AI Insights and Decode AI: Navigating the current landscape and the evolution of our podcast

In this episode of the Decode AI Podcast, hosts Michael and Ralf discuss the evolution of their podcast format, focusing on the current state of AI, customer perspectives, and the importance of understanding use cases. They explore the challenges businesses face in implementing AI, the significance of data strategies, and the role of AI in enhancing efficiency. The conversation also touches on the hype surrounding AI, its impact across various industries, and best practices for successful integration. The episode concludes with insights into the future of AI and emerging technologies.

• The podcast is evolving to include more general discussions about AI.
• Customers are often behind in their understanding of AI.
• AI implementation requires a clear understanding of use cases.
• Data management is crucial for successful AI strategies.
• AI should be seen as a tool for efficiency, not a job replacer.
• The hype around AI is still present, but practical applications are emerging.
• Industry-specific impacts of AI vary significantly.
• Best practices for AI integration include training and knowledge sharing.
• AI can help break down knowledge silos within organizations.
• Future developments in AI will continue to shape business practices.

AI, Microsoft Build, OpenAI, language models, AI development tools, hardware advancements, Google Gemini, technology development

Transcript

Copy

1:40

Hello and welcome to Decode AI Podcasts. Welcome everybody. are back again with an interesting episode about, I don't want to spoil you, but I want to welcome Ralph. Hello Ralph. you? How Hey, you out there and hello folks. Welcome back to our podcast. Happy to be here, happy to see you. Michael, let's go on. Yes, I'm happy as well. So we had a little break. We thought about the structure as well about how will we continue? What are our thoughts about this podcast? And we realized we had a way too many people in the first place in almost every episode. had a guest, which is brilliant actually, but we also plan to talk about some general stuff and usually with a guest we dive into a specific topic and going deep into a specific topic. So we try to not reboot, I wouldn't say after seven, eight episodes, it's kind of reboot, but we start with mixing up a little bit. we start today with a new episode about general stuff coming from... from maps without any gas. then we're looking forward to have guests again but on a more random way and we continue with the podcast having guests on specific topics. Yeah, that's true. So it was really difficult for us to follow the path we had in mind when we started this thing. So we're kind of a wrap up of what happened so far, what is our journey. And it speaks out of my heart, if you say that we had too many guests and we were not able to cover general things at all. That's true. And also it was like We planned to have a dialogue in between and that didn't happen. So we ended up being an interviewer for our guests with very interesting topics. And we were very glad and thankful that we had the ability to host those cool guys here in our podcast. But yeah, it's almost a thing, Michael, that we should have some talks in between and discuss. several topics here, so that's really cool. Absolutely. So right now we would start with something like a general discussion about AI. So we make a full stop with all the fancy stuff and look around where we are, what's going on with AI, is it real? And we would like to talk about our personal experience, how we see the market, what's going on. around us. So I would like to clarify my personal role again, just to make sure you're aware where I'm coming from. And then we'll... Michael, we can take a step back. what the guys can expect from Decode AI in the further future. before we go into our topics, maybe it's worth have a round up with that as well. So our first decision we made is to keep English as the main language for Decode AI. Yes, we're here. we will not have any German decode AI talks from now on. That's first decision. Second was like Michael introduced you to is that we keep going as well with just dialogue driven podcast series where like Michael and me will have discussions or explain several topics as well as we will have for sure. the cool kids with us to talk about dedicated topics here, Michael. Do you have anything else? No, that's a good summary. And I think that's something which makes sense from our initial thoughts, talking about the different levels with different details and also getting more details on our website. As you are aware from the show notes, we have a dedicated website, so we put more details on that as well. Yes. All right, so just to make clear where I'm coming from, I'm this guy, how can I put this? Usually I say I don't have any ideas about AI and then I realized, I do. When I talk about AI, I always think in the first place about co-pilot for Microsoft. because I'm a Microsoft MVP coming from Microsoft 365, which is very close to Co-Pilot. So that's first place. But over the last months, I was working with Azure OpenAI, I'm working with customers on AI strategies. So I have more about just Co-Pilot. And on the other hand, I'm also using on... the community side. AI products. So I discover more products. I'm looking into the latest news about some AI tools. And Ralph and I share some things about what's going on on the AI market, what's coming up with AI stuff. yeah, I would say I'm the guy coming into companies talking about AI in general without the real deep knowledge of configuring everything. So that's part where I'm still discovering things almost weekly. Great. Yeah, a great introduction of yourself and reframing yourself because I mean, there is a development in between. So from the beginning of Decode AI to today, and this we have to say, we both made a development