Shaping the Future of Health with Boehringer Ingelheim [Integreat 2025]

Transcript:

I’d like to introduce our next speaker, Ralf Schundelmeier from Boehringer Ingelheim.

So I love working with this team. For those of you that are not familiar with Boehringer Ingelheim, it’s the world’s largest private pharmaceutical company and Ralf will tell us more about the company but I’m also deeply grateful whenever I have the opportunity to work with this team because they saved my father in law’s life.

Boehringer Ingelheim is one of the foremost innovators on stroke prevention and management pharmaceuticals. So my father-in-law had an ischematic stroke and he was happily saved in the emergency room and so I’m grateful to researchers that are in this space.

I also have a Doberman who’s a very high need dog and I’m also using Boehringer products to keep my dog healthy. So, Ralf will tell you more about them but for me personally I’m just grateful for the work this team does.

Please welcome Ralf up to the stage.

Good afternoon.

I was asked this morning, what song I would like to play, and I couldn’t think of anything better, because it sometimes is also the motto in IT. You always have to look at the bright side of life because things don’t always work so well as we demo them here on stage.

So good. What are we talking about today? It’s really more the journey that we have taken.

So it will be a little less tech and more what are we trying to do, why are we doing it, and then some of the lessons that we have learned. So maybe things for you to watch out for.

But one key here is really self-service. Yeah.

And that’s really one of the key drivers for us. Unlocking data and empowering people to do it themselves.

Yeah because we cannot scale our organization. So, Jeremiah already did a great sales pitch for the company.

He actually, there’s not that much for me to add. But as you mentioned, we are family owned pharmaceutical company founded in 1885 in a small town called Ingelheim in Germany.

Focus is on human pharma and animal health. I think that’s also a little bit of an exception because most of the larger life science companies kind of divested their animal health business.

We’re about 55,000 employees worldwide. We are a research company.

I think this is also super important and that’s why you see how much money we spend on research about 6,000,000,000 every year. We have 26 R&D sites and in revenue we’re at about 27,000,000,000 per year.

So it’s as he mentioned probably one of the largest companies that you’ve never heard of. Good.

What do we do? I mean, Jeremiah already described it. It’s a wide range from cancer to respiratory diseases.

There was just a product launch just recently. It was the first product launch for Pulmonary Fibrosis in the last ten years.

Really a big big launch event for us and then of course animal health. And as you pointed out it’s for pets, not just dogs but also cats.

One of the recent products was a diabetes treatment for cats that did not involve an injection. For those of you that have cats, try to inject your cat.

Good luck. But also livestock, so that’s also a core part of the business.

So then I had a challenge. How do I transition from this company transition to what we are facing? And I have a strong affinity to the Japanese culture.

Actually my wife is Japanese. And I was thinking of this beautiful print called the Great Wave of Kanagawa by an artist called Hokusai.

It’s this giant wave and I thought it’s a perfect perfect description maybe what what we are looking at because this AI wave is really changing dramatically how we work, how our jobs look like, whether companies will be successful or not. So I also like numbers and we’ve seen a lot of numbers today.

And one of the jokes is don’t trust any statistics or numbers that you haven’t falsified yourself. But there’s a lot of good data out there, right? This one is coming from McKinsey and they actually did a study primarily for life science companies.

And I’m only calling out to three things actually. First is we talked already a lot about these agentic workflows and McKinsey predicted about 75 to 85% of all workflows in our daily life can either be augmented or fully automated by Adjendic AI.

So this is a dramatic change how people work. Then of course cost efficiency, about 6% to 8% just from increasing productivity and optimizing operations.

And then of course this also has an impact on your bottom line. So in a time horizon of three to five years they predict 3.

4% to 5. 4 increase in your earnings.

So really a lot to deal with. So I don’t think there’s a safe harbor where we can safely wait for this to pass.

It’s really facing this wave and best get a big surfboard and ride it out. You can tell I lived in California for some time.

So how do we get ready? And I mean there are many many aspects and we only have very little time here today. But today I really focus about one thing that we always forget.

AI needs data. Without data there is no AI.

Yeah and also keep in mind there is not just generative AI, there’s also still a lot of traditional machine learning out there that needs data. Also a lot of the use cases that we talk about talk about talk to your data, right.

