What happens when you blend the power of AI with the world of healthcare? Join us for our conversation with Tetiana Telensak, the Head of Product for Data and AI at Merck Group, who reveals how she uses AI to improve healthcare products.
Featured Links: Follow Tetiana on LinkedIn | Merck | 'Five things we learned at the #mtpcon + Pendomonium roadshow - Berlin 2024'feature by Louron Pratt
Episode transcript
Lily Smith: 0:00
Hello and welcome to the Product Experience Podcast. This week, Randy has a chat with Tetiana Telenczak of Merck Group. She's the Head of Product for Data and AI in Healthcare and R&D. The Product Experience Podcast is brought to you by Mind, the Product part of the Pendo family. Every week we talk to inspiring product people from around the globe.
Randy Silver: 0:26
Visit mindtheproductcom to catch up on past episodes and discover free resources to help you with your product practice. Learn about Mind, the Product's conferences and their great training opportunities.
Lily Smith: 0:38
Create a free account to get product inspiration delivered weekly to your inbox. Mind, the Product supports over 200 product tank meetups from New York to Barcelona. There's probably one near you.
Randy Silver: 0:56
Tetiana Telenczak, thank you so much for joining us today, live and in person, here in Berlin.
Tetiana Telenczak: 1:01
Hi, good to meet you as well and thank you for having me here.
Randy Silver: 1:03
So for the people here in Berlin, they've seen you on well and thank you for having me here. So for the people here in Berlin, they've seen you on stage, they've had a proper introduction to you, but the people listening not so much yet. So can you just give us a quick introduction? What are you doing these days and how did you get into products in the first place?
Tetiana Telenczak: 1:17
So I am Tetiana Telenczak. I am Director for Data and Air Products in Healthcare Research and Development at Merck KGIA Darmstadt, Germany, and Merck has a 350-year history. It operates in lifetime electronics and healthcare and has more than 60,000 employees. So I joined Merck approximately three years ago and I was the first product leader in our group data office. So my focus was always, and still is, on building internal products that help Merck to be competitive on the market, and this month, actually, I transitioned to healthcare business and currently I'm responsible for building a new platform to analyze some results of our clinical trials so that America can get all the approvals and release new drugs to the market.
Randy Silver: 2:11
Oh, fascinating.
Tetiana Telenczak: 2:12
And actually my background is consulting, so I used to work at a big whore company and consult a lot of large German organizations on AI strategy, AI governance and building data offices. That's why this topic is really something very special for me, and I have spent this space last nine years and had this pleasure of observing hypes, successes and failures, and I'm very looking forward to our conversation.
Randy Silver: 2:43
So on stage earlier we talked a little bit about all this and I want to talk about your approach to building with AI, but you had a really spicy take that I loved. You're leaning into the AI hype. How are you using AI hype for good?
Tetiana Telenczak: 2:59
Well, there are different opportunities, how you can use this hype for good. In my past, before 2022, before this hype around Chattopadhyay and LLMs, it was very hard to place this topic around AI. And are we AI ready as a company and is our portfolio is ready for AI transformation? It was very hard to place such topics to the senior management. They would not be interested. Especially if your maybe financial results are fine, you would think, okay, I'm fine with that. Why do I need to look on all that stuff?
Tetiana Telenczak: 3:32
And with this hype, this sense of urgency is there and, as one of the presenters today showed, like last year, each CEO at least once what do we do with AI? Are we ready for that? And I think it's a great trend because we want to be competitive, we want that our economy grows further and this AI revolution will be relevant for any kind of business or any kind of industry, and you better start early to experiment, to understand how that works. Is it something for my business model? And if yes, in which regard? Because you don't start just to build it just for the sake that you can show it on your slides. Maybe you still do it if you're a startup, but at least for established companies like Merck, you still need time to figure out how that works.
Tetiana Telenczak: 4:21
So when someone from the business or from your management approaches you and say, hey, can we do this with AI or can we build this LLM-based chatbot, I would always first say yes, but then you can think on your specific situation.
