people at Pharma, because sometimes they get excited and they're all bought in, and you hear some really big claims, right? The AI is gonna cut our development times by we're gonna be five times faster, right? We're gonna be do you hear some big numbers floating around? How does, and maybe this is situational, but like how does that land with the executives versus trying to take more of a stepwise approach? Are we looking for like major disruption at this point, or are we looking for continual improvements?
SPEAKER_01I think that's like such an interesting question, right? Because if you go to um, for example, conferences like JPM or conference conferences like technology conferences, there's a lot of excitement about like this will transform the industry. But the hard reality is that AI isn't new. When it comes to transformational technology, right? Like maybe a decade or two ago, there was a lot of bioinformatics stuff, and then there was a lot of like, oh, let's go digital, even before that. Like at the end of the day, most new technology introductions, and AI is just the latest one of them, they are doomed to failure without like really careful change management plans. Because just because it's AI doesn't mean how pharma conducts its business is going to change immediately. And the more transformational it is, the more different functions with different group or individual motivations for how they want to work or the vendors they engage with, the more of that you have to change. So even my company now, like there are aspirations. Certainly, you get you hear all the time, like, oh, this is going to change the way we work. And there are people that say, like, oh, let's completely change how we do things. This is going to accelerate like so much. I'm like, at the end of the day, you still need to get your incremental gains because that's where you're actually going to show the value now. Obviously, I'm going to disagree with some people who believe that this huge wholesale transformational change immediately is possible. I strongly believe that that's not possible. That the actual transformation is going to happen when you have done so many of those incremental things that one day you look back and go, wow, this is way different than we did things five years ago. But like to do it immediately, like the number of even people that you have to get into a room to all align, just thinking about the protocol. How many different like a safety team, clinicians, clinops?
SPEAKER_00Exactly. CMC, like everything is based off the protocol, right? It's all these dependencies and dependencies on the dependencies, right? So it's not as simple as, oh, we just changed this. It's like, okay, great, but how does that affect your manufacturing or whatever, right? There's always unforeseen issues.
SPEAKER_01It's going to be incremental. But you get those incremental, like 10% speed up here, sometimes 50% speed up over here, but maybe that in the grand scheme of things, that's only like three days on the critical path. But you get enough of those, and suddenly you look back one year, two years, three years down the road, and you see that we're doing things like 50% faster than we were before. That's how I think it's going to happen.
SPEAKER_00I can tell you, in the last week here at smaller tech startup, AI has probably saved me a couple days of work. Like I had to give my team a new commissions plan for the year. And I give the bare bones and it creates it. And so it's like, okay, well, that's a couple hours instead of a half a day, or responding to an RFP. I don't really feel like rewriting boilerplate about my company history and and whatnot. Like I still, I still heavily need to go through it because I think people are getting, hopefully, people are getting good. I feel like I'm recognizing more when people are sending me AI generated stuff. And I'm like, all right, well, I don't really want to read this if you couldn't be bothered to write it yourself. But some of the stuff is just really time consuming. And I'm sure every job has that, where there's things that take a lot of time that aren't super, super important versus those things that really do move the needle. You might need assistance in a different way or might require a higher level review if it is truly critical.
SPEAKER_01You're spot on. I mean, I think in general, there was a lot of trust harmed when Chat GPT first came out. I think there was like this, oh, this is amazing. And then the more people use it, the more people realized I can't really be sending that kind of stuff to my boss, or oh, if I sent that to the FDA, like this is gonna be a problem. So now we're like, I think we're in the area, and I detect it from the lay people, the people that are not like technologists, that trust has eroded. Like people have used either internal like big chatbots or even ChatGPT or even Gemini, frankly, enough, that they know that like, you know, this is not the level of quality of work that can be submitted immediately. So that's why I think the next era of this whole LLM space is really gonna be like how do you build in the safeguards, attain the AI, to then have it be trustworthy enough to actually have people, you know, use it enough that you get those transformations. I think it's also gonna be individually too. Like every single person themselves is gonna have to grapple with that.
SPEAKER_00In the meantime, for our team, I still do everything you said is 100% accurate, but we still do use it for some of our own back and forth and messaging. And what we just do is say, hey, comes from Chat GBT, which is almost shorthand for like take it with a grain of salt, looks okay, but it's almost like citing your sources because I think that where it starts to feel very uncomfortable is when you're positioning it as your own work. And then it ultimately gets back to this idea. This is another separate conversation, but who's responsible, right? If if you're making a recommendation, you are responsible whether or not you put the words on paper or you copied it over from a browser window. And I think that's what makes me nervous is like, oh, I want to be able to stand behind everything I say, but I also want to trust things that are coming into my inbox from people in my company, people on my team, whatever, whatever. But everybody's busy, right? And sometimes work's happen. That's I think exactly what you're talking about, right? This trust gap. So I don't know, I don't know when that gets filled. Again, this is completely off topic.