You need good data that you can talk to. You need to get your data AI ready.

And we believe that self-service integration and self-service data is really our key to success to make this happen. And that’s what keeps my teams busy.

So we are a very old company. We’ve been collecting data for a long long time and a lot of this data believe it or not might still be on paper or it’s on some legacy systems.

Yeah. They’re very hard to get to.

Most of them don’t have AIs, APIs. Data is not cataloged.

So really how can we unlock this data treasure so that we can use it? And this is how we came up with what we call Dataland or Dataland Vision. So the idea was really to create an enterprise wide data vault where all what we call crown jewels, all the data that we believe has a lot of value will be brought into this ecosystem to make it easily consumable by either analytic use cases or AI or Gen AI use cases.

And if you look closely, I thought, we talk about AI. Why don’t I generate some of my flights? With AI, it’s also a testament that AI is not quite there yet.

Sometimes I feel it behaves like a child in puberty. I actually have two of them because they never listen to what they say and I say it five times and it’s still not changing the color or it’s not correcting the wording.

So if you look closely in the slides, you might might find some little typos. So why integration is important for this? I found this slide somewhere honestly I don’t remember exactly where but this is quite frankly still how a lot of our integration scenarios look like.

Right? Tons of point to point integrations. We also have a very wide variety of tools.

That’s actually how we started talking with SnapLogic because we have a lot of modernization projects going on. Yeah, so we really felt do we need a new approach and this is how we came up and I’m not boring you with a lot of architecture but just a simplified view how we think about managing this data ecosystem.

So data sources cannot only be within your company. I think that’s also super important especially in life science.

We buy a lot of data to really augment our research, to augment market data. So that’s also one key aspect And then as I mentioned we have all these systems that have this data and our idea is really to leverage SnapLogic not just for building the data injections that we build but also really to provide it as a tool for self-service.

Yeah, because we cannot scale. The demand that we have is too high.

We also did more like traditional IT projects and it takes too long especially in research. It’s not weeks.

They want data now. They run the experiments and then they determine is this a good result or bad result and I move on.

So a lot of our processes also change, right. We are heavily regulated so a lot of our processes are quite heavy.

Yeah, but this is a balance that we have to make. How we can empower people to do it themselves quickly.

So then the data is ingested into our storage and processing area initially primarily on AWS. We use S3 and Apache Iceberg to maintain primarily the originally raw cleaned and curated but then more and more for analytic use cases we also moved more stuff into relational databases.

So that’s the kind of the technical aspect and you see there’s a lot more but I don’t want to go necessarily into detail. But there’s also a second aspect that we call data as a product.

And I think when we started this was really absolutely valid because our first use cases were primarily analytic use cases, right. So you really have more, call it longer term view on the data.

So it really makes a lot of sense to have very well defined data products. We have data domains.

Every domain has a data domain owner. Now every data domain they have their own data product teams.

So this is definitely still valid for analytic use cases and it also helps us tremendously with data governance because it’s super important to control the use of your data, to be able to track the lineage, to also track who used the data for which purposes. For some of the AI use cases we are currently rethinking this a little bit to be more dynamic because creating a data product typically takes time.

We also talked a little bit about who and where do we all use AI and we find it across the board. So these are just some few examples when for example in production it’s about predictive modeling to determine whether there are any issues or deviations in the production process before there is actually an impact.

Or when you think about clinical trials, it’s about visual dashboards, then it’s about analytic models for rebates and pricing and then all the way to the right and this really is becoming more and more important. How do we use AI to find new molecules, to find new medications for for patients and really speed that up.

So for us, as I mentioned, we did traditional projects and sometimes we were super fast and sometimes it took us nine months to create the data product really end to end from getting access to the source system, doing all the end to end work. That might still be valid for certain use cases but we really believe if we make self-service integration available it will really enable us to scale but also to accelerate innovation.

We talk about citizen developers but we also talk about citizen researchers, right. So that we can provide all the data that they need to validate the hypothesis fast, test it fast, and come to decisions quickly.

It also helps us to break down silos. Yeah, we have manufacturing data, we have sales data, we have research data.

