Hey, guys. Welcome to the podcast. My guest today is Luca Dezzani. He is VP US Medical Affairs Oncology for Johnson and Johnson, and we talk about generative AI, artificial intelligence today in medical affairs. Awesome conversation. I learned a lot. Think you guys are gonna love this. Don't forget to check us out on MSL Talk Live, which is the 1st Tuesday typically of every month, and that's on LinkedIn live. And connect with me on LinkedIn if you haven't already. Thanks for your support.
Welcome to MSL talk with Tom Caravella, a podcast specifically designed for MSLs and all things field medical. Hey, Luca. Welcome to the podcast. How are you doing today, my friend? Hey, Tom. Thank you for having me. Very, very good. I am so excited. We've guys we've been talking about this for a long time. I know Luca. I see him all the time. You know, it just Luca is always at the big conferences and and, you know, we've certainly become friendly.
And the inspiration for this specific episode came when I saw Luca do a talk at Mass East earlier this year on the same, if not similar topic, and it was amazing. I learned so much. And I was like, you know, I've been I've been hounding him ever since. So before we get into that, Luca, why don't you do a quick introduction?
Yeah. Yeah. Yeah. Yeah. And actually, let me just tease you a little bit, because I need to say even just from that part given how quickly, and and I think this is a little teaser for for the topic for today. Even just from, what what was that? February? Well, whatever. You know, few months ago when that meeting had actually happened. Probably, it wasn't even February. It was I think it was April. Yeah. April. Yeah. April.
Yeah. Even just from from April, I need to say things have been evolving so so quickly, and I think I may have some juicy new things for you already. So, again, just a little teaser. We are we are gonna jump into it in a in a minute, but, it's really a field that is really skyrocketing in terms of how quickly things are are changing. So, but just just about myself, I'm physician by training. I spent my, kind of first half of my professional life in clinical practice.
My area, has always been oncology. That's my kind of area of expertise. And then, obviously, second half of my professional life in, in pharma. I've been with a few different companies, Novartis for many years, and then most recently with AstraZeneca before joining about 2 years ago now, Johnson and Johnson in, in their medical medical affairs team. So, as I said, oncology for pretty much my entire career with maybe, like, a few exceptions here and there.
So, that's, that's who I am in particular. I need to say, I I have a personal, I would say, interest in anything around innovation and technology in medical affairs, but in general or more broadly in healthcare. So that's really like a big passion of mine and that's something I'm always excited to speak about. And, guys, just so you know, Luca is being incredibly humble when I say because he is one of the leaders, not just in medical affairs, but in oncology.
So I am I am very grateful and pleased, and, excited to have him on the show. And I do wanna announce we actually, this episode is being sponsored by, mslmastery.com, which is a brand new program that's founded by myself and Sarah Snyder. And we are introducing our first program, which is called Aspire MSL, and it's for all the aspiring MSLs out there that are looking to break into their 1st, medical science liaison position.
So if you want more information, just go to mslmastery.com, and you'll get more information on the Aspire MSL program. So, Luca, let's get into it. So a lot of talk about artificial intelligence. I think the place that I would love to start is if you could describe we hear the term generative AI. Can you just explain what that means for us? Yeah. Well, let me caveat, that by, you know, saying that clearly I'm not, you know, technical expert. Right?
So everything I know is really just based on my experience as a medical affairs professional trying to learn from the experts, trying to work with IT folks, programmers, people that really have the technical expertise. So by any means, you know, just don't never quote me on anything I say, especially when it comes to the technical aspects of it. I feel that my contribution is really try to find interesting use cases within medical affairs or healthcare more broadly for this new technologies.
Obviously, artificial intelligence is a perfect example. But again, definitely not a technical expert by any means. But just to try to frame this a little bit, we've been actually working with artificial intelligence now in medical affairs for, I would say, quite a few years. Certainly over the last 5 years or so, artificial intelligence really became more of a mainstream technology that we have been leveraging.
Now the use that we have been doing in medical affairs of artificial intelligence has been very much around kind of analytical power. Right? It's really been around trying to crunch big volume of data and try to make sense out of out of it, try to find signals, try to find trends.
So this has been, really the use we have been doing of artificial intelligence in the last few years in different data sources and different domains starting with clinical data, claims, but also MSL insights or or CRM data.