Tetiana Telenczak: 4:39
So maybe you need more budget and you're saying, I can think about that, but actually I need additional funding for this. Or maybe you were struggling with getting more engineers in your team and you can say, well, for this topic, I actually need two additional data scientists and one LLM ops engineer so that they can help me building this. And as a product manager, you still have your job to figure out what will be useful and usable, together with UX research designer, what is good for your users. But you can use this opportunity to negotiate more resources and also maybe do something good for your career because you will be working on some trendy topic that will also increase your value as a product manager, because you get experience in the most cutting-edge technologies on the market. But yeah, that doesn't mean that you just build something, releases it and then everyone is happy. I'm just telling users opportunity and still take care of your users and that they get value out of it.
Randy Silver: 5:41
Right. So if somebody, if my boss, comes up to me and says hey, can we use AI to build this, the first instinct is say yes, even if this is a slightly silly thing that you don't agree with, because you'll be getting the resources, you'll be learning, you'll be getting everything, and by the time you launch that, you'll find five other use cases.
Tetiana Telenczak: 6:01
Which can be helpful. So that's why I would say yes to it. And the second aspect here is to have a dialogue with your manager or with CEO of your company. What is AI about? Because you need to start building expectations on what is possible and you start it as early as possible in this overall journey.
Tetiana Telenczak: 6:23
Because I had a situation in the past as I was a customer success lead at DataCure and I had a client from automotive industry and DataCure has a product so that you can do some data analysis just with clicks, without coding, and the engineers from this automotive company thought that you just do a couple of clicks and then your AI model is ready and everyone is happy and it works in deployed production just with this couple of clicks. But it doesn't work like that. But I think a lot of people who don't have background in AI, but maybe also don't have generally technical background, have very high expectations of how that works, and that's why, when I would answer yes to this executive, I would add but I need budget and please let me explain what is possible to do with it generally.
Randy Silver: 7:14
Right. So AI is not magic pixie dust. The word just doesn't mean we can do something magically and have it just happen. Do you want your CEO or your sales manager, or whoever your requester is, to be really educated, to know the difference between LLMs and ML and whatever else, or do they not need to know that, like in the same way that they don't really understand the rest of the tech stack? How educated and involved should they be in understanding these things?
Tetiana Telenczak: 7:44
It's a great question. It's a bit similar discussion like does any product manager have to code or to know Python by heart and such things? In tech world, there is one role which is very important, which is not a data scientist or data engineer. It's so-called translator and it's people who understand partially business side of processes and they also understand technical processes and they're not necessarily subject matter experts in the business area and they're not necessarily the best coders or engineers, but they can connect these two words together. And why I'm saying this.
Tetiana Telenczak: 8:21
I think that it's important that these managers or executives can understand potential of AI and have basic understanding what it means from technical perspective.
Tetiana Telenczak: 8:33
They don't have to be experts in deployment of AI model or be experts in coding in Python. Some basic understanding is important because later you will ask maybe for a new infrastructure or you ask for additional technical experts for a particular niche in this area and you have to explain to people why. So they need to have a conceptual, general understanding of what is AI, what is possible to do with it, and then leave the rest to the product team to get it done in a proper way and which algorithm it will be. Either you would use more traditional machine learning, like, I don't know, decision tree, linear regression or maybe hierarchical clustering, or your product team will decide to use large language models. Leave it to them. But I think what is very important to have this high level understanding what is potential for AI and what kind of commitments I, as a manager, have actually to do in order that my organization is enabled to deliver an AI product.
Randy Silver: 9:38
This episode is brought to you by Pendo, the only all-in-one product experience platform.
Lily Smith: 9:44
Do you find yourself bouncing around multiple tools to uncover what's happening inside your?
Randy Silver: 9:48
product In one simple platform. Pendo makes it easy to both answer critical questions about how users engage with your product and take action.
Lily Smith: 9:57
First, Pendo is built around product analytics, enabling you to deeply understand user behavior so you can make strategic optimizations.
Randy Silver: 10:05
Next, Pendo lets you deploy in-app guides that lead users through the actions that matter most.
Lily Smith: 10:11
Then Pendo integrates user feedback so you can capture and analyze how people feel and what people want.
Randy Silver: 10:17
And a new thing in Pendo session replays a very cool way to experience your users' actual experiences.
Lily Smith: 10:24
There's a good reason over 10,000 companies use it today.
Randy Silver: 10:28
Visit pendoio slash podcast to create your free Pendo account today and try it yourself.
Lily Smith: 10:34
Want to take your product-led know-how a step further? Check out Pendo and Mind, the Product's lineup of free certification courses led by product experts and designed to help you grow and advance in your career.
