Rise of AI Conference 2025 · Berlin · Applied AI Stage

Beyond the Hype: A Practical Blueprint for Production AI

Christian Winterfeldt · Dell Technologies

At Rise of AI 2025 in Berlin, Christian Winterfeldt, CTO & Director AI Solutions at Dell Technologies, argues that organizations fail with AI not because they lack talent or tools but because they lack a common strategy. He divides the AI market into three phases: a 30-year-old pre-generative-AI market of roughly 2,500 highly specified use cases, the recent large language model market that made organizations more efficient, and an emerging enterprise market built on agentic inferencing that changes business outcomes rather than just optimizing tasks. He illustrates this with a procurement manager, Anna, whose agents monitor markets, prices and contracts overnight, and a product manager, Lisa, whose work depends on six core AI capabilities including bringing corporate context to LLMs and managing trusted data. Winterfeldt presents a "strategy house" model, developed and applied inside Dell itself, that moves from vision and use-case selection to a scalable foundation of culture, data, technology and ecosystem, and finally to execution. He closes with Dell's own on-premise AI platform results and a checklist for building a production-grade AI environment, insisting organizations start with a business goal rather than with technology.

Key takeaways

  • Winterfeldt frames the emerging enterprise AI market around agentic inferencing, where multiple agents act on an organization's behalf before a manager starts the day.
  • A standard LLM knows only about 1% of an organization's internal knowledge, so context must be added through RAG or fine-tuning.
  • Dell's internal sales chatbot reached a 96% usage rate among staff after being tuned with reinforcement learning on real answers.
  • Running AI inferencing on-premise rather than in the cloud can cut costs by 75% at scale, with an ROI within 12 months for Dell's own platform.
  • About 80% of enterprise data sits on-premise, so Winterfeldt argues AI workloads should be brought to where the data already is.

Detailed notes

Three AI markets and the shift to agentic inferencing

Winterfeldt separates AI into three markets: a pre-generative-AI market from 30 years ago covering about 2,500 highly specified use cases, each needing its own architecture; the large language model market built over the last two years with tens to hundreds of millions in investment, producing versatile LLMs used for chatbots, writing and image generation; and a new enterprise market focused on business outcomes rather than optimization. He introduces agentic inferencing through the example of Anna, a procurement manager whose agents continuously track the market, monitor price fluctuations, review vendor contracts and generate negotiation recommendations before she starts her day, arguing this changes whole organizational processes, not just individual efficiency.

Six capabilities organizations need to master

Using product manager Lisa as an example, Winterfeldt lists capabilities organizations must build: bringing corporate context to LLMs (since a standard LLM knows only about 1% of an organization's knowledge) via RAG or fine-tuning; deciding what data to trust and which agents or models may access it; building simple applications such as chatbots without needing to be a programmer; using analytics agents to monitor product performance against competitors; and optimizing physical infrastructure such as production lines and assembly assets. He notes that many organizations run successful pilots that never reach production, and describes a manufacturing company where IT had no knowledge of AI initiatives running in other departments, calling this misalignment the common failure pattern.

The strategy house model

Winterfeldt presents a "strategy house" as a mirror rather than a fixed model: start with vision and strategy (what makes the organization special, which processes can improve with AI), then select use cases, build a scalable foundation covering culture, data, technology and ecosystem, and finally execute with board-level backing. At Dell, this combined a top-down question (what differentiates the organization) with a bottom-up employee survey that produced over a thousand candidate use cases, narrowed to about ten. Dell identified four strategic areas for AI investment: go-to-market, services, R&D and supply chain, and built a sales chatbot now used more than four times a day by over 96% of staff after refining it with reinforcement learning on the underlying data.