So a variety of different use cases, but, again, very much around how do we tap into, the analytical power, right, the, ability that artificial intelligence has to find patterns and and identify signals where humans, just because of the volume of data, that's why I think historically we have been calling it big data. Right? Just because of the large volume of data, it's very hard for MSS or medical directors or humans in general to really process that large amount of data.
Now a game changer, it's been, I would say, probably over the last year or so, probably even less than that, it's been generative AI. I think everyone is probably familiar by now with chat GPT. I think that's the closest experience that pretty much, I guess, anyone has now with generative AI. The beauty of generative AI is that not only computers are being used to crunch numbers or analyze data, it can actually give you something, generate something that didn't exist before.
It's actually using data to, educate and train a computer, a machine, in order to enable the computer to give you or generate something that didn't exist before. It can be, very articulate or thoughtful, answer, to a query. Right? That's what we basically do with chat GPT, but it can actually have, different shapes and, and different nuances. So happy, obviously, to, get into some of the details in in our conversation today. I do wanna ask you, like, how you use it.
Before I I get to that, you you you got me really curious because I attended your talk, and you said there's things that and that was in, like, May, April, or May. And you've said things have already changed since then. There's new stuff. You wanna talk about that now, or do you wanna incorporate that later? I'm just curious what you mean by that. Yeah. Yeah. So, again, I need to say, you know, for me, my personal journey with with, generative AI, it's been a little bit of a roller coaster.
I think I was a big, big skeptical, even just probably 6 months ago. Beginning of the year when, you know, the big hype with that, ChargeGPT and in general, generative AI, became a bit more prominent on the news and and everywhere, I would say. That's when I really obviously started to think, okay, what are the use cases? How can I apply some of this in medical affairs? I need to say my initial reaction was, well, I don't really see anything that I can truly do today.
I mean, I can see the potential. Right? I can see how it can potentially be a game changer in a few years from now. But I was looking into different dimensions of medical affairs. I was looking at CHAT GPT, and, you know, I couldn't really find much that, I could really do, right away. However and that's how, you know, this you know, I describe it as a roller coaster and how my journey has been evolving.
You know, the more I started to, get closer to individuals within our, organization who really are technical experts in, in generative AI and GPT in particular. And that's when I started to realize that generative AI is more than just a chatbot or like a smart chatbot. There is much more that can be done with generative AI. Maybe I'm gonna give a couple of examples to illustrate some of the use cases that I see as stuff that we can actually do today.
So, one of the items, I think, everyone by now probably in Making Affairs is familiar with social listening. Right? It's, it's something that I think we became very familiar with.
Actually, when all conferences went dark or, let's say, all in person conferences, medical conferences went dark, with COVID, you know, a few years ago, that was pretty much the only way we could still have a scientific discourse with, peers and other KOS, digital opinion leaders, all of that, dialogue that was happening and asked for some of the other conferences, all of a sudden switched to, Twitter or other social media.
And that's when we really became all or became proficient, with social listening. And one of the biggest, I would say, limitations, at least for me, has always been it takes a lot of time. Right?
Despite, obviously, now very sophisticated platforms being available for social listening, still, you know, it does require, the MSL or a medical director to spend a lot of time doing manual work and looking through tweets, looking through sentiment analysis, looking through a big volume of, data points to really come up with, you know, the insight, really come up with the so what. Now ChatGPT, or let's say, more appropriately, GPT is very good at summarizing things.
And, it's very easy for, GPT to take all the tweets on a particular topic of interest, a particular product, a particular therapeutic area, and, let's say, all tweets that were published within a certain time frame, a week, a month, one day if it's a a conference, can be summarized into 1 or 2 paragraphs by GPT.
And that really gives the MSL or the medical director, very good starting point to, in 5 minutes, 10 minutes, read the paragraph and really get a sense of, okay, this has been the reaction to the data we presented at this particular conference, or this is the reaction that, physicians had on this particular study that was presented by one of our competitor. So Before let me let me interrupt you for a second because this is this is a great use case. But my question is, how do you get that?
So what's the prompt? So you go to chat gpt and you put in a prompt that says, please provide tweets from this event, this date, this subject. Tell me how you get that information. Yeah. And that's where the technical aspect comes into play. We don't use chat GPT at all. Right? So we actually have an ecosystem that is homegrown within within Johnson and Johnson, and it's it's purely exploratory. We are really not using it for now. Mhmm. It's it's pure, like, pilot and the exploratory technology.