How can I sometimes create meaningful insights across of all these data silos? So really using this ecosystem to bring it all together and again without necessarily requiring help of IT. Scaling and AI adoption.

It’s key for AI to have the data and if I make it easier to bring the data into the system, I immediately get adoption. Yeah, very often we have the data scientists ready.

They want to act but they cannot because they don’t have access to the data. So here really using self-service integration to scale this AI adoption is key.

We also talked a lot about compliance and trust and especially in the EU with all the AI regulations. Yeah, I mean we cannot just do AI anymore.

We have to actually have an audit trail. What data have I used? Where did it come from? Yeah, so all of these things and then of course also other regulations like HIPAA and so on and so forth.

So but we’re not doing this just for the sake of technology. Yes of course we’re doing it to to ride the wave, but there’s also some some key business value in the end.

Quite frankly it makes my life a bit easier because we all have to ask for budget. Right? And suddenly we move from, apologies if I say it, but plumbing.

Yeah. Because we’re sometimes viewed as plumbing when we build integrations and as long as the plumbing works, everything is fine.

But now we are really transitioning to be strategic because of the lot of the things that we talked about today. They are not possible without a good integration approach.

Yeah that involves API’s as well, API management of course. So but it’s really for us also a game changer.

So we talked about accelerated decision making, right. Sometimes hard to put a direct money value to that, but when you think of a research project, the sooner I can determine whether this is the right path or not, I might be able to save a lot of money because I don’t have to ride all the way to the end to determine that it’s a poor outcome.

Agility, security and compliance, right. Composable enterprise architecture really helps us to be more agile providing enterprise security.

It’s super important especially when we talk about agentic AI. A lot of these agents can sometimes access data they probably shouldn’t access because they don’t impersonate you.

They might use a generic integration user that has rights to all the data. So really thinking about this is key but here this architecture really helps.

Cross function collaboration. This is also super important for us because it really changes our role.

We transition from being gatekeeper to strategic enablers. Yeah and this is I would say also for our team super rewarding because it’s a very different role.

We are not just being handed orders to execute, it’s really now we are strategic partners and strategic enablers. Cost, who doesn’t like it when you can list cost optimization? Yeah, I mean we don’t have exact numbers yet for us, but it’s already quite obvious that through reduced manual efforts or especially reduced duplicate efforts or through automation that significant cost savings.

And as we mentioned before innovation enablement, we can just move so much faster and to me this is a big big outcome. I typically say we need to get out of the way of the business.

Yeah. So they don’t perceive us as a blocker that you know we hold up our hand wait wait wait we have an SOP here or we have these processes.

And and gain more speed. This one I thought a lot a lot about like what what are all the the challenges.

And quite a few of them are not related to technology. The first is governance versus autonomy.

I don’t know how many of you have dealt with data scientists. They are really used to do whatever it takes.

Using those words and whatever they want. Yep, so they’re really used to load data on the laptops, run their models on the laptop and now there we come and say no no no no now we have this beautiful data land vision, we have to be compliant, we have these governance processes.

So it’s really a very careful balance and you have to educate a lot. And sometimes you also have to change your approach.

Like we have to for example change our definition of the data product to enable faster use of data. Because the next one speed versus perfection, right? You can spend months or years to define the perfect data product, but is that always the right way or do you just want to say let’s get the data in.

It could be almost raw data, especially in data science raw data is good enough, right? So be more flexible. Data quality and trust.

Yeah, yes it’s self-service, but we need to make sure that they’re using clean and validated data and consistent data. Yeah.

Which is then also a challenge when you allow self-service data ingestion or self-service file uploads. Right.

So here we also have to be very careful. For me, a no brainer, but sometimes it’s not source system access.

Sometimes it takes us a very long time to get actually physical access to the source system because the system lead for whatever reason takes forever to get the approval to have access. Yeah.

Security and compliance, we mentioned that already a little bit. Compliance is key especially in a heavily regulated industry like pharma.

We really need to make sure we are playing by the rules. GDPR, HIPAA and all the other regulations while enabling access to the data.

And then we talk a lot about citizen development but it’s it’s actually a challenge to enable users and making sure that they are upskilled to do it themselves. Key takeaways.

