Dr. Marcell Vollmer is the CEO of Prospitalia Group, formerly Chief Innovation Officer at Celonis and Chief Digital Officer at SAP. He joins to discuss Machine Learning advances in MedTech and how practitioners can be thoughtful about when it is appropriate to deploy ML.
[00:00:00] MV: I start by introducing the term bionic. Why do I do that? Because technology and humans need to merge, and really, we as humans need to level up the technology.
[00:00:13] RS: Welcome to How AI happens, a podcast where experts explain their work at the cutting edge of artificial intelligence. You'll hear from AI researchers, data scientists, and machine learning engineers, as they get technical about the most exciting developments in their field and the challenges they're facing along the way. I'm your host, Rob Stevenson, and we're about to learn How AI Happens.
[00:00:42] RS: Joining me today on How AI Happens is the CEO of Prospitalia Group, Dr. Marcell Vollmer. Marcell, welcome to the podcast. How are you today?
[00:00:50] VM: Many thanks. Thanks for having me. Great being here today.
[00:00:54] RS: Really pleased to have you. We have loads to chat about. You've had such a storied background in the AI and machine learning space. You are currently the CEO of Prospitalia Group. I'll let you explain what that is in a moment. You were also a former partner at Boston Consulting Group and the Chief Digital Officer at SAP, loads of experience in the space and there's a million different directions we could take this conversation. So, I'm really excited to get into it with you. I guess first, would you mind sharing a little bit about your background and how you kind of wound up in your current role and how you approached that role?
[00:01:22] VM: Happy to do it. So, basically, I worked 15 years in the logistics supply chain industry at DHL and then I joined SAP. I stayed there for 14 years, helping to transform the back-office functions. This was my starting point, chat services, as well as also then getting very quickly into post-merger integration what I have done in my previous role to integrate business objects and Sybase. I will say integration lead for the analytics and mobile business when SAP decided to change from organic ghost to ghost by acquisitions, an exciting journey.
I was leading a kind of strategic in-house consulting organization, and helping also the co-CEOs for the new set up, 12 Strategic Initiatives at SAP. One was to become a cloud company powered by HANA. So, some might know SAP and has heard about it, which is basically the foundational CS for HANA system, what SAP runs nowadays, as an ERP. I continued my role becoming Chief Procurement Officer focusing also on the functional side and moved on by getting the opportunity also being part of the acquisition and laid down the integration of amoeba feed class and conquerors and market leaders for source to pay, when also SAP expanded in that area.
As a chief procurement officer, there is nothing what I could envision is better than basically getting access to the leading cloud solutions. When you transform your procurement function and really bring it on the next level, also on the IT systems for, what's around 6 billion spent what my team managed on the SAP side. I also integrated them Ariba class, moved to the United States, Atlanta, Georgia, and continued as Chief Operating Officer, really bringing the business together, and also helping more and more customers to define their digital transformation and work from the United States or basically all over the world.
Later on, as Chief Digital Officer, before I left and joined Celonis. Some might have heard about Celonis, a market leader for process analytics, process mining. It's an amazing technology. We might cover this a little bit later because it's using AI machine learning as well. So, last one and half years prior to my role as CEO, I was part of a Boston Consulting Group, also focusing on digital transformation, bringing a little bit reality into what artificial intelligence can do and can't do.
Now, Prospitalia Group is basically a kind of – let's call it a marketplace in the healthcare industry with cloud analytics supply chain solutions, as well as advisory what we have. It's a group of around 500 people, a little bit more than 100 million on the revenue side mainly focused on the German healthcare industry. If you want to say, and I explained it to my friends very often, it’s a kind of Amazon for hospitals, what we do.
[00:04:14] RS: Got it. Thanks for walking me through your background there. Where does the AI and machine learning piece come in with Prospitalia Group?
[00:04:21] MV: Basically, AI machine learning, it's a fascinating topic because everyone talks about it and everyone wants to talk about it, because this really brings it to the next level and people have very high expectations, what artificial intelligence can do. We are using it internally also for our solutions when it comes to the millions of products, what hospitals need from all over the world. Basically, you want to ensure that you have the best for your patients available at the fingertip. And that's basically what a marketplace functionality is a catalog.