That where we download or the or automatically, the system downloads all the tweets on a particular, a particular topic or keyword or hashtag or whatever. And then and then GPT, which is, you know, an open source, technology, does the magic. Right? So, again, I'm not gonna be able to explain you how that happens and the technicalities of it, but the idea is there is a platform. There is a chat gpt equivalent. Platform. Yeah. However, it's much more, obviously, structured.
It's much more codified. We we we obviously control very tightly the data sources. You know, one of the biggest issues, as you know, with JetGPT is that very often makes makes stuff up. Right? It very often gives you information. They call it hallucination. Right? It gives you information that is just not real. And that's, you know, something that obviously we need to try to mitigate.
And again, you know, purely exploratory, it's something that we are doing today and is, I think, already helping probably, MSLs and medical directors to get that initial sense. But again, probably there is, actually, without probably. 1st for sure, there is much more work, that needs to go into this before it truly becomes mainstream. But but again, if you think about what I just described, then you can replace the tweets with MSL Insights, but the exact same process applies.
You can say, okay, I'm going to take all the MSL Insights I have from the month of June on this particular topic, and I'm going to ask GPT to identify 2, 3, 4 themes, and for each one of them, give me one paragraph.
Again, that's never going to replace the human processing and translation of that paragraph into something that is actionable, but still, it's going to really alleviate all the burden of having to go through thousands and thousands of lines of text and usually unstructured data that historically has been done purely manually or insights from an advisory board or potentially one day combining all of that into one data lake and you just query that data lake
and you get the information you need on a particular topic in a very, very quick, fashion. I mean, it saves so much time. Once you start using some of these tools, it saves so much time. So let me ask you, how does does this GPT sit in your CRM to collect all that data? No. It doesn't. It just pulls data from the data sources you, are looking at. So it's obviously like a stand alone, I would say, data platform or or or, data lake that pulls information from the data sources of interest.
Whether it's social media, whether it's your CRM, It literally pulls data from, from the data source you identify, and that's where GPT analyzes and processes and, and, and gives you that summary, for instance, of everything that has been tweeted on that particular day. Right. And so that in itself have access to this? Everybody within field medical has access to this tool? No. No. As I said, it's a pure, it's a pure exploratory, you know, pilot that we are conducting.
It's really just a handful of people that have been playing with this. Yeah. Again, I just want to be very transparent and candid on. Right? It's even though I said, you know, this is stuff that we are doing today because that's true. That's stuff that we are doing today. But again, I I I just want to kind of manage expectations. If you are gonna try tomorrow to do it, don't expect this to be a game changer overnight. It's the case.
It's definitely gonna take much more time for the technology to mature. But again, even today, it can already alleviate some of the workload. It's just a matter of doing it and making sure that, you know, it's compliant, it's consistent, it's it's reliable, I would say. So so what's the potential for this? Like, do you see that this is the way things are going where each organization maybe even I mean, obviously, j and j is enormous. Right?
But let's just say x y z company that's 1 quarter of the size of j and j eventually may have their own internal homegrown system, maybe like a super CRM that has all this generative AI built into it so that it just makes the efficiency of how, whether it's MSLs or even commercial folks, can access the data and analytic tools and and have it at your fingertips so that you can have more meaningful conversations and your preparation for your
your KOL visits is all curated through this homegrown system. Is that part of what you think might happen? Yeah. So, 2 things. Number 1, I I think the in the foreseeable future where I see value where I see value for generative AI is really for, decision support, right, or decision augmentation. What I mean by that is I don't think, we are nowhere close to any kind of decision automation. Right?
What I mean by that is I think, GPT gives MSS or medical directors something to react to and, again, alleviates all the manual work, but it's not gonna make decision. Right? The ultimate decision is always gonna be on what to do, what the next step maybe is always gonna be by or at least again, for the foreseeable future is gonna be done by a medical affairs professional. Now, still, the system can recommend a next best action. Right? It can really help a lot.
That's why I'm I'm using this expression, you know, decision augmentation because it can really augment the ability of a medical director to make informed decisions. But again, that particular action, that particular decision, is still gonna be with the medical director. Now what you mentioned though, I think it's very, very important and very interesting. Meaning that I do believe that generative AI will indeed become more of a mainstream technology.