step ahead here. And it's also with me so that I'm talking with my customer about strategies as well as we're talking about methods. And we're also talking about how to implement, how to provide a valid platform, how to secure all this stuff, and how to get it in an operational mode. And all these topics are pretty interesting. So we will cover in this session as well a little outlook on this, as well as on some general use cases as you may, as they... considered to be IP as well. So we cannot go into the deep details, but we will highlight some of the ideas we see in the real world. And we also want to give you a statement of what's the state of AI for today. Yeah. That's actually our first point in our agenda for today. I feel it's like we had a real hype over the last two years. And I want to have a look into what's the actual point in our AI journey. So what's going on? What is real? So when you... Talk to customers, Ralf. What do you think? Where are the customers right now in the journey of AI? They are always four years behind me, but... Yes, that's true. So, no, seriously. the customers at the moment have played around with all this stuff like we all did. They were fascinated by the fact that they could utilize human speech to get valid answers from a machine and they... either could negotiate over data and validate data and get stuff done. Then the introduction of Co-Pilot and the progress of Co-Pilot and other tools which are helpful during your day to manage your everyday tasks came up and thus they also have a raising in the industries. And the fascinating thing is that from all the concerns, it turned into you cannot be without AI these days. In some areas, I do have a fantastic example for that. Talking with customers about software development these days is a tough thing because they have their fantastic idea. There is GitHub Copilot and you have to use GitHub Copilot, so I pay you less. because you're using GitHub Copilot, right? So GitHub Copilot is doing the work for you so I can pay less. We both, Michael, we know that's not really true, right? So we do not have less work because we still have to prompt Copilot. We still have to have the knowledge about the structure, the software itself and everything. So Copilot cannot do everything for us. It's a time saver and... Maybe it can automate some processes here at this stage. Yeah, so customers are going over to have it partially in production in their everyday business and use cases are developing from test state to more and more into production forming out their use case and as well the business case behind because I mean you don't have a project without a business case at any stage. Usually. And this is really interesting. And before, we saw projects coming up without any business case behind just to play around, just to understand. But I still see that we have a lack of knowledge out there, lack of understanding as well. And I'm pretty interesting to listen to you. What's your experience with that at customer side? Yeah, well, it's pretty much the same from the sort about I'm coming to customers and they tell me, I have an idea about using AI. And then I say, whoa, whoa, coming back, please one step back. Let's talk about the real idea behind that. Let's talk about use cases and what you may have discovered already. So I tried to... reset the customers because they, it, it, for me, feels like they are using this phrase AI for almost everything. And they discovered some things because they played around. As you mentioned, there was a period of time where almost every customer, every, every IT related, company was, And you know something's there. you have to put AI in every name and description and every tool and you don't know what's actually working or not. So they tried to figure out what's going on and they discovered something like, I heard this from Google using Gemini for this specific workflow and it was great. It's absolutely fine. I heard from Microsoft about some use cases and I tried to use this. Brilliant. And then I realized it's not my business. It's not doing anything. So it's still the issue. from my point of view, everyone touched AI on a basic level. And now they have to discover what are the actual use cases? What are the benefits of AI? And most of the time the companies realize, yes, AI sounds good and the use cases we have discovered are good as well, but I have to do some work with my data. So Copilot, that's a... Sounds ridiculous, but Copilot is an easy one. So especially M365 Copilot is an easy one. We have to follow some specific rules. you have to put the data you want to work with online, you have to connect it at least online to make it available for the search, you have to set the security boundaries so not everything is visible for everyone. And that's it on a very high level view. The rest is more about prompting and how to use it, but the data is prepared. be kept because you need to on your keys. So, it's there. But the real use case for the whole company instead of just some specific departments is not about the data in M365. It's about all the data like SAP, CRM system in general, marketing reports and so many other tools. you have to... we think about. coming from and how to integrate this data in your AI strategy. That's my point. Long story short, everyone thinks they have an idea about AI and how to use it. And they just don't know how to get the big benefit out of AI. that's something I see as well. it turns out that also at your customers, it turns out that knowledge is something which How can I say that friendly? Is missing in some kinds regarding AI. And for a success story, in my opinion, it is really necessary to train the whole company on everything regarding AI. So starting from the management down to... all the hands that are working with that company to get an understanding of AI to enable them and to also get concerns covered as well