We really look into all the data to ensure that we can streamline the data, that we get it into our systems, and that hospitals from all over Germany, as well as some as our countries can access it, and really work with that. That's one area, sounds pretty simple, I can tell you and some of you might know about the complexity what master data can bring into the business, that it is fairly complex and you need a lot of machine learning to do that. But we’re also using it, when we manage ORs, operation rooms, where we try also to look into the data in real time, whenever operation starts a little bit later. It might be only 10 minutes, something like that, and then the surgeon will come in and will basically perform an intervention with a patient.
Basically, when you see, it's something that’s already delayed, you can automatically predict, okay, what does it mean at the end for the patients later? And what basically also might happen that the last one on the list might, by no way, be able here to get his treatment today. That's something what you can only do. When you have access to the information and use some algorithms on that. These are only two examples of what we do. There's much more on how we integrate the systems and the supply chain sides. Basically, whatever is used him in a while, like an implant is automatically also then recorded into the warehouse in the inventory management. It is also ordered on the procurement side, but it gets auto reimbursed from the insurance companies. That's also where we use the technology for multiple aspects here and it's getting more and more basically what the teams are working with on a daily basis.
[00:06:36] RS: On this podcast anyway, the use of technology like you're describing is often assumed, right? And you know, the podcast is called how AI happens. So, that makes sense. But I was hoping you might indulge us revisiting the premise a little bit. In this case, why is machine learning the right approach for this use case?
[00:06:54] MV: So, I think first of all, and let's bring a little bit reality in that. Sorry, I'm a German, as you can hear from my strong accent. I'm always very straightforward and very open. Also, you have seen, I use AI machine learning also like, hey, that's the most natural thing to do and just use it. But let's be realistic, it is not so much really what the intelligence side is already what is available. To be fair, it is algorithms. It is great stuff, what you can do with that. So, predictions what I have mentioned in an OR, it's amazing help also for patients knowing, “Hey, you might get delayed, or basically there's something we need to move you to the next step, great things to do.” But basically, it is only an algorithm, what you use here and how you basically start also by leveraging now the artificial intelligence, which is a broader term, machine learning is a subset of that, and deep learning is a subset of that. As part of the data science, that's how I describe it always.
But face the facts, it is algorithms what you're using here, and a lot is not so much intelligence or artificial intelligence. It is basically also linear progression, what you do, and some very basic statistics of what you use, and you try to automate and also combine it with external data, which is a good thing. Well, I think it's fair to say, we at the beginning, I know that and when I was at Boston Consulting Group, so 58 billion globally investment market, what the total was that last year. So, you'll see, there's a lot of investment going into that area. But basically, there's also a lot what we still have to do before we come to the autonomous robot machines, what we're using self-driving cars as another example. We see the challenges what we have with drones, with other technology right now. So therefore, I always like to be very realistic on that and also manage expectations, what it can do and what it can't do when you're talking about artificial intelligence as a broader term.
[00:08:52] RS: When you say we're the early stages, do you view current applications of this tech as more Spartan, I suppose?
[00:09:01] MV: You can say that. I like most things that basically – and I start by introducing the term bionic, why do I do that? Because technology and humans need to merge. And really, we as humans need to level up through technology. That's how I would describe it at the beginning. Bionic might not be a good term, at least, I like it also by showing the process on how the machines are working for us, and how we can benefit from collaborating and integrating machines in our work. And then the learning side, deep learning, machine learning, artificial intelligence is basically then the next evolutionary step. How you use it and how you basically can also advance the algorithms and technology you have access to.
[00:09:44] RS: So, when you bring bionic in, do you mean more like human machine interface, kind of like the integration that's at play there?
[00:09:50] MV: This is one element how you can define bionic. I like more talking about and I use the term bionic company in a way that the technology, and the technology, it does not mean that we are only talking about artificial intelligence. It can also be different, emerging or disruptive technologies, what we have access to. You can use IoT, you can use 3d printing as part of that. It is more how you leverage the technology and basically, that humans use it, and basically, how you combine it from the process side of technology with seeing humans to combine it that the technology is serving the humans on whatever the task, it whatever the process is, they need to perform.