Today, it's gonna be fully embedded and integrated into the systems we are using today, CRM being perfect example. I think the day that we are not even gonna realize anymore that, there is some artificial intelligence, engine integrated into the CRM, and we're just gonna get recommendations or best best actions, by the CRM, but we are not gonna see any of that. It's everything is gonna be in the back end. Everything is gonna be happening behind the scenes.
I think that's really when the technology may, really become, mainstream. As I said, today is still a very, you know, clunky and, and, you know, IT heavy process that is just not for the masses, I would say. Well and you make a good point. I was at a conference recently for Forbes, and it was all about the future. It's called it was called the Future of Work Conference, and there were some folks from Microsoft actually talking about artificial intelligence.
And they were saying that in a very in a very near future, no one's gonna be using that term anymore. It's just gonna be built in. It's gonna be the way it is. We're using it now. We're referring to it as artificial intelligence as a thing, but that thing as it exists is just gonna be built in. So for example, your if you use Microsoft Outlook as your email editor, it's going to have artificial intelligence functionality built into Outlook.
You're not gonna be aware of it, but there's gonna be things that are gonna happen. So it's going to learn how you typically respond to certain emails, and you're literally gonna start typing and it's just gonna finish the whole thing for you. Or you could just hit certain prompts and the thing will populate like that. So is it going is your email getting replaced by an artificial no. No one's gonna be talking about that.
It's just gonna happen, and you're gonna have a whole new way of doing business. So getting back to medical affairs, what do you think people need to know, now? And how do you think people need to prepare for the future as it relates to some of this technology and these these advancements? Yeah. So, again, I I think there is a a big difference between, you know, the current state and the future state. I completely agree with you.
Future state to me, and that's again when I think this technology will become mainstream, is it's like Google Maps. There is a lot of artificial intelligence happening behind the scenes to the point that Google clearly knows what you like, what you don't like, what kind of food, what kind of restaurants, and when you start typing restaurant, it will show you results that are aligned with the history, where you usually go, where you usually like to have dinner or lunch.
There is a lot of artificial intelligence there. You don't see any of that. You just get results that are customized or tailored based on your history. To me, that's the future and that's what you are describing with Outlook, the Outlook example. So ideally, we're gonna get to a point that no skills whatsoever are required.
Obviously, I'm exaggerating, but this is just to illustrate the point that when the technology will truly be fully integrated, you're probably not gonna need much of an upskilling or much of an understanding of what's happening under the hood. Now, it's a different story because again, as I mentioned, most of it is not happening behind the scenes. You really need to work and partner with the technical experts.
I think to me is what medical affairs professionals need to understand today are use cases. What are the areas where artificial intelligence, generative AI in particular, can help, and what are the areas where probably we're never really going to see any of it, or at least again in the foreseeable future. Number 1 is understanding what's possible. Right? What are the areas that is are currently being explored? What are the areas that, some of the experts are are are currently working on?
That's one thing. Number 2, I think it's starting to maybe learn a little bit of the lingo. Try to familiarize with, again, GPT and chat GPT being a component of it, what's generative AI, what's regular all of that stuff. I think starting to get a bit more familiar with the technology and the nomenclature and taxonomy, I think it's really important. Number 3, I think it's about working with the experts.
If you are curious, if you want to explore some of this, if you want to see how this can help today, your team or your organization, you really need to, again, rely on the technical experts that will do the technical work for you, the programming, all of that stuff, as well as the, you know, compliance and regulatory folks. Well intended that, again, nothing is being used externally at this point in time.
It's pure internal work but still you really need to make sure that your compliance and legal and regulatory are fully aware and fully in the loop. So I think these are the key things that I would recommend people to at least start to think about. Not necessarily that you have to start doing something, but at least start to think about. Well, and I think it's a lot of people I think are afraid of it and they keep putting it off. Well, I don't need to know about that.
That's not something I need to know about. I mean, there's nothing I think I started to play around with it and learn about it, research it, and start to use it in my day to day. And, it's it's not scary. It's helpful. And I think a lot of people, they get scared of things that they're not familiar with. People don't like to embrace change. This is a change. It's a necessary well, not I don't know if it's necessary, but it's it's happening. It's gonna happen whether we we want it to or not.
