Latest / AI Meets MedTech / Inside AI Transformation in a Large Healthcare Company
Transcript
- Anna Nesterova: We hear about AI transformation all the time. But stripping away the buzzwords, what does it actually look like inside of the company of 50,000 people? Today we find out. Welcome to AIMets MedTech, the podcast for anyone who wants to understand how AI is reshaping healthcare, whether you are investing in it, whether you work in it. Or simply depend on it as a patient. I'm your host, Anna Nisturova. Today's episode is part of our mini-series from the second MedTech Mits AI Conference in Barcelona, where I'm sitting down with a few keynote speakers from the conference ahead of their sessions. And actually, their sessions start tomorrow on October 14th. Today, my guest is Carolina Hensel, Global Director of Data and Analytics at GE Healthcare. Carolina, thank you so much for joining. Karolina Henzel: Hi, thank you for having me. Anna Nesterova: Carolina, you work in one of the largest med tech companies in the world, and you are the one who sees how AI is actually being adopted on the inside. can you tell us a little bit more about your day-to-day work? What does your team do on a day to day basis? Karolina Henzel: So at the high level, my my role is to connect data analytics and AI with the real business decisions and also focus on the business outcomes. We do not try not to focus on the IT solutions, but on a business outcome. So our work is not simply about creating, developing dashboards or AI models. We are focused on the data foundation and AI capabilities. but also to be the advisors for the business. Listen, what are the challenges, what are the goals, and advise how AI or advanced analytics can help them achieve them, or deal with the challenges. So I would say that day-to-day work focus on the or combines a few pillars. first is data foundations. I know that for years we talked a lot about the data quality, but even though we talked a lot about that, it's still not solved and still challenging, especially if we talk about the foundation for the AI agents. Sometimes it's not about the having not good enough data, but I mean not enough granularity, or even we don't have some of the data in our enterprise systems. So, what we build in G Healthcare is our data products. We standardize key business definitions, work with the data owners to strengthen governance and data quality. And the second area is integrated analytics. What I mean by that, we have, and in all organizations, we have multiple reports, BI reports to monitor performance. But what we try to built on top of that on top of that is a agent with with this agent you can have the conversational analytics and ask the question about the data. But this first step needs to be achieved, meaning we need to have a trust to the data and good data foundation to be able to have the conversation based on those data. And what agent also and how The agent can support the leaders is to detect early signals, find underlying drivers, give the recommendation, and it really offloads the leaders and data community in finding relevant data on reconciliation that is needed between the report, or even get to the number that is hidden somewhere in the and in the report. And the third area is is AI solution. And we started I know like in G Healthcare we have AI embedded into product for for years. but actually one year ago we started AI transformation of our operations, and it was it's really cross-organizational effort because we know that It's not the most difficult is not to implement the AA model, but is to adopt, to scale the solution. That's why from the day one, the business leaders, change management office in is engaged in in this process and in this program. And actually about those key areas that we choose to this transformation, I'll be talking during the the conference because we had a lot of ideas. A lot of use cases coming from the leaders, how AI can solve the problem problems, but the challenge was to pick the right one, to Anna Nesterova: Mm. Karolina Henzel: have the focus of the organization on only a few that can that have the biggest impact on a bida. So I will be also talking a lot about the adoption during during my my speech at the conference. And and finally. A very important part of my job is is adoption and value realization. So AI solution for me is not successful when model works in a technical environment, but when it becomes the the the the part of the business processes, day to day work. It's a part of the business business workflow. Anna Nesterova: without giving up your keynote too much, what you just mentioned about the adoption and the challenges with adoption, can we touch on that? So do you think the challenge is mostly lies in people or in culture? Because you said the technology is not the challenge, it's the adoption part. So what it is, it's a leadership problem, maybe some leaders are not. familiar or maybe they're interested but they don't know how to engage their teams to adopt it. So what's your view? Karolina Henzel: Actually, it's like everything what what you mentioned, but I would say that it starts with a IT team, data science team, that we underestimate the adoption effort. And when we think it's it's very quickly we can move to the proof of concept and prove the value, show the opportunities, but then comes to the realization and making that part of the business process making a part of the tool that business use to make decisions and some sometimes we for forget about it so starting i would say that this is starting starts from the leadership so without a top down support of the of the leadership that this is our goal that we as a company want to be an AI driven and explaining why educate being the advocate of FAI Also dealing with the fears of the fear of the people explaining why it's so important for for