[00:10:34] RS: I appreciate that you spend a lot of time thinking about and stressing what AI can and can't do. Could you maybe give an example of how that conversation plays out at Prospitalia Group?
[00:10:44] MV: A good example on that is when we look into the data, and also try to predict, for example, if you want to suspend behavior, what you have when we use benchmarks when we compare, for example, the vast or the big set of data we have access to, to really look into that, what can you learn from the data? What is possible, really here, to understand from these complaints, for example? It might be the most efficient and effectively helping patients. So, that's something where we really see also the benefit, when we combine also just the data, and we are talking about procurement data. But we can link these data to all what a patient needs, which is much more than the implant, it's the surgery, it's also the treatment of the care afterwards.
When you combine this data, and when you really see it, you can derive meaningful analysis from that. But I believe, in the future, coming also to these products for this group of patients might be better than others, we are not at this level to be very fair. But basically, here we see already, at the beginning how we can use and leverage the available information and the data we have to really predict also, this is probably what a hospital needs. This is what patients in the future need. We can link this also to market information on the supply chain side, where, and we see right now due to the war in Ukraine, due to COVID, Shanghai currently is closed as we are having this interview, Shanghai, one of the largest in the world, and we can predict by that also, what is the impact? Which chips are impacted, and what are the products impacted by that? This is really where you cease to benefit from that overseeing and getting full transparency, monitoring your supply chain, and this is also what we use, to a certain extent also here for what we do at Prospitalia, as one player in the healthcare industry, driving the digital transformation and also using the available information to serve our clients best the clinics, to help them understand early.
This is what currently, basically a bottleneck might be. This is how you can mitigate it and this is basically the data we have that you can also compare use it for your analysis to really make an educated decision about a medical doctor here, is that you, for example can go for another product here to help your patient to cure as fast as possible.
[00:13:11] RS: When the practitioner, let's say, the machine learning engineer, for example, is considering where to point this technology. What are the questions you think they should ask themselves to determine whether it's the best approach?
[00:13:23] MV: Big question. I always love to start with a use case. That's what I learned from whenever we talked about technology, and it doesn't matter with SAP, BCG, and here now at Prospitalia, I always like to ask the question about what exactly is the use case? You can do so many things and as access to great technology, what you can use, there's also tons of data what is available. But at the end, what is the purpose? How can you really have a business benefit from that? You can play around, fine do that, and probably also you learn something, you generate interest from young people using technology. But at the end, I always like speaking to the tech guys in my group about, “Hey, think about the persona. Think about the use case we want to have, and then let's think about how can we really get the best out of what we have access to, and let's identify what is it? What could basically help us to become better in the future?” That's how I like to start a conversation and I speak with a lot of tech guys. We have startups as part of our groups, where you have tons of great ideas and innovations. But you need also to balance a little bit and people sometimes really need to focus on how to serve best as a customer at the end. So, think about a persona and a use case to where to adopt this, would be my recommendation.
[00:14:48] RS: How important is that design thinking approach for the individual practitioner because as you say, to put themselves in the shoes of the persona of the individual using the product, it still has to be easy, right? It has to be like somewhat straightforward for folks, or they want to use it. How do you kind of tow that line between it needing to be easy, but also appreciating that it's fantastically complicated, the computations that are happening?
[00:15:11] MV: I love design thinking. It’s a great way and how you bring values and development teams together. I know it's developed in Stanford as well as Hasselblad, rewarded, for example, to SAP and I was part of lots of design thinking groups working on that. Basically, it's a great way really where you focus on the persona and also solve for problem. And also, what I love to do is a prototyping. Whenever you do a prototyping in software, it's a mock up what you do. Basically, it helps you a lot by making it tangible. We are human beings. We talk a lot about abstract things like artificial intelligence, robots, and all this stuff — what we do. But at the end, seeing a prototype, and bringing a group of smart people together to solve one very specific defined problem, this is definitely a great way to do and to work, and then use a drive to really focus on small teams with also high autonomy, to take decisions to get something done. And then if you fail, fail early. If you succeed, really try to leverage and expand and share the learnings with the others.