Culture, data and ecosystem as the hard part

Winterfeldt describes culture as harder than technology: training managers through encouragement and education, and starting with the roughly 10-20% of employees eager to try new tools before bringing along the rest. He stresses that data is the core problem, since the same product spec can be interpreted differently by R&D, marketing and customer support, requiring reliable "data products" and pipelines. He states that no single company can cover every layer alone, so an ecosystem of partners, including Applied AI in Germany and Hugging Face, is necessary.

Building a production-grade AI platform

Winterfeldt describes the technical stack: running inferencing on-premise on physical servers can save 75% of costs compared with cloud at scale, with an ROI within 12 months on Dell's own AI platform investment. He notes 80% of enterprise data sits on-premise and that AI should be brought to where the data is, including the edge. He outlines a system built with Nvidia, the "AI factory," combining data sources, a data platform with data contracts and vector databases for RAG, a model hub with lifecycle management and governance, and IT-operated infrastructure, which Dell's engineers can stand up as a demo in 30 minutes.

Q&A highlights

Asked how to overcome cultural resistance in traditional Mittelstand companies, Winterfeldt says to start with the roughly 10-20% of employees eager to experiment, show results, and bring the rest along through education rather than mandates. Asked how to demonstrate ROI to convince a manager, he describes setting KPIs and sub-KPIs, citing a customer-service chatbot measured on resolution time and customer feedback quality, and recurring employee surveys tracking time saved. Responding to a comment that starting with a business pain point may be more effective than starting with a business goal, he agrees in principle but notes that organizations investing during times of crisis tend to emerge from it in stronger shape than those that hold back.

Quotes

This is not a model. This is kind of a mirror. So take a picture and really compare this what you're doing at home.

Christian Winterfeldt

Don't start with technology. Start with a business goal worth scaling.

Christian Winterfeldt

Organizations who are investing in times of crisis are going out of the crisis exponentially better than other organizations who don't, who just keep their money together.

Christian Winterfeldt

Questions answered in this talk

Why do AI pilots often fail to reach production?

According to Christian Winterfeldt at Rise of AI 2025, pilots often stall because organizations lack a common AI strategy and because departments, such as IT and the teams running pilots, are misaligned and unaware of each other's initiatives.

How much can on-premise AI infrastructure save compared to the cloud?

Christian Winterfeldt states at Rise of AI 2025 that customers running inferencing on-premise on physical servers save 75% of costs compared to cloud services once they scale to thousands of users, with an ROI on Dell's own AI platform within 12 months.

How should a company choose which AI use cases to pursue?

Christian Winterfeldt describes at Rise of AI 2025 combining a top-down question about what differentiates the organization with a bottom-up employee survey, then narrowing candidate use cases, in Dell's case from over a thousand to around ten, based on data availability and the six core AI capabilities.

Mentioned

agentic inferencing, retrieval-augmented generation (RAG), strategy house, data contracts, vector database, model hub, AI factory

Keep exploring

More from Christian Winterfeldt

More from Rise of AI Conference 2025

More on Enterprise AI & Adoption

Transcript

Auto-generated transcript, not edited. · 6,332 words

Show the full transcript

That's the spirit. And Christian, I have something for you. Tada. Thank you. Thank you. So, first, thank you to my my previous speaker to Dr. Shif. So, great segue to my presentation right now. She talked about strategy. She talked about AI um platforms in the stack. And by the end of this uh this session in 20 minutes or so you will have a very practical guide a very practical blueprint how to build your own AI great production environment and with a very practical checklist which is giving you a clear understanding okay what I need to do and where I'm actually very good at. So if you believe it or not three years ago it was super tough to talk with organizations about AI.

two years ago, every executive I spoke to said, "Hey, I have an generative AI agenda." But they don't. So, um, but AI is, as you know, much more than just ChatGPT. And behind me, you see the broad universe of different simutations, optimization, machine learning, all the things we're doing today. We separate AI into three distinct market. The first market is the pre-GNAI market and this market was created 30 years ago from now. So this was about highly specified use cases and we know about 2,500 use cases of those. Every use case needs an own architecture and was bringing quite good business value. The second market this is the LLM market. So we created large language models based on a huge amount of data and with tens and hundreds of millions of investments.

