0:00Hey, guys. Welcome to the podcast. I'm excited because I have a good friend here. I've been trying to set this up for a while. My buddy, Vera Kudsenko, who's, like, the AI expert genius. I know all you guys wanna know about AI and everything about AI. Well, she knows all. And we've been talking about doing this, and it's taken a while. But here we are. Vera, welcome to the podcast. Thank you so much for having me here. Super excited to talk about this this topic with you, and and so happy to have finally
0:32figured out a time right before the holiday season. Yeah. Yes. So, guys, Vera and I met, like, on the conference circuit, really. See her at all the conferences. You just became friendly and have talked about this, but, you know, super busy with things, and, it took a while. But she came up with the idea to talk about the AI gap in medical affairs and why 95% of Gen AI projects fail.
1:03But before we get into it and that's what we're gonna talk about. Before we get into it, I'd love, Vera, for you to do an introduction. Just kinda tell everybody who you are, where you're from, all that good stuff. Yeah. Absolutely. So hi, everyone. Name is Vera, founder and CEO at Atrix AI. We work closely with medical affairs teams to help them operationalize AI so they're not in the 95, but rather in the 5% from our discussion today. And I'll throw out a a really fun fact. Over ten years ago, when I was working at Facebook, one of my projects was actually to
1:37take news articles that you would share at news or, like, on the feed at the time and to use AI to help summarize those articles. So it's funny because you fast forward to now when GenAI, you know, now with Chad GPT, it has been kind of societally viral for, what, three years, and you can probably do that now in five minutes. And and back then, that project was, like, months long. Right? And and and definitely, like, very hard to accomplish. So it's it's very much like a full circle moment, I feel like, in a lot of ways to see what's happening now.
2:08Oh, I love it. I love it. And we're gonna get into it. Before we do, guys, I just wanted to recognize for the past, like, jeez, I'm gonna say couple weeks, even couple months. I've been talking to so many people that are either out of work or are desperately on the job market. Guys, hang in there. I know it's it's a tough time. We got the holidays coming up. So I just wanna give you guys a word of encouragement. It's not you. It might just be the market. It's timing. Things are gonna get better.
2:39You're gonna find a job. Check out my book. This is not a shameless plug. I really think you should buy my book. It's called job search mastery, how to win your dream job, only because it's helping so many people. But I just wanted to mention that because we're this episode is gonna come out, like, around Thanksgiving. So want to make sure that, you know, I had a real brief word of encouragement for you guys. So, Vera, let's get into this, and let's start with the basics. Because you mentioned the AI gap in medical affairs.
3:10So tell us, like, what does that mean? Yeah. Absolutely. So I I think when it comes to the AI gap in in terms of what I see when it comes to Teams, you know, over the last, like, year to year and a half, we've been doing a lot of training and operationalizing AI. And there there's a sense of, you know, from a training perspective, everyone wants to learn about it, and that's wonderful. And you absolutely need the the training piece of it. But what we oftentimes will see will happen is, you know, you do the training. Maybe you open up Copilot for that thirty minutes, right, of the training to try out the
3:44different prompts, and then you see after that training, things can oftentimes fall off. And part of it, is is that, you know, you want the training to enable consistency in in terms of integrating those skill sets, right, because there's a reason why you're doing it. And the gap is almost going from that training to trying to operationalize AI solutions without getting folks properly, set up to to to use AI consistently.
4:17Mhmm. I'm assuming are you at it? Learning curve. Right? It's so you're talking about, like, kinda like that learning curve is the gap where it's like, you can learn something. There's training, but then it's like, you can't just expect it to be operationalized because there's that that period where people just have to kinda play around with it and figure it out. Like, I think I understand what a prompt is. I you know? Right? Am I right about that? Yeah. Totally. I mean, think about, for example, like, when the iPhone came out. Right? You know, was the first thing that you did was it, you know, read a book about iPhones, right,
4:49and and how to use them? You probably went you know, you got your actual phone, and then, like, you you started playing around with it. You saw the texting feature, and you're like, oh, cool. It's, you know, it's virtual buttons now instead of, like, physical buttons. You saw, oh, you can send images or, like, oh, that didn't work or, like, oh, I can't unsend it. Right? So you learn you learn what works and what doesn't by actually using it. So so in essence, I feel like with AI, it's it's the same thing, where with the trainings, you are getting some of that practice, but you need to almost continue that practice.