The first one actually right follow-up for the citizen development. Don’t just think about technology, yeah.

To have a data driven target operating model, yes sure. You need to reorganize, you need to refocus, you need strong technology, but also think of your people.

That you need to enable the organization and enable the people to do it themselves. Yeah, because self-service means that the person who is doing it, they need to be able to do it.

Next one is also interesting. Technology does not make bad processes better.

Yeah. Agentic frameworks will not fix your bad processes.

Yes, sometimes you have to re engineer your processes. You have to change your rules in order to be successful.

And then of course we talk a lot about self-service. Also an API first strategy.

You know something that I took for granted for 20 and now suddenly you have to convince people again that hey maybe you should start with an API. Very new to me.

I also had to relearn and enable people more by explaining why API first is so important and especially in a Genic Ai. When MCP actually came up there were some articles all of the death of APIs, you don’t need APIs anymore and we had a very heavy debate in our company and said no it’s actually the opposite.

You need more APIs in order to make MCP successful. The last three are actually more things that I didn’t talk about today but for you to keep in mind.

Right? I think Jeremy already alluded a little bit about event driven integration. It’s not new.

It’s actually has been around for quite some time but for us when we look at moving data, yeah, I don’t need a full refresh of my data every week, every day or even incremental. Maybe I should use events or event streaming approaches like change data capture to have always the freshest data especially also in in agentic environments.

It’s key that your data is always fresh. So event driven architectures might become more relevant here when it comes to data freshness.

Then of course MCP and Agenetic AI. We’re actually actively working on all of these things but there was just not enough time in this little slot.

And then also really leverage AI whenever you can. In all aspects API creation, with the snap GBT, when it comes to data quality, when it comes to data transformations, yeah.

It really helps the people to move faster. Always look on the bright side of life.

I hope it will not be replacing us. I hope it will give us the tools to move fast and have more time to focus on the important things.

And then I have lost completely track of time because the timer went the opposite way but I think we’re good and have some time for questions.

Thank you very much. That was fascinating and really interesting to see the broad span of the work that you’re all doing.

So you having taken in the broad span from you know system modernization to the data view application processes etcetera. Think of your hard questions.

You have a captive expert in the midst of one of these projects but I’ll kind of kick it off maybe. First, I think one of the things we talk about with legacy modernization is how to build the business case to do it in the first place.

And I think by my last count, I think your team is in the process of I think at least five legacy middleware solutions migrating whole or in part modernizing those systems. And it was really from my understanding driven by the business case and the business benefit of where you’re going and not necessarily the end of life of any of those systems.

But can you talk through a little bit how the Boehringer Ingelheim team looks at modernization because I know you’re very financially responsible as well in terms of how the company is run. How did you approach that challenge?

It was actually a dual approach. I’m not sure what was the chicken and what was the egg.

First there was a need to modernize because we also, I mean we looked at some of the timelines here. We have some very old stuff and that runs out of support.

So here it’s a very easy risk-based calculation where grudgingly people are willing to take money in the hand. But if you then come on top of that and that’s really where I would say this JetGPT wave really helped, because like I said suddenly it becomes strategic and you become a strategic enabler with your technology to a lot of these use cases.

I think we still face some challenges not necessarily in selecting a tool and wanting to modernize, but sometimes the business you know they are moving a little slower or they don’t feel that there is a need to to modernize. So again end of life helps and also strategic impact really helps in that matter.

Excellent. So kind of a dual approach to make it clear this is not merely a business value neutral exercise.

There really is value on the other side of of the exercise. Of course, the business users have to participate in user acceptance testing and have to support the the project as well.

Well, I’m a box of questions. I can monopolize the time but, do we have any questions, you know, from attendees? Just go ahead and raise your hand.

We’ve got mic runners. Looks like we’ve got one over here.

I know we none of us wants to talk about it, but essentially AI sooner or later takes jobs away. And it is more relevant to the organization like yourselves, which is quite old or old school maybe from inside as well.

How do you guys deal with that? How do you sugarcoat that message?

I don’t believe in sugarcoating. I think it will be a challenge for all of us, these difficult conversations.

I actually believe for a company like Boehringer it’s probably more challenging because it’s been so old. Throughout all its history there was only growth.