So this was a huge endeavor organizations did over the last two years. The outcome great versatile highly opt highly high quality LLMs we using today to making our life easier. So helping with chat bots, uh writing text, creating pictures, doing lots of stuff in organizations are using that to improve efficiency kind of bit. And now we are coming to the enterprise market and this market is about really creating business outcome which is bringing organizations to another level not just optimizing stuff. So really creating new products or creating a better process giving speed to an organization to really leaprog um and and outperform competitions and a new paradigm is coming up right now. This is what we seeing coming at space.

This is a gentic inferencing and what is it? Think about a um yeah a simple procurement manager. Think about and let's call her Anna. Anna is responsibility today is just purchasing stuff, monitoring performance and renegotiate contracts whenever needed. And now think about Anna's working with agents. An agent is looking continuously what the market is doing. The other agent is monitoring price fluctuations. Even other other agent is looking on the contracts uh we have with the vendors and even another agent is making negotiation recommendations um to to get better prices to get a better supply performance etron everything has been done even before Anna woke up in the morning and opened her laptop and now think about every organizational process improved like that.

So this is not just making Anna more efficient and gave her more more purchasing um um demand. It's it's about bringing the whole organization to a next level. And this is what we're talking about now. But to get there, AI is not just a fluffy tool. It's not just a one tool we are using. There are core capabilities organizations need to master. This is what we saw over the last two years. And let me give you another story about how to master these capabilities. Think about a program manager or a product manager and the product manager we call her Lisa and Lisa Lisa's task is to bring a new product into the market and she's using AI with that and what are the capabilities the organizations needs to master.

First of all bring more context to the LLMs. So standard LLM is just knowing 1% of the organizational knowledge organizations have today. So they need to bring all the corporate data which were selected over the last decades in the organization to the LLM. We're doing this with rack or we are fine-tuning models etc. So these are the capabilities you need to master and now Lisa needs to create some collateral for bringing this product into the market for marketing for press releases for uh let's say quality management or for for customer service and she's creating this and then we have other capabilities like what are the right data to use and this is the biggest hurdle we see today what data to trust and who's able to access these data which agents which AI models are allowed to access which data.

So these are data management agents we're using today and and and Lisa she's not a she's not a programmer however she's really it's very simple for her to create a new applications which are helping her or helping um quality management to have a chatbot easy chatbot to interact with customers and then you need to monitor the performance of a product. So is a product going in the right direction or is it not? What should you do? Um and now you're having uh deep field of analyzing your performance against competitors etc. So uh analytics agents and last but not least you have lots of infrastructure in your organization not only the data center but only in production in on assembly lines in optimizing optimizing the assets you already purchased the assets who are really bringing value to your organization.

This is a optimizing what you're doing. And if you master all these things, you're you're getting advanced. You're you're doing more than just pilots. And here, quick question. Who of you raise your hands have seen a great demo on a laptop but never saw it in production? Never hit it round running. Okay. One, two. Okay. Lots of you. And and this is a this is a very common theme we we we saw over the last two years. So, we saw lots of initiatives. And I spoke to a manufacturing uh company and he said, "Okay, we we created a great great AI application for predictive maintenance for our organization. We have some pilots in HR to support our HR team and that was great.

So nothing to say about that." But then I talked to the IT department and they didn't know they didn't know about these initiatives. So this is the common theme. So that we have a disalignment in organizations. So the IT department they have a governance they have they know how to secure the information of the organization and they have a they have a treasure treasure of of data which should be used in order to make the use cases better. And when I asked the the manufacturing guy hey how did you do that or why did you do that? Yeah, we just want to start with AI. And right now, I'm a deep believer in that we're not lacking on talent.