5:20And and the second point there, I feel like, is having proper expectations. You know? I I think oftentimes if you try one thing and you don't get the result you're hoping for, that doesn't necessarily mean the AI model can't get you there. It's that you need to be open to, you know, experimenting a bit or or maybe this thing fails, and that's okay. Because similar to any technology that you're trying out, you're gonna have things not work. And I think sometimes I also have this undue expectation of AI to, like, perfectly work on
5:52everything. Well and you you hit the nail on the head. Like, no one reads the instructions. I didn't read the instructions on my iPhone. I just grabbed it and started playing around with it. Right. And you go through training, and the training should be the instructions. But at the same time, until you use it and make it a part like, it a habit and part of your routine, you're not gonna get good at it. And I think part of the reason there's a gap and this is just me in layman's terms. Part of the reason there's a gap is because people aren't they don't you know, they're not
6:26always receptive to change. And there's a change in routine. There's a change in habit that and behavior that has to happen in order to really truly adopt AI. So I think one of the things you should do is schedule it. Put it on your calendar. Schedule time to get on whatever tool you're supposed to use, whatever training you got, you you know, you received, put time in your calendar to implement that. And that's how you know that it gets operationalized. So let's talk about you had said that 95% of Gen AI projects fail.
7:02So let's let's kind of break that down and figure out why there's such a big disconnect between the expectations and the actual results. Yeah. Yeah. Absolutely. So I I feel like with with the premise of executing and operationalizing AI, oftentimes, what how typically teams can approach it that can set up, the project to potentially fail would be, one, going in with the wrong expectation.
7:33So maybe if I had to summarize, it would be, like, two to three things. I think, one, oftentimes, folks are looking at AI as a hammer, and they're trying to find any nail to to, you know, use this hammer on. And they look at AI as AI to solve literature review, AI to solve field insights versus starting with the problem first. So meaning, you know, if we look at our team's objectives, right, or as an individual, if I look at, like, my day to day, right, or week to week, which things are taking the most amount
8:06of time that will help me achieve my objectives, you know, personally or professionally to the team to the org level. They're typically not starting there rather than they're starting at the feature level. So what I mean is we need a chatbot versus we're trying to solve insights reporting, or we're trying to solve the literature review process because we spent half of our time on that where we wanna spend more of our time on strategy. So I think that's that's one key one. Got you. So let's talk in medical fair specifically.
8:38Where do you see most teams stand right now, like, as you look at your clients in their AI journey? Yeah. So I would say there there are kind of, I would say, three three buckets or maybe three phases of the journey. I think the first phase of the journey is generally the the change management piece. So this is kind of what we're talking about initially with the trainings. People have to be comfortable with the concept. They have to understand how AI can benefit them versus what are kind of, like, the risks and
9:11downfalls, if you will. So a lot of folks are starting to establish trust when it comes to AI by doing training. So there are some folks that are still at kind of that stage. Then the second stage, I would say, is really folks who we who started to then establish Copilot Enterprise or GPT Enterprise, which kinda ties in with the training. Right? Then the bucket after that is companies who realize that Copilot and GPT are only gonna get you so far. Right? You can't expect Copilot and GPT to review a million pieces of literature to, you know, dig
9:45through all of your medical strategy plans and, like, come up with all of the unmet needs, you know, that you should be targeting with your content. It's just not realistic. So then that goes more into kind of custom AI or more advanced AI solutions where teams are are starting to pilot various use cases. So I would say a lot of folks are in in that bucketed category, and then there's few who's who were in this bucket last year who are now starting to scale what they saw work from the previous year when it comes to some of these, like, custom AI solutions.
10:18So it's almost like training enterprise GPT Copilot to piloting early high impact use cases to now, like, we're gonna scale some of those cases. That makes sense. And so why do you think the adoption of AI in for medical affairs organizations has been challenging Yeah. Compared to, like, other industries? Yeah. And I I think part of it comes because medical affairs, you know, is they're such scientific
10:52experts. Right? Everyone's a PharmD, MD, PhD, like, so absolutely knowledgeable about their therapeutic areas and how science and proper education to clinicians and patient needs to be done. AI is also a very specialized field. And, you know, they while they're MDs, PharmDs, they're not AI, you know, PharmDs or AI PhDs. And I I think because AI can feel sometimes, like, this black box of how do I use it, when
11:25do I use it for what, there's almost this kind of gap of communication of medical affairs understanding, okay. I have this problem. So how do I understand the language of AI to know how to to solve my problem? So what ends up happening, and this ties back to to the hammer analogy, is, you know, a team in medical affairs knows literature reviews the problem. Okay. Great. So now I have to go and find a literature, you know, AI or tech solution.
