FutureTalks: Beyond the buzz – AI in business with Hilmar Hohmann, IT Director at TMS Trademarketing Service GmbH and Krzysztof Szabelski, Head of Technology at Future Processing

We recently sat down with Hilmar Hohmann, IT Director at TMS Trademarketing Service GmbH and Krzysztof Szabelski, Head of Technology at Future Processing to dissect the hype surrounding artificial intelligence. While companies are eager to embrace AI, concerns about significant investments and potential pitfalls often hinder progress. 

Our experts will guide you through the essential steps to leverage AI effectively, starting with the foundational element: data. They’ll share insights on common mistakes to avoid and inspire you with real-world success stories. 

Hosted by Justyna Szymańska-Laskowska, DACH Business Director at Future Processing, this discussion delves into the future of AI and explores whether the current hype is justified. Join us for a thought-provoking discussion that will help you navigate the complexities of AI and make informed decisions for your business. 

Justyna Szymańska-Laskowska: Hi everyone, welcome to today’s panel discussion on the transformative power of artificial intelligence, its business potential, and real-life Justyna Szymańska-Laskowskae studies utilizing AI for their profit and growth. My name is Justyna Szymańska-Laskowska and I will be your host for this exciting session. Together with our esteemed guests, Hilmar Hohmann, IT Director at TMS Trade Marketing Service GmbH, and Krzysztof Szabelski, Head of Technology at Future Processing, we’ll discuss if AI is still a hype or a necessity. Welcome gentlemen, thank you so much for joining me today. 

So Hilmar, the first question is for you. Before we dive into the whole discussion, um, could you quickly share your interest in AI? Do you think everyone should jump on the AI bandwagon? 

Hilmar Hohmann: Um, “jumping on the bandwagon” is quite the expression. I would say that you need to dive into [it], or every IT Department, every company that has a certain size should be looking into AI. Um, you shouldn’t necessarily jump into every, uh, new application, every new AI capability that has been brought into, uh, into light. But, um, I wouldn’t say it’s overhype, but you need to pick, um, your applications or or pick the the areas where you want to apply AI quite thoroughly because you can sink a lot of time, uh, into that. 

Justyna Szymańska-Laskowska: So how do you know which ones you pick or which ones are interesting for you? 

Hilmar Hohmann: Uh, tough question. Uh, it’s sort of, um… let me phrase it this way: um, you don’t know what’s going to be successful before you you’re you’re trying trying it out. So what you should probably apply is a “fail fast” um, approach. So if you see at an early stage that it’s not working out, then you should probably skip it. 

And then we’ve had the discussion earlier, um, so Kristof said that he’s looking for external factors, external, um, pushing factors on on how to apply AI. So is there a customer that’s requesting something that would be beneficial for, um, the collaboration and that you could potentially also use internally within your processes? So that should probably [be] something where you should spend time on, um, and not, um, yeah, jump on every bandwagon so to speak. 

Justyna Szymańska-Laskowska: So Kristof, uh, do businesses suffer from the so-called AI FOMO? So is the fear of missing out and the pressure to stay up to date… do they seem to be quite a challenge for companies, especially those related to technology? So what’s your opinion here? 

Krzysztof Szabelski: I would say, uh, we had very, very strong FOMO in, uh, 2023. 2024 is [a] slower shift toward, uh, reality. Reality, yeah, that’s a good word. Like the cost of deploying that technology, its limitations, uh, accuracy you can reach. So it’s more and more now we have that discussion. So I see the potential, how I’m approaching that, and that’s the, uh, yeah, that that’s the way to go: to analyze, uh, shortlist potential improvements, uh, try it with a allocate some budget—typically low in the first place—uh, and see whether we are getting even nearer to some expected results. 

And even if the answer is not, it’s too cost… it’s too cost, too expensive to go, um, Next Level, there is one thing which… to it might be the answer. Like if you have that feeling of FOMO, and probably most of people have, there’s one thing that nearly every IT leader can do right now and that will be a good thing: is to take care of data. So we tend to tell that “garbage in, garbage out” and the AI machine learning algorithms are powered by data. If that technology ma matures over time, it’s ready to solve your problems, help on your challenges, and your data is garbage, you will have like another couple of years cleaning it. So investing it right now, deploying the right technology… sorry, data maintenance stores, investing into Data catalogs, data quality—that’s the right investment to prepare for what [is] coming in. 

Justyna Szymańska-Laskowska: So would you say that’s one of the reasons that businesses fail with AI implementation? That the data is not ready and people kind of jump this step? 

Krzysztof Szabelski: And it it is one of [them]. So there there are others right now. That the… for many Justyna Szymańska-Laskowskaes that I would say the cost curve—cost versus accuracy curve—is really getting steep after some point. And some, uh, uh, well some many Justyna Szymańska-Laskowskaes that are in theory possible to improve on with today’s technology, it’s, uh, it’s just too costly to achieve level of accuracy that satisfies us. So, uh, so so there there are many reasons but data is definitely one of the main one. I mean not not good enough data to train model to fit the large language models with, uh, and so on. 

Justyna Szymańska-Laskowska: Mhm. So so what would be the, um… Hilmar, what would be the good advice for the AI overhype in business world today? So where where would be a good place to start? So if if we feel that, uh, AI is being overhyped in the business world today, how we can kind of jump over the the hype or skip skip the hype and become more real? 