We're not lacking on on tools. We just need to align to a common strategy. And the strategy is what missing in most organizations. And therefore, we are systematically tackle the strategy with a strategy house. And this is not a model. This is kind of a mirror. So take a picture and and really compare this what you're doing at home. This is this is a mirror which is telling you where are you today. And this and like a normal house, you shouldn't start with the roof and you shouldn't start putting furniture into the house. You need to start with a with a with the fundamental things and this is the strategy. So the vision where where you want to start what is really the things you need to achieve and then you need to um get the right use cases and then you need to build a foundation which is able to scale based on the right culture based on data based on technology and an ecosystem because nobody in the in the entire industry is

able to do this alone and then you need the execution. So you need a um board decision. You need some power in the organization to actually make it done. And I will um guide you through these uh these different layers. This would be the red threads throughout the next minutes. And you will see how this really is a is a tangible approach to get an AI grade uh or a production grade AI environment. And let me start by a quick example. Think about a company. Let's let's take the company I'm working for. more than 100 billion in revenue, 100,000 peoples working there, more than 2,000 applications. Um, one exabyte of data we are using right now.

We have more than 500 AI uh programs uh going forward, 2,000 applications. So very complex and I'm a technician by heart. I love technology. But we didn't start with technology. We starting with a strategy and this strategy was really about hey um we asked uh two two questions. And the first question was was we what we need to do. This was a top- down approach. What makes us special as an organization? So if you're crafting an AI strategy, you need to ask these things first. What makes us special in organization? What are the processes which are supporting our advantage and are these processes really able to to improve by the use of AI etc. So you need to have this process first and the second process is really practical.

How do you do this? So coming bottom up and we made a survey with lots of employees and we are coming up with more than thousand use cases and we didn't implement it all. So you don't need thousand, you don't need 100, you not just need 10 or less use cases to really bring your organization to a next level and this is what we've done. So we identified patterns and then we looked do we have the right data to supporting these use cases? Are we have the right data foundation? And then finally you're deciding on an application, you're deciding on a tool. So this is this is easy. Are you going to this to do it by yourself or are you buying this?

And then you're matching this. So we're coming back to the six capabilities. Are you really able to leverage the six capabilities all of them to bring these AI initiatives forward? So as an example, what we came up with was there are only four strategic areas where we think we are really good at this is differentiating us from the market and we need to improve those four. And this is what we've done in go to market in services in research and development and in in our supply chain. We are the leading supply chain in the industry. And we came up with a um easy um sales chat for our our our salespeople to connecting all the information we have or our our organization that they are really able to get lots of information improving the time to preparing customer meetings getting all the documents together.

What is the customer right now doing? What's the industry? what is the financial report of this customer saying compared to what we are offering very compelling and saves lots of time of every employee but it wasn't easy so I was in this project for for quite a while we we went live uh in March this year it was easy to find a tool it was easy to find the stack but the get getting the data right was the real problem and the acceptance of the of the of the people was depending on how good is the outcome so we put lots of in reinforced learning looking at the answers and really made the database uh better made the data better.

So we we have now a acceptance or a usage rate of more than uh 96% of our people using this more than four times a day. So this was a huge success and with a survey and this is also something you need to do measuring the outcome with a server every other months we're looking how much time are people spending on what and what is the real tangible outcome and another view to look at this is the AI vision. So looking at um not only um bigger strategy looking at are you really going to improve a product or creating a new product to create new revenue streams or you want to improve your processes in your organization and you need to be very clear in what you are going to do.

So it could be either this or this way. There's nothing worse or or better but you need to decide. In the second level of the AI strategy house, you have to boil down the use cases and boiling down use cases to the use cases which are really bringing you forward. We are working not limited to but we're working together for instance with applied AI. I think most of you know this. It's an AI initiative very successful one in Germany backed by industry by government by education and we're doing workshops with together with them. And these workshops are very practical and they're showing where to focus and we are looking at the demand side. So what is really needed for your organization looking at different processes the customer journey for instance looking at a process map and benchmarking this against the industry and also leveraging own KPIs and on the other side you're looking at the supply side.