11:56So then now they start engaging with partners or maybe their internal IT team to build something out. And then the challenge becomes in communicating with with the IT team or the vendor partner in how to translate my literature review problem to know how to evaluate whether a vendor is the the right fit vendor or how to tell internal IT teams, like, here's what I want built. So what ends up happening sometimes is you, you know, you might work with an internal IT team
12:26or you pick a a vendor partner, and then only three, six, nine months later, you realize that there's actually limitations you might have not known about in that process. Or IT delivers the project, and then you realize, oh, actually, this doesn't fit our needs. So now we have to go back to the drawing board and iterate again and again. And this is where I see one of the big execution gaps. And part of the reason why I think the 95% exists is because there isn't that, like, common translation, if you will, that helps bridge the needs with the IT requirements.
13:00And you said to me that technology isn't the problem. You said it's culture, governance, and people. Can you unpack that a little bit? Explain that. Well, I think some of those points that you talked about are or that we just covered are kind of encapsulated in that in the sense that, you know, you need the the right change management, which kind of comes with training. It comes with leadership. Right? Setting up expectations that you should use Copilot, and and you should be open with an
13:32experimental mindset. And having champions within those teams who are AI forward, if you will, showing others, right, how they're leveraging AI to to just help the entire team. And that's kind of, I I would say, like, the change management, like, cultural piece of enabling I get you. Yeah. And then oh, go ahead. No. I I get you. I think it's we're all we're still talking about the same issue. Right? It's it's all kind of that adoption and operationalizing it and making it a new
14:06standard for how people are going to work. I just I love the idea because you have me thinking as you're talking about, you know, starting with the problem. Like, I I really kinda feel like that's something that organizations really need to rely on as they continue to expand their efforts efforts in in in AI versus just categorically saying, okay. It's for this, this, and this. Because we hear so much about use cases.
14:36Oh, you should be using it for this. Well, is that solving your problem, or are you just saying that you need to use it for that? Right. Well and and maybe for, you know, for that team or company that's speaking about AI for that one use case, that is also tied to a core problem for them. But, like, if that's not a core problem for you, I mean, you could use AI. You could spend a lot of money using it. But, like, is it worth the ROI? Right? And you're right. I mean, so many trade shows, we we see various use cases, right, or folks talk about how they're applying AI.
15:08So it's almost like collectively, we're aggregating, like, here's a library of use cases. So, oh, should we pick from a menu versus, hey. What's my biggest problem? Is my biggest problem, you know, helping identify the right patients because I'm in rare disease or helping educate clinicians on how to identify the right patients? Well, if that's the highest unlock opportunity and problem, then how then it then how do we solve that? And maybe sometimes AI is, like, not the solution. Do like, maybe you don't need it sometimes.
15:38So Yeah. What So maybe it's a group. Maybe it's a department. Maybe it's a person. Maybe it's a committee. Maybe you already have the solution, and you're just throwing it at AI because that's the new way to do things. Yeah. Yeah. Maybe. But I do like hearing about use cases because I think it's really interesting to see what people are doing. And, again, to your point, how that's solving their problems, and then it gives you the incentive to, like, well, I'm gonna try that.