as have an idea about use cases, they can utilize AI sensibly within their company. I listen to a phrase these days and it was kind of, if you start a project without AI these days, you're going to fail. And the response out of the audience, of the... Audience was... You claimed the same shit for blockchain five years ago. And I I read this, I will try to give you an idea where I read this. Maybe I will find the source of it again. But it is somehow true. Because think of if AI will have an impact to your project, and if so, plan with AI. But don't start any project including AI just by default. And there is a famous guy out of the DevOps community who said, don't ever accept the defaults. So, and now this comes together and I say, think of if it makes sense, utilize AI. If you don't see a value add, skip it. And maybe you just use it for running the project, but not involving it into the. product or the result of the project as I would state here. What do you think Michael? Yeah, I totally agree. My company developed a tool to kind of measure change management. So you get an idea about how successful was your change, how many resistances you have and so on. And this tool is based on questionnaires because you need some answers from multiple users and you aggregate this. all the stuff together. I think one year ago, we got the first question about, what do you do with AI in your tool? And I said, nothing right now? you have to implement AI. And I asked why? What is AI making it better in the first place? And the answer was that nowadays you have to implement AI, otherwise the tool will fail. And this statement was something I really thought rarely really, really long about. And I realized, nah, it's not true. As you said, it's an option. And you said you should consider it. What I see right now is every company tries to put AI on a product they may have already and improve it. And it's real quotes. improve the tools with AI. the bandage is so small. I think it's not useful. When I use it to make the next step into my project, when I use it to improve the work in the backend, so the stuff I do to complete the project, that's fine, but it's not necessary to put it in every product just for the sake of there's AI into it. you can use the same code as we have. Don't change anything. Use ProPilots or Excel to get the results summary. Now we use AI. and say, yes, now we use AI in our product. And that's ridiculous just for the sake of AI. So it makes sense. to really think about it, identify what's going on, does it make sense, it? Do you think, well, do think it makes sense to... Now let me rephrase the question. Do you think AI will be the driver itself for the next couple of years? No, no, no, I don't see that. I don't see that. What's happening right now? So as I've said already, so AI is at the moment the handy thing you can have to optimize your daily work, right? It can be a time saver. What ends up in that you utilize this time for something else. So it is. It is not that you get more free time. It is just that you get maybe more things done or that you have an easier overview about things or that you automate some toil processes which were in your manual tasks before. This said, we cannot expect that AI will be in the driver's seat for the next years. It will make, from our point of view, serving customers and supporting them to implement AI. It will be a definite thing we will have a look on. I mean, remember things like IoT, things like big data, blockchain as said already, they became so usual and common in general that it is no longer such a hypey thingy, it is there and we utilize it. And as we've seen now, after two or three years, we don't get less jobs and not this great wave of releasing people is happening due to the fact of AI is there. So that is not yet happened, right? We see back. things, so the misuse of llama, for instance, by as it is an open source model. And this is utilized by other governments in the military sector, which is prohibited by the license, but nobody can do something around it. But it's still not that PCAI as a driver here. So it is utilized for something and it's not the driver's seat, like what it takes over. I have a strong guessing that Elon Musk would discuss that hardly with me, but still it is a thing valid to think of, valid and necessary to have in mind, to understand, to be aware of and to integrate if possible, but it's not the driver. I totally agree. Do you think AI makes a difference or a different impact for specific industry sectors? I mean, the misunderstanding out there is that AI is a brand new technology. It's not. Like IoT, it was there before. The NASA already had telemetry data from their rockets when they traveled to the moon. And thus we had already AI within the industry. So no, it is not. It will have like, is, as I said, the most common use case is being utilized as a time saver to automate 12 processes, which had like negotiations in between requests, which needed to be understand and stuff like that. And as well as like summarizing up documents or whatever data you want to have, it'll support you with by creative things, but you still have to be the creator. You still have to prompt for the creative things. And the nice thing is you don't have to understand the paint stuff anymore. This will be done by AI. So I hope that answers your question partially. Yeah, partially. As you described this, my understanding, it's still the point... Let me go a back. I totally agree about, yes, this is nothing new, understanding about AI and what's behind this phrase, these two letters, this has changed over the last two years. Yes, So what I think about is you mentioned big data, for example. Everyone was collecting data and getting everything together, but no one had an idea how to work with the data. And they were messing up everything with it, yeah. Yes. And then a couple of years ago, got machine