Randy Silver: 10:45
Learn more today at pendoio slash podcast. I think back when I was more of a hands-on operational product manager, I very much parted myself in being that translator. I can speak geek to executive and executive to geek, and it was. I would tell my dev teams if you can't do this, if it's harder than we thought, you need to be able to explain it to me well enough that I can then explain it to somebody else.
Randy Silver: 11:18
And it was similar, asking the executive why are we making this decision rather than that decision? But, yeah, being able to say I didn't have to know how to code, but I also needed to know. Well, connecting this to that is possible, but there are complications and we need to give them time for it.
Tetiana Telenczak: 11:36
Yeah.
Randy Silver: 11:37
Okay, makes sense. So, for people who are coming into this and learning how to in this space for the first time, what do you recommend for product people who are really just getting started starting with their experiments? How do they learn the details and the intricacies of how to actually deploy these models?
Tetiana Telenczak: 11:56
It's a good question would start with some conceptual work, even looking into already available use cases in your domain, because probably someone already maybe tried something out and you could find this information or use your network for that. So kind of first get an idea what is this exact business process where I would like to try out how I can enhance it or improve this process or maybe solve some particular business problem. So we kind of have to idea about the domain, what exactly you would like to solve with AI. That's the first thing. Second thing is, of course, you need to have in your engineering team people who are knowledgeable of how foundational models work, like In the past with traditional kind of machine learning models, there was a particular life cycle of building a product. You would start with business understanding that you have data, understanding that you would start coding your model, you refine it a lot of times and then at some point, when you're happy with the results, you would like to deploy it. The interesting thing with large language models is that this phase of prototyping is very short. You can do it within a couple of hours and already see some results. I don't say you get very good results, but you definitely get already something that you can work on, iterate and so on, and that's why you as a product manager need to understand how that works conceptually and have an idea how AI can help in your very specific process or user problem, and on the other side, your engineering team also needs to build this knowledge. That's why we had this discussion today, also during the conference.
Tetiana Telenczak: 13:39
I would start with some idea which might not be the best or most mind-blowing idea how AI can improve your whole life, but with something very tangible, and maybe a simple chatbot is already a way how you get there. Like if I talk about healthcare R&D, there are huge expectations how AI can help we start to like. One example here might be to do analysis of data across different studies. Another example would be, especially in the areas of generative AI, that it can help us to develop faster new molecular structures for the new drugs. But it's a long way to get there, and it's not only about that. You have now one new machine learning model. It's also about changing how we work and how we think about our work, so it doesn't happen over the night. That's why I would always recommend start small and then learn on the way, and maybe at some point you will get up to this very fancy use cases that Google, meta and Po are working, but also in that companies. It didn't happen over the night.
Randy Silver: 14:45
No, Makes perfect sense. So in lots of companies we can start with these experiments and it's okay if we get it wrong, but in healthcare it's not so much okay. So maybe it is for experimentation purposes, but you're not going to be rolling that stuff out. What are the things that you need to keep in mind when you're trying to move from experimentation mode to actual useful production mode? What kind of things do you really need to worry about and bring into the process?
Tetiana Telenczak: 15:15
So currently, for this platform for analyzers of clinical trials, we don't use AI. From our analyzers of clinical trials, we don't use AI. We do statistic analysis of results of the studies, because this is what is required by the regulatory bodies in order that they make a decision yes or no. There are, however, a lot of other products that we build in Merck in different domains that also have AI capabilities. I think the most important thing is, of course, that it brings actually expected outcome for this user group, but apart from that, we have a lot of topics around data protection, data compliance, upcoming AI act, and all these regulations are very important for us so that the users, but maybe even moreck shareholders, can trust us, what we build, and this is very important.
Tetiana Telenczak: 16:05
For example, we have currently an internal version of ChatTPT because Merck has official partnership with OpenAI and people like to put their personal data there. So we build specific functions so that as soon as a user enters this personal data into their chat, it will be automatically deleted because it's not allowed. We don't want to deal with it and it's a very bad practice, but you still know that some users will do it, so you need to build mechanism how this product can be still trustworthy, and there are a lot of such tricks that you need to think about. And whatever we build in this internal version of ChatGPT, we always agree with our workers' council. We go to our internal IT security team, data protection team and a lot of other stakeholders that are important for us to make sure that it's legal and compliant, and it's a lot of work. Yeah, and it's not only something for a large organization. I think that doesn't matter what is your size or how fast you want to be in your delivery. You still have to consider all those aspects in your work.