16:10Yes. Right? And and I think to to add to what you're saying where I think sharing use cases is extremely helpful is to really plant seeds of ideas. Because maybe in seeing, okay, AI can be applied in this way, it doesn't mean that I, as a team, have to apply it this way, but maybe this processing seems in parallel to another process that is relevant for my team, and then I can kind of translate that. So so ironically, actually, maybe use cases can be can be helpful in bridging that gap we
16:40talked about between IT language, if you will, or properly evaluating vendors and, like, the actual business needs of medical affairs teams. Yeah. Let's talk about skills and and what capabilities medical affairs professionals need to have nowadays to speak AI. Yeah. I mean, I will say, to to your earlier point about really encouraging folks who are in the job process. One thing I've heard consistently at trade shows, like, recently cohosted the AI Immersion
17:14Lab with medical affairs, and a lot of the leaders that were at the lab talked a lot about trying to source for folks for their team who had a medical background and AI knowledge. So it is something that I'm hearing too on the ground that's kind of bubbling up as a skill set to to have. So I just wanted to highlight that. And then I think from a skills perspective, the the best way, honestly, is to just you. Like, use Copilot or use GPT, whether it's in your personal life, because that'll help you
17:46get familiar with it. Just like you use a phone in your personal life, so then you also know how to use it in your professional life. Right? Right. And and using that and also from having a, you know, a previous career like, if you were an MSL, right, you you already know the pain points of an MSL because you've been an MSL. So the more you kind of experiment with that technology, the more you're able to the pieces will click, and you'll start to get ideas for how even a tool like Copilot or Chad GPT can be
18:18helpful on the day to day. Right. So I think well, I'm curious on on your perspective, Tom, on this as well from a recruiting standpoint. Well, you know, it's funny what you just said. I'm like, wait a second. Let me see if I got this straight. So you're saying that you went to this lab, and they were talking about this thing bubbling up about how AI is sourcing candidates. I'm like, jeez. I'm out of a job. That's it. They don't need me anymore. The quintessential fear of AI. You know, it I mean, that's an interesting point because there's also been you know, I
18:50think New York Times recently put out an article about new grads, you know, struggling to find jobs. But I you know, I've not dug into the statistics that are around that. But, you know, personally, as a leader who is also recruiting and sourcing, I would I actually would be totally fine with junior talent because they tend to be the ones who use AI. Do know what mean? I do think in in the new world, you know, with with this new technology evolving, the more comfortable folks get with using AI, the more things that you can do, the more, you know, the
19:24things you can be productive on and and actually become competitive in the job market. So I actually have a very maybe opposing viewpoint to to to the statistics. But Yeah. No. And I think that when you look at both scenarios, the scenario of me jokingly saying my job's gonna be replaced by AI, I just think that there's certain things that a human's going to do a lot better than a machine. And and, again, coupled with what you were saying about hiring, whether it's a junior candidate or any candidate, you, as a human, need to win the job.
20:02You need to I need to prove my worth and my value because if I'm more valuable than AI, then my clients are going to come to me versus using a machine. It's the same thing with a job seeker or job applicant if they can show their value to the organization. So you as a business owner and as a CEO, you want people that are gonna bring value to you, to your organization, help your organization grow.
20:33So how's how are you gonna do that? Yeah. There's gonna be certain things you're gonna do with machines, but there's gonna be people that you need on your team that are gonna do it. So I think that's the distinction for me. Yeah. And and I think to, you know, to your point, I mean, I would almost look for people who use AI all the time. So I'm like, that's great. You're very familiar with it because I and and here's why. Here's why. Like, to your point, I think that there is specific value. It has to be a symbiotic relationship between humans and AI.
21:04I I don't think AI will replace, especially, human centric jobs and also especially in a regulated industry like pharma. And and here's why. This actually kinda ties to the the compliance piece or, like, the governance piece that that you earlier, which is ultimately, at the end of the day, if you wanna leverage AI to help streamline anything, you know, whether it's insights, whether it's personalized outreach, you know, whether it's content, US, anything, it it has to you have to make sure it's
21:35correct. You have to make sure it's accurate. You have to guide the AI model. Because an AI model, when you use it, think of it as, like, an intro. It doesn't have your subject matter expertise. You have to guide it with your subject matter expertise for it to output what you think is gonna be ultimately be relevant, and you have to you have to check it. So there's always going to be a human in the loop. And an AI model, guess what? The other thing is AI models were trained from data that guess who created.