Yeah. And now these are some difficult challenges.

Right now the key focus is really on upskilling people. Yeah we talked earlier about AI training for example.

We actually started what we call the DataX Academy that we really use to educate our people in terms of data analytics and AI related topics. I think that’s key.

Like I said always look on the bright side of life. I also hope that there will be a balance, but I mean we already see at least agentic AI is used as the reason why people are being laid off.

I don’t know I always like to talk to people on the support line, if they know what they are doing rather than talking to a chatbot. But there will be an impact.

I think we also as a society have to face it. It’s challenging and for me it has to be an open conversation.

We have to think about it. We also need to think about reskilling, so maybe there will be other work for people.

I mean just in engineering, it’s also you know we talk a lot about byte coding. Will it replace software engineers? A friend of mine actually they did an internal project where five senior engineers really focused on using byte coding.

And yes you can reduce the effort. But you know there’s also a shortage of skilled people, so hopefully we we might see a balance there.

I’m still optimistic. I think from what we’re seeing across the SnapLogic community, the work that is currently being done to implement these systems is typically focused on domain areas where they’re already resource constrained. So at least right now we’re not seeing large scale, let’s say complete automation of a domain area.

Quite the contrary, we’re seeing the prime the lead use cases are typically somebody who comes forward and says, hey, I should not be doing this thing and I’m doing this thing all the time. And typically, it is a high context process that, you know, involves both unstructured or semi structured and structured data that requires a human to be that bridge between those two worlds and then execute the process.

And so, like, most of these cases we’re seeing are coming from people saying, I’m the domain expert for this area. I spend an undue amount of my time doing this thing.

Please automate this thing so I can work on other things I wanna work out. And I think that’s the other thing we’re not necessarily appreciating is the impact and gain in the positive space.

And even in the areas where we’re seeing this productivity gain, you mentioned support for example and support deflections. Most of the companies that I’ve worked with that are seeing a lot of impact in the support deflection, one of the members of our CIO advisory council said they saw an 85% reduction or deflection automation of HR service requests.

They’re hiring additional HR business partners to get closer to the business because they’re showing a higher strategic impact of working a different way. So I think this is going to be a very fuzzy picture and I think the statistics we see, talked with one analyst at dinner last night who said, you know, what we’re finding is that, there are a lot of layoffs being blamed on AI, but none of the layoffs were due to AI.

So, this is a murky space right now. I don’t think we should sugarcoat anything, but I think we should have truth over, you know, ego and perception and actually find out what’s actually happening in a bit of rigor, around that.

Other questions? Up front.

Ralf you mentioned you are for R and D company mainly right. So generally I’ve worked with lot of pharma company and what I’ve seen is like it takes long from discovery to market.

Do you see is going to be positive impact going forward? Absolutely. Or have you seen any sorry so far?

I would say we see the first results. Yeah.

I mean AI or not but for me COVID was a big game changer, right? How quickly suddenly if everybody wants to can provide vaccination for people. Yeah, took like what six to twelve months not ten years.

I think when you also look at some of the developments in The US, right, increased price pressure, etc., so you might have to augment revenue loss in some of your key drugs with new products. So you need to be able to develop faster and more cost efficient, right.

I mean today it’s still a lot automated handmade I would say, right. In labs we see a lot of automation but now it might be more and it’s not just in the research it’s also in the regulatory processes.

You know if you file with the FDA just to create all this documentation that you need to submit for approval, you move from handmade to generated through AI. So you can really speed up, the processes.

But when you also think how many trial and error, how many molecules you have to look at in order to get one final product, yeah, I think there’s a tremendous potential, but I think we are scratching the surface at this point. And just to add related to the earlier question like so it will be kind of a compensate like okay you can do more discovery because you have automation in hand. So you don’t have that many job loss in that aspect as well.

So you can utilize resources in doing more with the same resources and because you can do it faster.

Okay. We can have one or two.

I’m definitely running over because this is my favorite part of the session. Do we have any other questions in the room? Any hands? Oh, man.

I could I’ll just have to monopolize your time later by myself.

Alright. Well If you buy a drink.

Deal. Well, thank you so much.

Really appreciate you sharing your time with us and sharing this work. It is really an honor.