SPEAKER_01These are just some of the things that you know I thought about too. I completely agree with there's a camp of people in AI space who believe that LLMs are a dead end. And they believe that it's a dead end for a very specific reason, and recency and provenance is that reason, right? Like how recent is the information and can you identify the provenance of the information it provides? And in both cases, unless you build systems around it, the answer is no, it's not recent enough. So there's a camp of people who believe LLM are a dead end. So for me, like I tend to think of again the AI as just yet the next new technology that can do more than what was in the past. But the same things that in pharma industry we have to think about, which is like safety and safeguards and quality of the work and also making sure that everything is in line, like that still has to happen. But that even that activity can be accelerated by AI. So it just depends like how you use it to meet in 2026, it's gonna be more about how people implement AI, unless and less I think about unless Google makes some another amazing breakthrough. I think that there's going to be some natural cliff beyond which LLMs are not going to continue to improve without some measurable improvement in the data.
SPEAKER_00So this sits on something, maybe this will just be a couple closing questions, but sounds like a lot of what you're saying is don't get distracted by the fanciness or the whatever the modality is that you're using. Use it, rely on big data, rely on AI for what it's good, but stay grounded in the goals, right? And the the business need for whatever it is you're trying to implement. Would you say that's accurate?
SPEAKER_01Absolutely. You can never implement AI for the sake of AI. Like just because Chatbot can produce something doesn't mean it's meaningful. Um, you know, again, I'll use just using an example because I think these are the best ways to do it. Like so, we're rolling out a tool soon for do generative insights, like conversational analytics for clinical trials. And one of the things that somebody asked me is how come you have like barely single-digit, double-digit skills identified? Because those are the ones that we put these tremendous amount of safeguards against that we know that the information produces are accurate. And that's like the kind of thing that we have to do in order to regain trust, which is to continuously show that we can safeguard it and that people can actually trust the information coming out of it. Um, and that requires a tremendous focus on what actually matters to the business.
SPEAKER_00That's quite interesting. And I think that again, you're talking about incremental change, you're talking about very clear boundaries around what you're trying to achieve. It's funny, the meeting I was in right before this was essentially a roadmap, a roadmap meeting where there's all this push, not push, there's interest from some of our clients on like having more agenc stuff in our software. And we're trying to define, okay, where do we go? What's the first one? And the it's literally the same conversation. It's like what can be bounded, what is achievable, what delivers the most value with the least risk, essentially, and risk as defined by risk of burning dev cycles on something that isn't gonna the clients aren't gonna be interested, but also like either can't be any risk on the client side that you're delivering bad information or wrong information, because again, it gets back to that change management. Like the first time something fails, it's really hard to recover that goodwill. So that's really good to hear. That's yours respectful. All right, let me ask you one final question, Jeremy. Right? There's probably a lot of people out there that are interested generally in this concept in this space. If you give one piece of advice to someone earlier in their farmer career about working at this, you know, this intersection of data, influence, AI, trust, what would that be?
SPEAKER_01I think the number one advice I would have for anyone early in your farmer career, maybe you're in a business function, interested in getting into AI, becoming more technical, or you're like a very technical person seeking to understand more about business. The number one advice I would have is to not stop growing. In January 2026, just look back at January 2025. It's very, very different. And do not tie yourself down to one way of doing things. Like you may go to a conference in January, February, and you have conviction this is how AI is going to be done. I guarantee you by September things may have changed. So just be curious, seek out knowledge, don't stop learning, but also get your hands dirty. I think the number one determinant of success I've seen people getting into AI and pharma are people who have gone through the pain of implementation. If somebody has never gone through the actual pain of trying to implement something and generate value, and then they try to tell you like what works or doesn't in AI, like you shouldn't listen to that person. Like the only people that truly understand are the ones who have the battle scars to prove that they know all the things that can go wrong. I have many of those. Like, so for somebody earlier in the career, I'm like, don't just don't be afraid to fail. You gotta take those licks, earn those scars, and then that will be what informs you like the next time how to succeed better. But you have to go through that cycle and you have to never stop learning and never stop at like understanding more, right? Wonderful. Thank you, Jeremy. I really appreciate you coming on. Yeah, thank you so much, Jonathan. It was a great conversation.
SPEAKER_00And that's a wrap on today's episode of Pharma Sessions with me, Jonathan Kaske. If you enjoyed today's conversation, don't forget to hit follow or subscribe and share it with someone else in the pharma world who might need to hear it. For more on pharma trends, career growth, and business strategies, connect with me, Jonathan Kaske, on LinkedIn. Until next time, thanks for listening.