22:07We did. I'm sure. Yeah. It's not like it's not like sentient. Like, what Right. It it's always going to be a symbiotic relationship. So that's I have zero qualms about, like, AI replacing anyone. And this is why for me, I look for candidates that use AI frequently because I know they're they're so familiar, and they understand that symbiotic relationship. So so to those who are maybe afraid of AI taking their jobs, I I certainly I I don't think that'll happen. So Well and I think, you know, there's been a lot of talk about the fear of of AI and which jobs
22:42are gonna get replaced, which ones aren't. And, obviously, so many MSLs and medical affairs professionals listening to this podcast, we've talked about it. And I really do think that there's so much nuance to the MSL job into what medical affairs is responsible for doing that I don't think you can replace what medical affairs is responsible for completely through AI. So that's maybe that's a biased opinion because I make a living off of and from medical affairs
23:15and, you know, and servicing medical affairs organizations. But I do believe that. And you're you just said the same thing. So we're on the same page. We're definitely in agreement with that. Yeah. I mean, I I don't think AI can replace human trust, human relationships, and and the the the expertise because the expertise has to be fed into it. You know? And human think about, like, your, like, a customer support. You know? Like, you call on the phone, and you have to click, like, 10 different numbers. Or, like, you go to the chatbot because everyone actually now has AI, you know,
23:48chatbots that try to questions, and you're only just, like, human. Like, that's what I type. I'm like, get me to human. Get me to human. How do I the bot? And, like, I I fundamentally I I don't think that an MSL, right, in in medical affairs is is just such an inherently educational communication relation voting job that it's the least likely to be at risk, in my opinion, to be replaced by AI. I agree. I agree.
24:18But let's talk about resistance to AI, fear of AI, and how organizations can get over that. I know when we look at the adoption and the organize the operational side of it, do you think and I say we're getting back to this. I know I keep coming back to the same thing.
24:49But the the reluctance piece and the fear of it from a user perspective, do you think it's it's in the the guidance from the organization, or do you think it's more from the what's the word I'm looking for? The capabilities of the individual. Because there's some people that are just good at certain things and not. But I keep hearing from people, like, things like, I haven't gotten there yet.
25:24I'm not good at it. I haven't adopted it yet, or it's something I still need to do. I don't understand it that well. To me, that's all part of this fear and and resistance and reluctance. Are you seeing that too, or is or am I just seeing it? No. I I see that a 100% as well. And I think a a part of I feel like there's kind of a a couple of components there. I think there's this, the first component of, you know, how do team leaders, right, and managers create a an an encouraged and safe environment for their teams to experiment with
26:01AI. I think there's kind of that component, and there are ways that they can do that to encourage folks who might have that fear to be a bit more open to leveraging AI. And then there's another component of, you know, what what you're talking about, I I feel like, which is, you know, I don't know how it works, or it's like, saw an output from an AI, and it was wrong. So, therefore, all AI is wrong. Do know what? All AI is wrong. And and is it on me because I'm not asking the right question, or or is it you know, the AI tool itself is is not good?
26:33So that's almost like another bucket of what I see. And part of that second bucket, if I was to dig into it, is, I have two thoughts there. I think, one, fundamentally, no one who ever uses any technology should blame themselves for it. I think the technology should be designed in such a way that it should be easy to use. So I would never want someone to feel like, oh, it seems like this intimidating thing or, like, I'm doing a prompt wrong or I'm doing a question wrong. I think that's you know, I I this is a product opinion.
27:05As as someone who builds products, I think you should try to design products in such a way where that doesn't happen. But the the opposing side of that of where I see that fear pop up sometimes experientially where, let's say, a person has had experience with AI before, whether it's a vendor or or internal IT, you know, AI product, and they're like, oh, you know, like, I tried it, and it was wrong. So, like, therefore, like, I I I you know, I'm more I'm I'm more hesitant and more fearful of AI now. Typically, why that happens is because they there hasn't been full transparency or maybe
27:43education on how AI behind the scenes is is making a decision. And what I mean by that maybe, trying to think of, like, a a specific example. Let's say I have an AI chatbot, and it's plugged into all my field notes. You know? And I ask it a question about a specific therapeutic area and product, and I see that the AI model pulled in some notes about that therapeutic area and product and then pulled some notes about a completely different product.
28:14And now you're like, oh, it's giving me a summary that has, like, what I was looking for and then, like, a mix of what I wasn't looking for. So, therefore, AI is wrong. I'm never gonna trust it again. Oftentimes, that happens because there isn't there has been education done on how the AI model is making some of those decisions to pull that data. And, therefore, if the the person doesn't feel in control of being able to tell AI how to pull the right data. And where I've seen it reverse, like, if someone comes in very skeptical and then starts
28:47to trust it more, it's when there is a feeling one, an explanation of how AI is doing everything so there's transparency. And then two, they're able to put in their input and kind of provide the AI model guidance. So if we take that same example, rather than pulling it from two you know, those two different sources, it now pulls from, like, the right source because of their input. And then they see progress, and then they start to build trust. I don't if that answers your question, but I I felt compelled that that is a big pattern of what I also see.