learning coming up and getting more popular. And machine learning is, when you look at the specification or some explanations about AI, is part of AI. So it's there for quite a while. Now we got some generative AI and that's the new stuff. So from my understanding, AI is, I totally agree, it's there for a long time. And AI is, coming back to the previous point, is more about learning to work with the data you have. And you can use AI to get better, faster input into your data. You can get more insights about the data. And talking about different industry sectors, we have customers in manufacturing. They are more interested into getting data about the machines they use and what about any recalibration, maintenance windows, something like that. On the other hand, we have other customers working within retail and they are more interested to get data about transportation. delivery tools, things like that. the data in the stores, how many objects they have and order. What is it in English? The point in time where you realize you need more of product X. thank you. quite high. So that's something I've seen over the last months to identify. Yes, there is something like the white collar worker sitting in the office working with Excel files, but that's usually not driving the business. except you are working just in an information technology company. But if you're using it on the industry, right, and using manufacturing, retail, automotive, whatever. So if you are doing something with other data, as I mentioned earlier, it's more important to understand there are more things to do as you usually hear about. And then you come back to machine learning is there since quite a while and many companies haven't used anything or just a basic level from machine learning with the data they already have. I mean, the difference is in the past, the data were more like machine produced data where they worked on and they had dedicated models for it to assume whether there is an impact in the data, a gap or something to negotiate with that. And you're right. But what we also write in is or what we have missing gap here is the data itself changed. So we're now talking about documents written by humans, no more longer that structured data. We have unstructured data. We have documents with dedicated topics. Roughly to understand would be even for some humans heard. And now we have AI with the GenAI here. Let's symbolically speak about that it's reading the content and it is summarizing the content and for me with a Highly educated guess that this one the output is correct and it's not hallucinating. So the other problems with AI are still there we cannot we cannot ignore this and It won't change and and and this is the change here, but I mean These are still... So the change is like we had in the past already chatbots. We also had in the past chatbots where you could have conversations with. The new thing is now it has such a broad knowledge about things you can ask. Like it is more or less like Googling in a conversation. And this is... This is the new thing. it is really the brand new thing. Using natural language, you speak to search, to reorganize, to request things. And that's been done then at the end of the day. I mean, that differs from the approach. I mean, no, it differs not from the approach. So the idea was all the time from 1942 back to today. that it should be something, machine, a human-like thinking, which the transformer model isn't, but it feels like. And I just wanted to point out that we're talking about artificial intelligence, which was named on a conference in 1950 as the first time there. And we're talking, and the moment it was spoken there, The approach was completely different. We talked about hardware running like for loops and stuff. And then we came over several technologies towards today. And that changed from over the time, like machine running does stuff, then programs run stuff. And then you already mentioned that we have machine learning there. And then we have dedicated services where you just fine tune your machine learning model with a dedicated use case. So it was pre-trained and then you maybe for recognizing images, could fine tune that one to get the job done you wanted to. And now we're talking about generative AI. The generative AI stuff is really dedicated to LLM. And in general, when we talk today about AI, is in 90 % all cases, referring towards generative AI. This is the point we have to talk about. Sorry for that going that far back to explain this, but I guess this is somehow what I was trying to mention earlier. Where we hang up now for a little while, but that's pretty interesting, And I've forgotten the original question here, Michael. Yeah, actually. So from my point of view, AI has got a real hype over the last two years. And most of the people forgot that still was already available. And now with Gen.ai, we got the hype to bring up the capabilities we already have. to use it more because it's easier to use and understand. For example, we get summaries about meetings we have, emails we get, someone created on a regular basis. So all this stuff we already have. is easier to access and understand. And sometimes it's still hard for companies to abstract, to think about the use cases. Usually it's not the office work which brings the business. It's necessary to do something in the background, except you're in just in this area. If you get your money from some products, some selling something, then it's usually a good support, saves some time, makes it easier to work with. and then we should work with. Yeah, to get it better. sound a little bit like we're over the hype cycle already. That's what I get from what you're saying. Is that your opinion? that your... Okay, I don't think so. Nope, I don't think so. So there's still improvement coming. So we're just at the beginning of the hype cycle. Why is that for me? So we coming up now with, we