Randy Silver: 17:18
So to stepping back a little bit to make sure that we're doing everything right. In a space like this, we're not just dealing with user-generated content. We have to have really high-quality data and we have to operate in an environment that we know how to navigate. As in the rules, the regulations, the laws, they have to be something that we know how to navigate. What is that like right now? Are we mature enough in society? Do we have good enough data? Is the regulatory structure there to really support this, or is it still something that we're catching up on?
Tetiana Telenczak: 17:54
I think it's a development phase. Today at the conference, we talked about what it means to have an AI-native product and what do we need to get done in order that it works like that and we talk okay, you need to have this knowledge in your product team, but it's not only about your own team, it's also about the ecosystem in which you work. So if you have partners in your ecosystem and they work with paper well, you can only automate or gather data about yourself. However, there are a lot of initiatives in this area. Even Merck KGAA is cooperating here with Palantir, and they have one joint venture which is called Syntropy, and this is something like a data sharing platform where you can host in one platform like healthcare, data from different institutions, from hospitals, from research departments of universities, from pharmaceuticals companies, all in one place, in trusted environments and with all their security compliance rules as needed.
Tetiana Telenczak: 19:02
And there are examples also in other industries like this about creating these data-sharing platforms between the companies, and it's very important because normally your value chain is not only within your own company. You have different contractors, you have different suppliers and all this ecosystem needs to grow together so that you are enabled actually to build this AI native product, and one example here is that when we submit the results of clinical trials, we submit them in SaaS it's a specific commercial vendor or in our programming language, but we potentially could not submit the results in Python because currently their regulatory authorities would not accept this in Python. Oh, interesting. That's why I think it's a journey for commercial companies like Merck, but it's also a journey for authorities and everyone else involved in this space, and it has to take some time.
Randy Silver: 20:05
Okay, so there's a challenge there. There's also a challenge in making sure you have enough people that can do so. We talked earlier about using the hype to open up in the approval to recruit more people, which is great, but when you are recruiting those people, what are you looking for these days? Are you looking for people who have seven years of AI skill, or are you looking for someone who's got the aptitude and you can train them yourself.
Tetiana Telenczak: 20:29
I think, probably like soft skills is always priority number one. What is your attitude? Are you hungry to learn something new? Are you curious about the last trends on the market? Do you own your topic? Are you burning for your job?
Tetiana Telenczak: 20:46
And then there is a second aspect about your technical expertise, or what I always call data literacy, and I think that the second aspect is also very important, and that's why, especially in my case, because I mostly build infrastructure products for me, it's very important that this person has some technical background. You don't have to have a master in computer science I, by myself, started journalism but you still need to have experience in this space, and when an engineer is complaining about something, you have to have high level understanding what he complains about and how you can maybe help him to get it solved. Understanding what he complains about and how you can maybe help him to get it solved. And that's why, in my case, I mostly look for people with a technical background, for example, data scientists who have interest in more customer and user facing roles, because they in the past normally already built several prototypes and they have then this intrinsic motivation.
Tetiana Telenczak: 21:52
Okay, I want now to build something that really works and I want to understand what needs to be done, that this model is not only perfect from technical perspective but brings that outcome that was expected actually from the business side, and that's why, in my case, I really work like to work with data scientists who want to be product managers. But I understand that there might be another way around. Of course, that product manager gets more technical, but it's a hard job and it's very challenging with all the last developments happening around foundational models. It takes a lot of time, but we have no choice. We have to be on the edge of everything that happens on the market because this is a job of a product manager and it's a lot of expectations on this job.
Randy Silver: 22:37
Well as someone who has a degree in biology and journalism. I very much appreciate that answer, tatiana. Thank you, this was a wonderful conversation. I really enjoyed it.
Tetiana Telenczak: 22:48
Me too. Thank you for having me here, and see you again.
Lily Smith: 23:02
The Product Experience hosts are me, Lily Smith, host by night and chief product officer by day.
Randy Silver: 23:08
And me Randy Silver also host by night, and I spend my days working with product and leadership teams, helping their teams to do amazing work. Luran Pratt is our producer and Luke Smith is our editor, and our theme music is from product community legend Arnie Kittler's band Pow. Thanks to them for letting us use their track.