So what are you able to to provide as an organization to make this use case happen? So the AI capabilities coming back to the first slide with the six uh um um capabilities and also looking at what data are available today and where these demand and supply are meeting. These are the use cases you should tackle and you should have a long breath to to work them through and this is really bringing you um um really competitive advantage. Next thing um applied AI is providing a repository of lots of use cases. They're already um leveraged. They you can use this. You don't need to reinvent the wheel. Use what what others already done and and learn quick.

And then you come into the um how to implement these use cases. Typical projects are linear. You're starting a project then there's lots of huddle in the middle of this project. then you're delivering this project and off you go. But AI projects are different. There are there's they're a loop and you have really to continue use case by use case in your organization. It's starting with assess the differentiators and then you need to determine which process to improve with AI. Then you need to discover patterns and then you're coming up with a hey what's the data foundation I can use. So this is as you remember the the left side of the bottom up approach.

The next step you're choosing the right capabilities in your organization and then you're implementing the tools which is quite simple the tooling. So you have lots of tools in the market today choosing the right one off you go measure the return and after measuring the return you need to scale. So in our examples we tested this with bunch of hundred of of users but at the point we put more than 30,000 people on this environment. You need to scale. You need to scale and and by that time you had lots of information what's working, what's not, which data are good, which data are bad. And you have to do these reinforced learning. You have to improve your environment and your platform must be ready for scaling.

So just having something in a in a pilot running for 100 people, that's a that's a piece of cake. But really bringing this into production and having these full advantage, this is something you need to do. And once you you've done with that, you need to assess the differentiators. Okay, what's next? And this continuous loop is something we implemented very successfully with our own organization and with other organizations. So this is the hard part but I'm feeling very comfortable there. So I'm a technician. I I love this. I love technology speaking about this but this is more about not not only about a technology that's about culture culture data getting the technology right and the ecosystem.

So getting the culture right. I have a organization 100 plus people lots of management structures and really training these managers giving them the the experience how to use AI, how to make market research, how to leverage AI to make their day-to-day job better. This this is the really the hard part. You you can do this with encouraging them. You can do this with education. But this is something which needs to be part of your plan. You need to get your data right. I had yesterday a long discussion with lots of head of AI um in in Vienna and this was about um the awareness of data. So data are the real problem. Which data to trust in in in in an example of the product manager?

What's the right spec spec features of this product? Is it the one which the R&D department have has created? Is it what the marketing team put on the website? Is it something which the uh customer support has created? and think about these information are not the same. They are interpreted in a different way. So you need to have um really good data products to leverage and these data products together creating a data pipeline which are finding new use cases. So this is the crucial one technology of course I will come through this you need to have a stable platform which is able to do that and an ecosystem you can leverage. So on all these layers there's no one company you can do this all but you need an ecosystem which is working.

Coming to technology um AI platforms what what are the areas to concern so AI is not running on any software this is running on infrastructure so you can say let's run it on cloud we saw we see that customers running inferencing on premise on a really a physical server in their data center they're saving 75% of their costs so when you're scaling to thousand 10,000 people cloud services becoming super super expensive and you have latency and all these things going on. So you need these eightway big GPU systems to run this. We have right now situations investing in an um in an AI platform on premise having a model um platform. This is you have an ROI within 12 months.

So within a 12 month you are you are much way of what you are paying for different models of the hyperscaler but you need to do it right. The second thing is about data. 80% of the data today are sitting in the real world, sitting on premise in corporations and 50% of the data created new, they're still sitting in the real world. So you need to bring your AI where the data are. So in your production lines, in your in where you're manufacturing, you need to bring your AI there on the on the edge. And this is about the location. Bring your AI where the data is. Now we are coming to a slide with lots of blocks and and and fields.