29:18No. I I think I think that's like, obviously, the whole purpose of this conversation is what's gonna resonate with people? What's gonna help people? Folks that are listening to this, what can they take away and say, oh, I'm gonna try that. Oh, I that's me, and, you know, I need to do this. So that's why I bring it up because I feel like there's so much talk about this. Mhmm. But at the same time, I think people all feel like they're just not good at it yet to a
29:51certain extent. They're getting better. I know I'm getting better, and we all are. But, you know, at different levels, at a different pace, I think everybody's going at a different pace, which is why I keep bringing that part of it up to get people to be more comfortable. I'll tell you one of the things that helped me, and this is actually kind of a funny story. But and my wife did this. She told she told me. She's like, you gotta do this. It'll change your relationship with I use ChatGPT. So you go to ChatGPT or whatever, you know, Gemini or Copilot, whatever you're using, and ask it what
30:28it wants you to call it. Okay. Give it a name. Right? So what's gonna happen is you're gonna say, hey. I what I said was, I really appreciate all of the help that you give me on a daily basis, and I feel it's appropriate that you have a name. What do you want me to call you? So it responded back and it's like, oh, thank you so much, Tom. That's really nice of you. Just call me chat GPT or chat or chatty.
30:59And I was like, no. I really want a something personal that would fit our relationship. So we came back and it had all these different things. And one of the things it said was something along the lines of, I feel like a lot of what you're trying to do in your role is to help people gain momentum. So there was a there was a name that had something to do with momentum. So I was like, I'm gonna call you Moe.
31:31Oh, I love that. That's so fun. So my chat GPT, I call and Moe loves it. Moe's my buddy now. Right? So I have this, like, different relationship. So I encourage you guys, get comfortable with whatever AI you use and give it a name. Oh, I love that. That's so and it almost makes it seem, you know, kind of to analogy of treat chat like an intern. So it's like, you know, Moe just joined the team. How are you gonna make make sure Moe is successful? You're not gonna tell Moe, like, you did this one thing wrong.
32:03Therefore, like, I'm never gonna trust you again. Do you know, like that that would, you know, that would suck. Like, how sad? And it really personalized it and humanized it to the to the extent that the response was, I love it. Call me Moe. Whenever you need me, just say, let's go, Moe. Oh, I love that. Where to god? So it's just a fun little thing you guys could do. Personality now. I love that. Yeah. So getting back to what we were talking about before, I know we talked about 95% of the
32:39projects fail. So what would you say the 5% are doing? Like, why are the why are five per what's the what's the secret sauce behind the success rate of the 5%? I think with with the 5%, you know, first, they start with with the problem. And and part of the reason of why starting with the problem versus just, you know, where where can we fit an AI and let's just try that is is successful is because the second reason, which is you have to take an iterative approach.
33:09So if if you're solving a critical problem for the team or the business, there's there's resources and support behind that. Right? And there's an understanding of because this problem is critical to solve, like, we will take an iterative approach to solve that problem. So, you know, second, with the iterative approach, it is also this understanding that, you know, Moe of your team that's gonna join, you know, as AI is a blank slate. So you you know, first, what what are you gonna do with Moe on day one?
33:41You're probably gonna give him the basic SOPs or her. And then you're you're going to see how Moe does, and then you're gonna be like, hey. You know, this was great, and, you know, these are areas for improvement. So now, you know, on day 60, Moe's a lot better. And I would almost look at the AI journey in a similar way. It's not necessarily day zero. It's going to be, you know, perfect to, you know, to to spec, but you have to take that iterative mindset. So that's the second key of of success.
34:13And then the the third is the change management piece, which which really boils down to leadership has to be bought in to this technology. And there has to be, in in my opinion, I think, org or team wide support and encouragement, which you can tell, right, if companies have trainings, if they are putting budget behind this, if your managers and and leaders are talking about it, ultimately, that is a really positive thing because it it it's it enables you, you know, depending on the team you're working on, to go and try these things out.