This is how I use design thinking and what I have learned what works very well, in both ways in things which are really go into the next level. But also, when you figure out, we are not at this point in time, really, at a level where we can use it already. So, let's move away, let's do something new and find an alternative solution. Fantastic using design thinking for that.
[00:16:43] RS: I'd love to go back to what you said earlier about how we are still in the early days of this technology. I think that's clear to anyone who really indulges their inner sci-fi nerd about what this technology could look like in the near future. When you think of the medium term, I guess, the medium horizon, what is the future of med tech look like? And what are the applications you see AI and ML disrupting? What does that look like for the industry? And what does it look like for Prospitalia Group?
[00:17:08] MV: I think first of all, allow me to focus on the patient first. Med tech, health tech will provide tremendous opportunities for patients to get whatever they need, and basically also be seen, the average life expectation goes and that's definitely a good thing. When you look at the global numbers, platforms are built. I think that's number one patients benefit from that. Number two is when you look on the supplier side, it is also amazing, seeing the investments of technology what is happening right now. Bionic, in that case, you can use the term how you started with the interpretation, which is perfectly right, and I just use it on the more economical side. When you say now about that, we have an exoskeleton, for example, that says people who are no longer able to walk, can walk again, or basically that you get help for machines for elderly people staying at home. And basically, having a robot supporting certain activities, having machines helping, that's amazing and that's all what med tech does.
When I see the wheelchairs, automatically going stairs up and down and safely transporting passengers, it's amazing. Getting indoor navigation, helping you going a little bit – how can help people. So, there's a huge potential and also where we see a lot of innovation happening right now. I only mentioned 58 billion on 2021 on AI. When you think about a man tech, health tech investment, it is a much larger amount, which definitely will help also that we as people can benefit from that, and also that patients basically get whatever they can do.
By the way, looking also in other ways and use artificial intelligence here, I didn't use it at the very beginning and we at Prospitalia are not exactly working in that space doing it. But we know that machines are excellent for certain defined task like identifying breast cancer, just as one example. We know the precision of what a machine has, is uncompleted to what a human can do, which is really great thing also, by seeing here the benefits of using the technology and med tech, health tech, it is definitely going industry.
Just in Germany here, the health care spent in hospitals increase from 60 billion to 100 billion in the last 20 years. So, it almost doubled. Here, you see also what is needed and also the investment, it is not only a price increase. We see now inflation, we have a problem to deal with that. But on the other side, we see also that basically there's a lot of technology available right now which costs money, no doubt about that. But basically, it helps people and just really what counts. At Prospitalia, we try to help really here, to get access to the latest available technology as fast as possible, and also integrating in rotating pulses that every hospital, every patient gets, what he or she needs when they need it or find alternative product in case something is not available, like we see right now, when we have a supply chain bottleneck of what's happening.
[00:20:07] RS: Fantastic. Thank you for looking around that corner a little bit for us. Last question for you, Marcell. For the folks out there in podcast land listening, forging their careers in this space, what advice would you give to make sure that they can continue contributing at a high level and making sure that what they're working on is truly disruptive?
[00:20:23] MV: Try few innovations and do something you believe on, test and play around, do the coding, and I hope that people learn much more in this regard. Because the future is bright, and there's tons of opportunities. I really hope that you'll find your passion and I think there is no better place I could imagine to innovate than really a place also where you can help people to develop innovations, where basically all the people in the world can benefit from. It's an amazing opportunity and I hope lots of young people are getting motivated, starts cording, and think about what I said about, think about persona, and the use case you want to do, and I hope you will be successful. If you're an entrepreneur, follow your vision and go for the funding for your startup. I hope you will be successful and I hope a lot of innovations with have all of us, and that we all benefit from it.
[00:21:15] RS: Dr. Marcell Vollmer, this has been fantastic. Thank you so much for joining me today.
[00:21:19] MV: Many thanks for having me. It's a pleasure.
[00:21:30] RS: How AI Happens is brought to you by Sama. Sama provides accurate data for ambitious AI, specializing in image, video, and sensor data annotation and validation for machine learning algorithms in industries such as transportation, retail, ecommerce, media, medtech, robotics, and agriculture. For more information, head to sama.com.