So don't get distracted by this. This is how a real production AI uh great AI system is looking like. So on the left side left hand side here you see the sources. You want to leverage as many sources as possible your own sources in the organization. Um structured data and databases terod data Oracle whatsoever. That's the easy part. Unstructured Hadoop clusters things you like to to leverage. Then then you're looking at IoT data. You're looking at surveillance information. You might need to connect HR information. Who was on the conveyor belt at what time to get a new insight of of of of some optimization, social media information from the outside. So lots of information you want to integrate.

You want to have robust data products. The second layer is with these diversity on information there comes complexity. So you need to manage these complexity which is the data platform. So in this data platform you need to create um data contracts. So I can trust those data and forward this to my AI model or you need to have federated queries from these sources. I want to have a continuous uh stream of information for these use cases. This is what a data platform is doing or um creating vector databases which is techni technically wise the way how you bring data into your LLM. So data putting into vector database put it into LLM. This is what we're calling the rec mechanism.

This is one of the capabilities of the six you need to master. So that's the data platform. Then in the middle you have your environment the muzzle. This is something where you run the models where you run all the inferences. You need to have a virtualized infrastructure small instances. Nvidia is helping us on that. So they have a great enterprise platform you can leverage. So where you're putting all your different use cases on these platform that's the next level. And then you have a model hub. So you don't use just one model on premise use several models in your organization and and have a have a model hub which is helping you to okay this model is now fine from this department in this model you had added this there's a life cycle management uh for the models and you're customizing what you're doing there and you have a your own um governance and and compliance into this so you need to get sure that these models are best case not biased um so that

you take all the the failures people done in the past are uh baked into the model. So you also have to have a look on this. So corporate compliance is also important here and then you need to operate this and this is ideally operated by the IT team. They know how to do that. They know how to life cycle that they know how to um um make AB testing in this environment. And this is um the next thing and everything as you see is based on a on a common strategy. It's based on on organizational governance. You need to drive. This sounds complex. It isn't. So that that's the the easiest part to get there.

This is as close as we can get to an um easy button. This is the DEAI factory. We built this together with Nvidia. Data coming in on the one side. Ready use cases are leaving this factory on the other side based on solid infrastructure. So we have validated solution. Um we have a reference architecture so you can really start very quickly to make this happen. The ecos system in the middle. So for instance we're working together with hugging phase. We have an own portal where you can download lots of um access to libraries where you can download different models so that you have an a model uh in production very fast. So our uh system engineers if if they are implementing this on premise as a demo hey they can do it in 30 minutes.

give us a server. We're implementing this um with that framework in minutes so that you see how a production grade um environment is starting and the services who are supporting organizations in every step of this journey and then hopefully you have the use cases which are really firing up your organization. Final slide as I as I said the checklist this is something you have need to look at. So if you're starting with your um AI environment there there is a huge checklist of the stack you need to know. So from infrastructure, cooling, power, the right GPUs, how to get the right network uh running to make this happen to to you getting your data right, building a data platform, what are the data contracts, the data platforms.

So and for every layer there are specialists. Some of those are in our organization, some of those are partners for us and we have a very detailed list which of the players in the market today fit into area. So speak to us, speak to us and start the conversation. So wherever you are, wherever you're struggling and this is not a model, this is a mirror. So think about where are your gaps and where you need support and speak to us finding the right people in that environment. So up to the use case builders. So the winners in this market are not the ones who building the big models um and have the fanciest tools. It's it's the organization who are really have a have a clear goal have a clear goal working towards it and having a platform which is able to scale because AI is not a project.