34:45So if I had to summarize, like, very high level, like, those would be, like, the three key characteristics. So when you say iterative what I understand that as is there's, like, first iteration, second iteration. So it's like a compounding effect. You start here, and then you work your way. Yes. Exact and and oftentimes, what's helpful in that process because I I think a big part of it, especially in this industry, is how do you how do you trust AI as you know, you need
35:15things to be accurate. You can't have AI hallucinating here and there. Like, at a certain point, especially for these high priority problems, like, you you need to be able to rely and trust it. And what's oftentimes helpful in this iterative process is you wanna know what you're iterating towards. Right? Right. Upfront defining what is the bar for quality, accuracy, right, or any other characteristic that's really important in solving this problem. And part of the iterative process is you're aligning with that rubric until it gets to a
35:45certain point where you feel comfortable. So an example is, you know, most humans roughly have, like, a three to 5% error rate. Like, even us. You know, we make mistakes too. So Moe if Moe comes in and, you know, he he or she is three to 5% wrong, that's kinda human level. You know? So maybe that's an acceptable bar. But in some cases, you know, you need it to be 99.999% accurate. Right? So it's important to establish that ahead of time, and then that's that's a really key
36:17component of that iterative process. So I read a book on AI, and and one of the things that I learned, and and I wanna hear your opinion on this, is when starting out on your AI journey, one of the things this guy said to do was, start by asking your AI tool, how should I be using you to do x y z? But be very detailed and explain. This is what I do.
36:48This is what who I serve. Blah blah blah. How should I be relying on you to help me? What do you think it like, and, obviously, you're a prompt genius. Like, it I know that that's probably rudimentary, but talk to me about where that fits in the iteration process. Well, so so the ways yeah. Yeah. I feel like what you're kinda touching on here is also AI is such a great tool to use to help you, a, generate prompts to to help you create form, right, and and ideate, which is kind of
37:22what you're pointing to. So when when I think about what you just said in the in the iterative process is you can't do that, and you can ask it like, hey. For this, you know, problem that we're trying to solve. Right? I know. Literature review takes a long time or patient identification. What do you think are, you know, good quality criteria for what an AI model needs to have to fulfill its job in solving this problem? And, likely, you're gonna see that it'll it'll, you know, spit out things and some things you're gonna be like, I don't know if that's relevant, but that's okay.
37:55But if there was one or two that are, you can continue to riff off of it. Right? So I love that. Leverage AI to help you with that process. Well and you brought it up. You just said it. There's the other part of that is ask AI to give you the prompt for it to tell to solve whatever problem or give you the details you need. And I'm like, wait a second. Because sometimes you're like, okay. What am I what's this prompt gonna look like? And so sometimes it's just like, please give me the prompt because I need to do this.
38:27Right? So I don't know. I I I I think it's I think it's such a it's a journey. It's it's, like, perfect. This this conversation is exactly what it is. It's a journey. It is. And I I was gonna say to to your last point, I actually just lost my my thought on because I keep interrupting you. I can't Oh, this is no. No. All good. Yes. So this is it. Because this is what I consistently also see in in trainings and as well as in trade shows,
39:02which is, you know, you you try to educate on the concept of what is a good prop. You want it to be detailed. You wanna take a role, etcetera, etcetera, just like you would tell, you know, your intern mode, right, on the first day. So that can be intimidating because you're like, wait. But my job is not to write essays. So, like, how how how am I gonna make, you know, Copilot or Chad GPT accessible on a day to day basis? Well, you know, great news. You can you can ask GPT to write you that essay so that you can use that to then get to the ultimate outcome that you want. So, I just wanted to highlight that.
39:34Vera, you are amazing. We are at time. Wow. We went a little bit long, but we had to because we had a lot to cover. And I just I knew that this was gonna be an awesome conversation. So thank you. Thank you so much for for coming on. Really appreciate you being here. Absolutely. Thank you, Tom, for having me. This was such a fun discussion. I wanna you about it on the recruiting side. Thank you. Gonna do it again. We're gonna do it again. So alright, guys. As always, thank you for your support. If you got value in this, please share it with others.
40:06Thank you for all of your comments and connections and all the stuff you do on LinkedIn, and appreciate you guys. Appreciate all your support. Love you all. See you next time.