have now AI agents. the... The whole system is getting bigger and bigger. The understanding of this stuff is just right now coming up. And what businesses are struggling with is the possibility and the ability to imagine their use cases can become true. And this is something which needs to be break up or which needs to break up so that the customer feels like enabled. it is okay to bring in an idea which sounds for me as a customer, as a user, fantastic, like a superhero story with Superman and all this stuff. But maybe it's possible to be done. And this openness to imagine or imagine about solutions is what you need to come up with cool use cases for AI. And yeah, it's true. I don't feel that we are on the edge of the hype cycle and it's going down. I still feel that we're going up. We have SLMs coming up. We have agents coming up. We have different systems to utilize agents within the products or within the projects and so on. And this opened up a new possibilities, new ways to go with AI at the end of the day. And this makes a difference still to me and there is still improvement. Thus, I'm saying that we're not yet at the tip of our hype cycle and it's going down. But it's okay if we have here different opinions. Yes, and I won't try to convince you with my different opinion, as usually. No, the point is I totally agree about the use cases. And you mentioned the scenario where everyone tried something with AI, they worked with it, and everyone was into AI and we have tests. And just for getting in touch with AI, for the sake of getting in touch with AI, they started projects. working with AI, whatever. So they haven't had any use cases. They haven't had any specific ideas. It has to be AI in the project. So just give it a try. say it was an AI-centric approach. yeah. And nowadays we got the shift to we think about use cases and how to use AI over there. So that's the reason why I, from my perspective, we are not falling down, but we are off the hype. So it's not just use AI for the sake of AI, but use AI where it's good. where it's makes sense to use AI. So it's coming from the hype up to a commodity thing. And in reality, it's still something we use quite often. As I said, before Gen.ai, it was kind of not existing in the heads, but yeah. Yeah, but still, when I remember back to the IoT hype cycle, it has kind of similarity and that was still an impact there. Because, I mean, what will happen next? Other AI platforms will race. Open source project will race. This will all grow and thus the hype cycle is not on the tip. Maybe you're... talking about the development cycle in a hype. So will there be the next big thing like Steve Jobs said? But wait a minute. One more thing. No, it is still in development. We still have not reached the tip. The transformer model got updates. We have, like I said, SLMs there. And this opens a lot because what will happen now is that We're going to see specific LLMs like for law, for medical, for whatever. So those LLMs or maybe they'll be SLMs will be specific trained on on the law, on medical information, on stuff, and they can understand medical measurements like blood pressure on stuff, and they can give out a suggestion about the cause of a patient experiencing whatever. And those, I'm really not sure if we've reached the tip of the hype cycle, but we can follow up with that and we can see whether you or me was right. I think it's a definition of hype. Yes. had the same in mind. Sorry, I had the same in mind. are not at the end of the development by far. are so many improvements, modalities, integrations. So this will continue and the technology part is going up like a hockey stick. that's still increasing, getting better and so on. for companies. They are switching to use it as a tool, as it really is. Instead of we are following the hype, we have to use AI in some cases. Or in every case, as we are using AI where it makes sense. Use cases is the point. Nowadays, we are looking into specific use cases. What's in it for me as a company to invest into? AI or use AI and that's something I think it's also what I think the next half a year will dominate the discussion about AI. I, on my opinion, you're more or less referring to a majority of this topic, less than about a hype. If, if maybe that's more, the better, description here for, for, for what you're describing at the moment. I'm, I'm, I'm with you. I'm completely with you. So it's, it's really cool to see that we are now talking about real business cases and use cases instead of. Can you. Can you show me the stuff, make it accessible to me and I will play around with it. There was some, I mean, boilerplate template you had to run with customers to bring the technology and neuro to them. And that was fun for sure. It's still fun, but yeah, we're over that. Great. We have several more topics. We run this already a couple of minutes. I don't think, so maybe we can answer the question, is AI still in the hype? My answer will be yes, and your answer is. Time off. Okay, great. So, and what we also can state is that AI becomes reality in the customer view or in the businesses and that it is having an impact to them. Absolutely. And I love the scenarios where it's not right in your face, right? Yeah, it's just somewhere in the background and doing some fancy stuff you don't see. You just feel it got an improvement in some ways. So that's the point I really love to see. And there it's really in our reality. So you feel it in... multiple services already. That's really awesome. So we've talked about where our customers at the moment, where do we see AI? What's our daily business doing and dealing with AI? Where are we personally right now at the moment on our journey with AI? What is the next big thing we tried to figure out? And we had some kind of answers on that. We put it in an order, whether it is a hype or not, is it reality or not? I like all those information. hope