the company. this is this is let's say a starting point, but also building such a program is not forgetting about this adaption plan. So in G in Healthcare, we from the same beginning we had a change management office who worked together with us, with the leaders. To define what will be the adoption path to Anna Nesterova: Mm-hmm. Karolina Henzel: change the ways of working. Also explaining how this AI solution can support what the people do. So this is these all elements need to come come together and the AI needs to be embedded into established process. So the recommendation. cannot sit in a separate application because it requires people to change the entire way of working to to change the entire way of working and it may never be adopted. So AI creates the value when it's support it when it supports an existing decision, an operational rhythm or problem-solving process. And also there must be clear ownership. If we work on a POC only within the IT team, data science team, we do not have if we are the owner until the POC is deployed. But then who will be the owner? So that's why it's so important to engage the business leader from the same beginning so they can own the outcome of the model, but also validate whether the AI recommendation, AI solution makes sense. in the real operational operating environment. And also we should build that. Yes, yes. Anna Nesterova: And also financial part, right? The f financial part too, because it has to make financial sense because what I hear a lot when companies try to implement AI, they come to conclusion that some some AI solutions are too expensive from what they actually thought initially. Is that part of what you do as well? The way out the financial part and what makes sense? Karolina Henzel: Yes, exactly. So with this program we have a our cost plan, but also the benefit that the solution should bring. So I would say that also in this AI some implementation, we need to approach a bit differently to the let's say traditional IT projects and be prepared for the low royal cluster. And do many times the pivot because we validate many hypotheses. So what we created and what kind of roadmap, what kind of assumptions we had at the start, it may appear it may occur that this is after the validation of some hypothesis, we see that other areas has high potential and all the forces should be focused there. And I I would say that also the organization have it's challenging to narrow down the scope. Because there is Anna Nesterova: Mm-hmm. Karolina Henzel: a lot of ideas. But France speaking, if we say that we need to engage the business leaders, and as you can imagine, they have a lot of their place. The AI's project, AI program, it's not the only one that they focus on. They have the business day-to-day operations that they need to focus, right? So we cannot overload them. But if we really want to take care. Or care about the adoption, we need to carefully also choose the business use cases that we are going to focus. And they should also solve the business problem that these leaders have. And they are committed Anna Nesterova: Yes. Karolina Henzel: to achieve them because only then we will have naturally their engagement. Anna Nesterova: Yeah. I have a question in terms of the company itself. Do you think this AI transformation, as you said, a year ago you started this AI transformation for the company, for the operations? That desire for AI transformation, did it come from the market when all the competitors are trying to implement AI in their operations, or it was more of an internal, intrinsic desire to change something on inside of the company? Karolina Henzel: There is no one answer, right? It's Anna Nesterova: Okay. Karolina Henzel: the combination. I mean, we observe markets and how other leaders implement or change their operation, how they approach to the AA transformation, and like few examples of Johnson and Johnson and some other companies that consciously narrow down the scope. So having many pilots just to validate the hypothesis but then choose only few ones that they will focus on, knowing that the adoption is needed. So we approach similar way to to that, but it all always also comes from the interlock need. I mean we've we operate in a on a global market, so with all those Political changes, the how the tariffs and how we were hit as all the global operatives, not only in healthcare industry, by the prices increases, right? It also it also started the discussions how AI can help to optimize our network, optimize our cost, logistics, and other operations. So This is the combination of of both. So the this cost pressure from the from the market, from those go political changes, I would say that also helped in accelerating this journey. Anna Nesterova: Mm-hmm. I do think there is a lot of pressure, on the market in general, because it's a lot all the discussions I hear that's all about AI, AI, AI. How do we implement AI? But also there are some fears, like you said, some people are scared of AI. How do you deal internally with that fear that people have? What if the those efficiencies will affect people in the company? Karolina Henzel: it starts with the education and changing the culture and Anna Nesterova: Mm-hmm. Karolina Henzel: educating by the leaders. So that's the first step. But also to understand these fears, I mean, being in their shoes, just giving the example of the Power BI reports, as I mentioned, that there is a lot of effort and time spent on Finding relevant data, preparation for the reviews with the leadership, reconciliation between reapers. So I would say a lot of work done by the highly skilled people that it's not value added, because value-added tasks starts when you create the recommendation for your leaders, understand what are the root causes of some drops or negative trends, right? And what is important to to communicate to those