It's a fly wheel and it's turning faster the more you put into this. So, hard for me to say this. Don't start with technology. Start with a business goal worth scaling. Thank you very much. Thank you, Christian. Um, do we have questions from the audience? I had two questions and you answered them both. One was about the pilots and how to get this actually to real scale. Um you talked about the bottom up from problem side but also about um well that you also talked about cultures and um I would like to know when you look into teams like how do you overcome the cultural aspect the culture aspect of um there is AI why do we really have to use that I mean you're not inme but you work withmemes and thememes in Germany are the backbone of our absolutely of our country and a lot of them are saying I don't need AI and that's Well, so there there's a saying if you want to best build the best boats, don't teach your

people how to to be the best woodworker. Just just give them the the the the love for the sea. And so they're building the best boats. And this is in in in my organization, you have always on every organization 10 20% of the ones who really eager to try something new to start and start with them. Start with them and trying something show the results and then from then uh um go there. And for the for the last 10 20% okay you need to have an education path you need to take them on the hand and show this. So they're always the the the people who are advanced and they're always the leggers and you have to manage this.

This is just a management task you need to do. So it also you said earlier boil down the use cases. You now also boil down the people that you can work with in the organization like try to find the believers and work with them and then you create miracle things. You don't need 100 people. You just need five. you're really working together and then you can create something super special. Can I ask uh in in your team is this an interdisiplinary team or is everybody coming from the tech side? No, they are not techies. So they're all salespeople. They are uh they love technology. Yes, but we are working together with other departments. We we just started this and I'm just sneaked myself into this project team and try to listen to learn and so I just um spread the virus in my organization.

So people starting using this and it's fun. So you you see results, we are sharing these things. So culture is a super super important point. I think that's a really a nice way to end that you you have to have a lot of fun and there is a question. Hi, I'm good Mark Magnus from Esteem. I've been 30 years in AI. Yeah, I look at come on guys. So basically um I am totally in agreement with what you're saying. This is why we just uh this is actually a shameless PR. uh this year we launched Langops core a platform that caters to multilingual AI and has the interaction with LLMs knowledge graphs and actually a aic um or shall we say an AI tool layer.

So I firmly believe in your uh hypothesis that we need to think on the basis of platforms that can interact with LLMs. I see LLMs after 30 years as simply an alphabet. It's a new pro programming dialogue end of story to build something. I need something more and that is what we wanted at a steam to underpin and uh so look up esteem.se because we're a Swedish company and in there we have Langop's core and please communicate with my CEO he's here somewhere probably not in this room he's in the other one uh Yin HL he's German and we also have a German company which is a knowledge graph company. So these are the combinations that we believe very strongly in and also the way that is moving forward.

I really like this. Congratulations. Thank you. Thank you. You're going to be my best friend. Oh, the CTO of Oh, Philip, can you get this? Sorry. Thank you also for for your um presentation. I really enjoy it. Uh my name is Philipe the Castro. president of XU Exponential University and we are very focused on digitization. AI is certainly one of the areas we study and make research and um in our teaching we have a great deal of project based learning with companies. Now one of the main uh questions that come up to our meetings and to our lectures is how do you demonstrate for example specifically by using generative AI um demonstrate the increase of performance in in companies.

So let's imagine that I'm representing any theme here in Germany or you can even choose a corporate a large corporate organization. I will not put names on it because that could be sound awkward. But let me uh ask you in a pragmatical way. So if I'm a manager in such a company, how would you convince me to set up the business case? What would be the key performance indicators on a practical way that you would set up? And you would say by monitoring by managing we can fix these goals we can reach these targets in six months in 12 months either you choose I don't know customer service area or product innovation or manufacturing thank you so the question is about how how we manage performance how we monitor performance to really convince demonstrate so that it doesn't work only on the computer but really on real life yeah so it's it's it's about measuring so where are you today where you want to want to go.

There's a goal, there are KPIs behind that. You're boiling down these KPIs and sub KPIs and then then you're managing those. So for instance, we are working in customer service with the with an AI chatbot which is next best action. So we leveraged all our customer engagements over the last last decade. What was the action in a certain situation? What is the best next action action to do? And then we're measuring um the time this this uh incident was solved. The the quality feedback we received from the customer. Was it timely? Was it correct? Um was there further uh concessions we had to make because we maybe not been an able to to answer these things.