you out there feel that this is a worth for you as well. But as we said, also, when we go to a customer, feels sometimes like that we're going five years back or four years back or two years back, but we're always going back as we are at the cutting edge, which is not an issue to us. And you explained, well, this is the reason why the customer is paying us for it. And I see that as well. saying that said that we're going back is it still may be that we have to talk about how to get started. Sure, and that's pretty common. Usually you take the experience from what went well and what didn't. So, from my experience, the first step is always managing expectations. As we said, the hype was there, the hype was huge, and everyone said... AI is the best, the biggest, the greatest, the extraordinary tool, solution. Exactly. yeah. now use cases is the story. So we have to talk about expectations. What are the limits? What are the experiences we've got? by just dropping AI into a company without telling anyone or without preparing your organization for using AI. So it's more about the first experiences we have with AI. I'm referring to a story I always tell about teams at the very beginning. So I'm working with Microsoft Unified Collaboration. Since I don't want to talk about the years, it's making me sad how old I am. Anyways, it's about the first experiences where someone created multiple thousands of teams just because they had no idea how to limit that, how to explain, how to work with teams and different channels. It's a no brainer for today, but six years ago, that was a new experience and they fall into some issues and traps. So they had to make the mistakes to learn from them. And other customers learned from mistakes other companies did before or had before. So this is the... Right now, not talking about AI is the biggest, best, greatest, awesome tool ever. It's about talking about issues, expectations, and how to avoid issues and handling the expectations. So when I summarize up, you would start with, okay, this is what AI can do for you. And these are the don'ts. And to negotiate the expectation of customer versus AI capabilities, right? Is that what you would start off with? Yeah, as I said at the very beginning, during the first discussions, I'm coming into a company. The idea is we want to use AI. That's the first statement. So you haven't had any discussion right now. You don't know if it doesn't make sense. it bring some benefits for the company? Is this a value? for the company or project or something. And it's just a statement. And then you have to reset the customer and say, okay, okay, going one step back at the beginning, let's talk about use cases. Let's talk about what do you expect? What do you get from an AI integration or tool? So that's... the first step I usually do. And sometimes it's not that the customers have no idea. Sometimes they thought about this and they already figured out where it makes sense. But it's necessary to ensure you... the customers have to, or had those thoughts already. they are... not unprepared, they know what they are talking about. Yeah, great. see that. OK, so when you... Let's face this out as a recommendation. So when you engage Michael or me, no, you as a customer, you should... What would be the first thing you would say? You want to work with AI, you as a customer, should... Think about the use cases. Think about the use cases. Great. I'd start a little bit earlier with get AI knowledge. And this will lead you easier to the case Michael is stating here. Get use cases for it. A little bit of this would make sense. So maybe a mixture of both. So without any knowledge, you're going to be trapped with that imagination stuff. So you need to enable people to think of what are the possibilities and capabilities of AI these days. And then you can get the use cases. And I got your point here pretty well. So instead of just having on the agenda AI, which was the AI centric approach we had seen in the past two years, get an idea about what you want to do. So you stated that you're coming into customer like doing stuff with Azure AI already. You also have customers utilizing copilot, I think. What is the reason that they are utilizing copilot? Do you have any idea? Or what is your experience with that? I mean, It's still not that we're coming to where to start, but it's out of the blue, and I'm pretty interested into that. So especially with Microsoft 365 Copilot, the most companies realize we have a lot of meetings and it makes sense to summarize these meetings and getting faster into the topics. So that's the biggest and most requested feature talking about Microsoft 365 Copilot, which is ridiculous. There's a cheaper license called Teams Premium and has the same feature. Other story. But the point is they just think about the use cases, the regular pros work process they currently have. And usually it's coming from the administration in a company thinking about co-pilot itself because they work with Microsoft 365 usually, Steams, with Exchange, with SharePoint, as I said before. And so. They live in this cloud world already, or they are on a journey to the cloud world. And they get the most out of Copilot because of this. The data is already there. Yes. So access to the data is crucial for the success of rolling out M365 Copilot. And they got the idea from Microsoft how they can benefit of working with M365 Copilot. That's also kind of the managing expectation part. Usually they have an idea and the story from Microsoft is, we invested billions of dollars into Open AI and we leverage to chat GPT for all and everyone just hears, it's chat GPT for all. So I can work with chat GPT in my Microsoft world. And so that's not true. It's not false, but it's not true. But by saying that, so maybe it's worse to talk about co-pilot in another session. What's behind that? Yeah, for sure, for sure. was just, I just was wondering