people is that the time that they spend on these non-value added tasks they can shift into the where they skilled they the the the the skills are needed so recommendation working with the recommendation given by the agents validating the the hypothesis and and different scenarios so so this is like a shift of the the the work and not elimination of the of the work so there is a lot of of communication and sometimes i would say that people who are close to the ai to to to analytics to data naturally they are enthusiastic about the ai and do not have this fear but i would give you one example of BI dashboards when I deployed a solution that should support the sales team. and I thought that it will be so helpful with the recommendation, friends, everything you have in one place on trust, trustworthy data. But then I talked to the people and they feared to use it, they were afraid of using it, and not to destroy the view for their leaders. So they really were they were afraid that by applying some of the filters, they can destroy the report. So that's why it's so important to really understand the process, understand what are the fears of those people, how day-to-day work look like, and this kind of thing, because only then we can and we will know how to explain them or present the tools so they they naturally become the part of day day to day job. Anna Nesterova: Mm-hmm. And can you tell me a little bit more about the global scale? Because it's one thing to have AI pilot and AI product in a small company. So you test something, maybe you you considered five different products, you chose one, you deployed it, and hopefully it works, right? Like and it works fine. But it's a totally different game when you are dealing with a large company like GE, right? So you have so many teams across so many countries. How does it work in such a massive global company? If you're bringing something new. Karolina Henzel: It is it is challenging. It is challenging, that's that's for sure. and without this support from top level leadership, it will be difficult to lead such a transformation bottom up, I would say. Anna Nesterova: Mm-hmm. Karolina Henzel: So that's that's the that's that's the important thing because they communicate those changes and and the need of such an AI AI solution, but having in How we shape this program. So every each of the regions and segments commit to the to the value that they should provide from this solution. So we were like personally assigned people who committed sign off that they will deliver this value from this solution. And they were engaged from the same beginning on the on the pilot. pilot stage in building the solutions on a regular we have our operational mechanism to to communicate to gather the feedback and it's not easy but also with the support of this the change management office they helped us to prepare the adoption plan to if needed change and the ways of working sharing one of example example of the of manufacturing where we digitalize a dating management system and the so it was explained also and in those ways of working also explain how the AI recommendations should be treated by the employees that this is how it's actually how other employees would advise you or give you advice but you should filter that through your expertise and then make decisions. So this is not like an AI agent that makes decisions, but you are. You just getting the recommendation or you get quicker some insights. You can test some scenarios yeah and get to the trusted data and and insights so there are multiple elements to to to to to make it or scale such such a solutions but we should not let's say underestimate this adoption effort and the Anna Nesterova: Mm-hmm. Karolina Henzel: same or even much more time compared to how much time is spent or building creating building the AI model should be spent or designing with the business leaders, representants from the segments and region, what will be the adoption plan? Anna Nesterova: Mm-hmm. Karolina Henzel: And commit them to the value that should be achieved from from the solution. Anna Nesterova: Yeah. And you mentioned manufacturing is one of the operations parts where you see AI transformation happening the most. Where else do you see it? Karolina Henzel: we select if you such a logistic for the cost optimization, inventory, so safety stocks, demand, demand planning, procurement, and we Anna Nesterova: Mm-hmm. Karolina Henzel: will there there is contract digitalization opportunity and finding the rebates in the contract and many other clauses that might be helpful for for the procurement office. commercial. so pricing ai and i pricing so and also tariffs so we use also ai for a tariff optimization or better management of the of the tariffs so these are like a key pillars of of our of of the operation that we decided to transform by ai Anna Nesterova: Yeah. So you will be speaking at the conference tomorrow. So you're traveling to Barcelona tomorrow. I would like to ask you, what is the core message you're hoping to leave your o audience with? Karolina Henzel: Hmm. that the the core message is that AI, that scaling AI is not primarily primarily a model development challenge, that it is an organ organizational and and data transformation challenge. So there is a lot of excitement around generative AI and agent, and rightly so. but in a large organization's especially one operating in healthcare, the difficult part is not to create an impressive pilot and show the opportunity, but to integrate AI into the way the organization makes decisions and works every day. Anna Nesterova: Mm-hmm. So you will be talking about that, right? So I encourage everyone to Karolina Henzel: Yeah, yeah. Anna Nesterova: come and listen to your keynote and I'm sure they will learn a lot from you. Thank you so much, Carolina, for joining me today. Karolina Henzel: Thank you.