So they have different KPIs which are leading us into a way okay our customer service has become better in terms of faster better outcome and more costefficient time efficient whatsoever. So there are always KPIs behind that and we are as we are keen to measuring KPIs and and and and doing this or another example um getting getting time back uh to to the team. So we are saying you have 30% more time doing more valuable stuff. So we are doing surveys every surveys every other month. How our sales team for instance is spending their time searching for information uh traveling doing whatsoever and doing these surveys we see how the did the the implementation of the next tool of the next uh genai application help them to save more time.

So we are try to quantify this in a in a way with surveys KPIs we are measuring etc. Unfortunately, we're running out of time. And that was the first question. Sorry. Just just very quick. I love the presentation. As an artistic person, I love also a lot of data and structure and colors. Amazing. Uh the only thing uh where I just wondering if the last sentence I might slightly disagree because uh don't you think it's from from emotional standpoint smarter to start with a business pain point worth eliminating? Because my experience is that people as the lady before said don't uh like most employees uh they don't want to learn AI because it's another thing they have to learn and I think if you if you just as you just said now maybe it's absence here in what I'm doing right now.

Yeah. But I think just to get it sold in because the executives they always want it but most employees because I teach also AI is like yeah I'm going to do it because I have to and smile but I don't actually want to do it. Yeah. So fully agree um if it's only a comment the the experience from the last weeks recent weeks I I had with lots of organizations in discussion the ones who are in charge for AI they're also in Germany they're a bit unsure how to start this they're starting with small use cases proving the value of the smart use case starting the next one but this is improving at this example Anna as a purchasing manager is become a bit more efficient.

So she's can do 20% more. This is not efficient. So bring organizations to the next level. So rechange bigger um bigger processes creating new products. So organizations are hesitate today to making really big step. And an interesting finding was organizations who are investing in times of crisis are going out of the crisis exponentially better than other organizations who don't who just keep their money together. And this is the time point in time we are today. be brave investing and I like your phrase looking at a business problem to solve. Maybe it's a better better way to talk to seuite to get this done. Yeah. Thank you. Cutting cost is not a strategy to um rise right.

Last words you as CTO. So same company, chief technology officer for ASI solution. So that's why ecosystem that's me. Um two two comments on on the previous questions. one for you was how do I move the organization? We've seen this before. Digital transformation, we've all been there. Anything Seale says we're doing this and everybody needs to follow and we're going to put you all into a training in it and you're all going to do this and guess what's going to happen? Nothing's happening. as a as a nice theory from the military and and other things. Find a couple of people and this is interesting. This is a KPI thing as well. Find a couple of people in your company that say I'm going to do this because I want they're not asking you for do I get paid more, do I get overtime paid or whatever.

It's probably about 5% of your your your organization. There are people that are taking that torch that light and go. And guess what's going to happen when the first leader like Christian has a result on this and say I'm AI. Everybody else go why is he getting this and not me? That's the culture thing. But good thing for you, for all of you. Um, first of all, we launched a format in Germany for CIOS or C- level. One and a half hours once a month. Three people live on a webinar. Ask any question about this. Thank you. Find me on LinkedIn. The professor Ga Kra from Klaus and the former CIO of Continental. Three of us just on stage.

Just you ask any question. The second thing we've written a couple of white papers what sea level need to do. It's all open source. So anybody who writes stuff use it. It's open source. We're not paying for it. Thank you. And with that we close that. And thank you for giving me the torch because I have to hand it over to the next speaker. Thank you Christian. Yes. Round of applause.

Rise of AI 2027 · Berlin · May 2027

Be in the room for the next one.

Ten years of talks started as conversations in one curated room in Berlin. The next edition brings 300 selected leaders together in May 2027.