about the common use cases, why they are requesting it. Okay. And then we realized there's more behind that. And it makes sense to work with other things, how to work with the data, interacting with different areas like the documents in your emails, working with email sentences, not sentences, with email summaries, putting out the data, working more with that. Yeah, I like the prompt about when I got cc'd. you know, when I got cc'd, what is important for me to know? OK, that said, so suggestion would be if you start with AI, think out of the box. Absolutely. No, just sorry. I know I talk a lot about, I talk too much right now, but I want to, I want to add one important thing. Not only think about outside of the box. Don't think outside of the box only because AI is also helpful with so-called low hanging fruits. So if you have a work process which is repetitive, the time. you're digging into a use case here, which is very, very dedicated to something. Is that something that we can bring into a compliance session more or less? Or do you want to go into that deeply right now? I feel it's about AI in general, not only co-pilot. Because sometimes customers try to make it more complex and use AI for the most complex scenarios, which are some breaking points for AI utilization, for using AI for good. It's more they have to use AI for the, because it's powerful, the best, as I said. So they try to solve complex issues at the first place, instead of going to the quick wins, helping the most people in the company. That's a thing I try to say. So think, think outside of the box and get easy things done first. Great. Cool. so recommendation would be built up some knowledge, think outside of the box and get easy things first done to save time to get into more complex things. I would add something here. Think of the data available. You can work on with. AI and maybe you're going to structure that a little bit for you. Cause yes, it's true. AI can work over unstructured data, but we humans deal hardly with that. thus structure your data a little bit that you have an idea about it. Great. Okay. That said, where to start else? Listen to our podcast. That's one thing. Go to our Meetup, another thing. Join the global AI community, for instance. Sign up for the newsletter. Look out for your region where there may be a global AI Meetup to have current topics presented by great speakers and network and connect with others on the same topic. It's also a great one. But in general, I would say these are the starting points overall. And furthermore, if you started with that journey, don't be shy. Utilize public available AI like copilot in Bing or Gemini or whatever. And have in mind, don't paste or put any data into it, which is endangering business. contacts or something like that. So keep it roughly, highly, no customer data into it. We here in Germany would do so. I don't know from the outside, but I think of that data is endangered then when you put it into a public AI. But you can start with that. to understand what's, what is, try to understand what is the thing of... I got lost. Sorry. Okay. Let's that there. I'm lost. I was in a flow and interrupted. Okay. going back to our topic list, we had a start. okay. One more thing. Get your prompting skills a little bit improved. And that said is... When you talk to an AI, think of a little child, you have to explain everything, but that child has a good understanding of the things you're saying. So it'll do what you say in the way you say with the data you share. And this is the secret for good prompts on my opinion. I totally agree and that's a good advice. Because most people don't know how to prompt. That's a good advice. I take that with me. So I learned at least something new to work with AI. Great. Okay, great that I could share that with you. So what are mistakes when you start the journey with AI? What would you say? So we had one new business case, we had another one, AI centric approach could be wrong. What are also mistakes? So don't pay attention to knowledge within your company if you're the only one knowing anything about AI. ensure that all others can follow you. So spread the word and share knowledge and get them into trainings or so that knowledge is there. Ensure that you've addressed all concerns, because that's the next pitfall where you're going to jump in if you don't do that. And I had an interesting discussion these days about the EU AI Act. Think of all your private use cases. So I know that teachers started to utilize AI within their classes and stuff and all around. Think of that right now and stop it maybe if you think that this would be affected by the EU AI Act. So another great pitfall is not thinking of your use case, the impact and trying to understand whether it is in what category of the EU AI Act. this until 26, we're going to see a lot of stuff coming up there and you need to be aware of this. As we've already shared with you, there's an issue with llama happening or happened that was utilized within the military use case, which was prohibited by license. I mean, Meta cannot do anything here because it is the license they broke up with. And how can you limit that? It is open source, which brought up another bad discussion about the evil of open source if misused. We can start that also about guns. Yeah. Okay, those are, mean, let's wrap up of where to start. Is it a hype reality? What is... What are big mistakes and what is the expectation management like? We covered that for my opinion. Do we? Cool. we did. Do you have any specific use cases you want to share and you want to highlight? use cases. Yeah, yeah. So as mentioned by the intro already, we cannot share the details with you, but there are some in fantastic or not fantastic. There are some use cases we can talk about. And the main theme of all use cases is time saving from an employee. So to remove toil work from that employee and