So as we look at the future, I mean, do you see any pitfalls, drawbacks? Are there areas of caution that you wanna kinda share with people as we move forward in this in this AI world? Yeah. No. 100%. And I need to say, that's why I'm, you know, emphasizing that I'm really, really scared and or at least afraid of the potential unintended consequences that you may have if you use today some of these technologies for anything that is externally facing.
What I mean by that is maybe one day we're gonna get to a point that an HCP goes to a website and ask question into a chatbot and generative AI will respond to that. Mhmm. We are nowhere I I really think that's the kind of stuff that scares me. Right? When I when I see the level of accuracy that, GPT has today, that's really good for me to react to it and, and, again, summarizing things, helping me to digest data, but I would never ever use any of that content, for external external use.
So I I think, it's not necessarily a pitfall. It's more about watch out. This is this technology is really not ready for prime time for anything that is externally facing, anything that is customer facing, patient facing, anything that goes outside or has been communicated to any channel, I just don't think this is ready for for that. Eventually, maybe. I don't know. I mean, I it's hard for me to predict.
Maybe one day it will, but I think that's the, as I said, that's the very, very big, watch out. Well, and I think that's, you know, that puts it into perspective. Use these tools and consider these artificial intelligence technologies that are available to help with our job as humans, not replacing us as humans, in our jobs to be more efficient and to help us, but be cautious.
Because like you said, especially in a medical affairs world where there's such a focus on, you know, the regulations that exist and being compliant, and we can't trust. We have to only trust the people that are most trained to do the job. So how fast do you think this is all going? Yeah. Well, that's the $1,000,000 question. Right? As I said, I I am seeing, how quickly I, myself, I've been, kind of changing my mind and going through this journey.
And even just, you know, 6 to 8 months ago, the whole field and I'm talking about generative AI in medical affairs. I think generative AI overall, I I just I don't think I can really speak to that in a in a formed way.
But when it comes to generative AI in medical affairs, I've seen such a dramatic uptake and such a quick and rapid uptake that if we continue on the same trajectory, I really think, at least again for internal, analytical power and processing, I think it's gonna happen very, very soon. But, now is it months or is it years? That I I probably still still don't know. As I said, I think there are a number of things that we can already do.
It's just a matter of transforming that into a technology that can be broadly adopted by a medical organization rather than just a handful of technical experts. But, again, we went through something very similar. I spoke about I used the example of the social listening technology earlier and, you know, that's been also, like, very, very rapid uptake.
It really happened very, very quickly, and then at some point we reached a plateau and now we know you know, what social listening is good for, what's not so, good at. And, we know the pros and cons. We know the limitations. We know that it's something that we can do, we can leverage, we can tap into, but we also know what's probably not gonna be available at least in the in the in the near future. So I I foresee a similar, probably, proof of adoption, for social listening.
Obviously, it's a much easier technology, but it literally only took a couple of years. My prediction, and I can be completely wrong, is that we are looking at a similar similar time frame. Well, here's what we're gonna do. We're gonna have the same conversation, like, 6 months or a year from now, and we'll see how fast things have evolved. But, Luca, let's leave it at that. I really do appreciate you coming on. That was a really fast, I don't know, 35 minutes, but always loved talking to you.
Thank you for your insights, man. I learned a lot. Appreciate you. No. Thank you. Thank you again for for having me, Tom. I think this is fascinating topic. I'm pretty sure that next time we're going to touch base, there's going to be much more to cover and probably also a bit more hands on experience that we can share with your audience. Well, I'm hoping that I'm still here and it's not like a chatbot. Like, the MSL Talk podcast is still me, and it's not like artificial Tom Carabello.
You know what I mean? That's what I hope. Well, I'm pretty sure about that. I'm pretty sure. But maybe you're gonna have a nice teleprompter with some fancy content that is generated by supercomputer. Maybe that's future. Well, thank you again, my friend. And thank you guys for listening. Thank you for sharing the show and for all your support, and we'll see you next time.
Thank you, Tom. Thank you so much for listening to the show, and if you enjoyed it, please subscribe so that you don't miss an episode in the future, and feel free to leave a rating or a review or a comment. Thanks again, and we look forward to seeing you soon.