to be done automatically. It's not to replace an employee. Cannot replace that employee. It is just to remove that toil from, to give them back more time to do things on a project, for instance. And thus we had for a researcher Institute, the pleasure to work on stuff. And they were like, Hey, we're doing projects and while doing so and running that project or even when the project is done, members of that project got asked by other projects or members of the research institute about projects in the past or they currently running and answering that questions takes up to 20 or 30 percentage of their work every day. That's a huge number. So they are thinking of making the data of their projects available through an AI, like ChetchiPT or so, so that all project requesters or all those requests can be run in that AI to be answered there, to be pointed out towards documents with summaries and all that stuff so that the question is no more longer answered by a member of a project team. it's answered by AI based upon the data given to it. So that's a huge impact on the explanation they gave us for them and they are looking forward to do so. That's a pretty common one. Sorry for the interruption. That's something I hear a lot from different companies. And there's a phrase I've learned during my education, actually. If the company would know what they know. And this is an old one. But now with AI, you are maybe in the position to realize that. Yeah, no, I would say not because you still have to ask Yeah, sure. And you still have to ask for dedicated knowledge. With AI, would say the possible response and output could match more to your question than a stupid old school search in like SharePoint or something like that. And thus it is like breaking up the knowledge silos in a company. and share the knowledge better. Yeah, that is one of the most common use cases. In this case, we have a dedicated time server use case, also dedicated to knowledge sharing, but on another level. And that was quite fantastic to see. And thus, we have several of such use cases seen. I still have one use case in seen where it is more or less a consultancy agent consulting a requester what can be the next steps. And that was quite interesting to see as well because that was also a time saver project based upon the data the company has already and based up on what this specific AI to learn from their data. It could point out recommendations for the next steps that could be possible to run something into success. So really cool things. As well as we see lots of prediction models running in combination with a chat GPT where you can request over a large amount of data. We're talking about real big data here to have suggestion or not a recommendation, it is like that they are like an educated guess giving out what is the possible fail or what is the possible issue. And this was also quite interesting because having that answer with a classic research would take ages to get the information. And now it is really, it is there and looks Pretty promising. The hit rate is not 100%, but looks promising to me. So these are use cases commonly which I can talk about with you and share as I'm seeing there. And as said, the main theme is time saving. As well as I see a load of work which has been done in the background with that tool. That's amazing as well to see. And this ancient technology grazing is so, helpful here. Yeah. Do you have use cases you want to share? Nothing I can share right now, but not for my customer projects right now. This is... They are working with some IOT and CRM data and matching those data, asking for specific scenarios. And it's to a high level right now, so I'm not able to share a proper scenario. But yesterday during the What's New co-pilot session, co-pilot meetup I do on a regular basis. I've seen or I heard about a scenario where a company had a lot of layoffs. And an interesting scenario was from one of the users. They created. And I would say agent to create resumes out of project experiences. So they trained the agent to work with standardized data like project details and create resumes for a lot of users. It's also a kind of time saving. thought about the use case and it makes sense. So one person spent one week to create this agent and many users saved a lot of time to get the resumes quickly done and start with a new application. The next one was also working with local manuals. So they had the need to create localized manuals with many different languages. And they trained the... I'm not sure about the right term, if it was also an agent or if it was just something with Azure OpenAI. Anyhow, they trained the data on localized newspapers. So they trained the, let's call it agent, to get... local catchphrases, local terms, local words, and get familiar with the localized behavior coming out of a newspaper. And so they got a more localized, familiar agent entering, working with Swiss data. Cool. Okay, so I guess we fulfilled the promises we gave at the beginning of our podcast here. coming to an end and I see you're writing here that you're going to have to jump into another appointment as well as me. So that said, I would say we're almost done with our today's session, which was a little bit long. I'm sorry for that. But I felt it was really needed and very, very interesting to talk that points with you to negotiate also all minds. and to have the different views on the topics, which was really nice. Absolutely. And as we said at the beginning, I felt this could be the right move for the podcast. it feels like, yes, it is. That's the next step. Talking together about general stuff as well, combined with details, guest discussions. So it felt great. Thanks. So let me look for the catchphrase we use for ending this. Stay tuned, stay interested. Once again, stay tuned, stay interested. Sign up, listen up. Here we go. Bye bye. Take care all, and thanks for listening. Bye.