Hilmar Hohmann: I I wouldn’t say it’s it’s overhyped. I think, um, the the use Justyna Szymańska-Laskowskaes are real and to, um, to say this this clearly, I think that AI or or GenAI in particularly is probably the biggest thing thing since the invention of the internet or the introduction of the internet to to mainstream. And I don’t think there is any question about that. It’s about the, um, meaningful application within within businesses. 

And I agree that you need to get your data together, you need to make sure that your data is adequate. And most companies that exist for let’s say 10 years plus sit on an abundance of of data silos that are… they they just grew. Like we all know what happens if you acquire businesses, if you grow in in terms of Revenue, um, you basically make sure that you’re able to, um, support business but the data structure wasn’t the, uh, top priority, um, at the very beginning or or at the beginnings of the of the businesses. So you need to make sure that that is cleansed up. So I I fully, uh, agree to what, uh, Kristof said. 

And the fear of missing out part, um… I mean we’re driven by… we’re driven by our bosses, right? So if my management board, um, they’re going to the… they they’re reading the same articles that I read, they get the same, um, news articles, they read about AI. So it’s on their agenda. If it’s on their agenda, it’s on my agenda. So it’s not fear… like it is fear of missing out because, um, if your bosses tell you we need to look into this, you need you need to look into this. 

Um, maybe to to navigate to to go back to your question, um, navigate the overhype, um, is: Implement, uh, AI where has already been, um, successful. So as an example, introduce AI as a service. Um, the best example, uh, and probably or or the the two best examples that probably everyone can think of is Co-pilot within the whole Microsoft application suite and then, um, also GPT. But then GPT is also a good example of, um, data quality. So with GPT 4.0, um, GPT started hallucinating, um, and it got the the results got worse compared to 3.5. Um, I actually don’t know how it is now so I haven’t been following, uh, through with regards to that. 

But take use Justyna Szymańska-Laskowskaes, take implementation that have been successful, that have been proven successful in the past couple of months and years… it’s fairly new still… and build on top of that. So while you’re introducing that, um, identify use Justyna Szymańska-Laskowskaes that could benefit you, could benefit your process, your internal processes, the process that affect your customers. And then, um, also make sure that when you’re introducing, um, those scenarios, those, um, areas that have been proven, um, make sure that you have a plan on how to introduce them and and how to to use them and how to follow up on them. Just because you’re introducing Co-pilot doesn’t mean that everybody’s everybody’s using it. It’s just a new button that’s within every application and people feel reluctant to use it. So you need to have workshops to understand the functionality, you need to read yourself into the the functionality, and then also make sure that it’s rolled out, um, within the whole company and not only IT. Because within IT we tend to be sort of ahead the curve when it comes to intern processes. We we’re adapt to changes because that’s, um, the the the whole notion of of the IT departments is basically realizing the the the changes that, um, business wants to have realized. So maybe that’s a a a sensible approach [but not entirely sure]. 

Justyna Szymańska-Laskowska: So to continue that, um, Kristof, it’s not that AI is a cure for everything, right? So what are some common misconceptions of about AI capabilities in solving business business challenges that you can think of? 

Krzysztof Szabelski: I think the first, um, I would name two. Um, first is about cost because, uh, white… uh, public got feeling that AI is for 20 bucks because that’s what majority of, uh, services cost around that obviously. So AI is for 20 bucks and you can chat with it and you can show/do particular stuff. Uh, and then when we talking about automating or augmenting, as I as I like to say, business processes, uh, then we are getting into high token consumption, uh, we are getting into training which consumes the actual, uh, gear down there in the metal, uh, and that costs. And that and then there is a cost of that accuracy as I said. So in many Justyna Szymańska-Laskowskaes well 80% of accuracy is well pretty cheap. Then 90 is kind of project let’s say. 95 is a couple million project. And achieving 10… uh sorry 100, uh, percent is is impossible. So it it’s getting very steep. So so that’s that’s the first misconception. As we had a many conversation with our client about the potential and then we ended up with POC and actual visualization how much it’s going to cost to achieve a next step of accuracy and how much it’s going to cost running it monthly. I was like, “Why is it, uh, thousands? I expected like 20 bucks.” So so that’s one thing. 

Uh, yeah and and another one is with that, uh, accuracy. I think more and more, especially on the IT side, people understand that this is completely different than making IT project in past. IT project in past can be quite well planned at least on goal. Uh, we can promise that goal is going to be delivered. The cost and the whole scope can flow over time but the goal is achievable. Only AI, all the machine learning, deep learning, uh, large language model integration, any of that is sort of research rather than a project. So we can say that yes, with that budget allocated we can confirm or not confirm some hypothesis, but no one really is able to say 100% we’re going to deliver that. 

Hilmar Hohmann: Right, and if you remember our discussion yesterday where you had the number from the conference that basically stated nine out of 10 AI projects fail. So imagine you’re training a large language model and you want to achieve the 95% and it costs you a million dollars and then at the end you realize it’s a failure. You’ve basically sunk a million dollars. So you need to be very picky what exactly are you going to or are you trying to achieve. 

Krzysztof Szabelski: And and that’s so what we are advising right now is is to is that step-by-step approach. So a lot of organization already done some analysis or if not we’re happy to help with that. Uh, build that portfolio or backlog of hypothesis, that that’s that’s a good first step. Then rate them against how complex or costly would that be to progress on it and how much influence on the business that would have. And that’s a good first step. And then you can sell a couple of them for a POC, allocate let’s say small [budget]—because small depends on the size of company and stuff—but you can allocate small budgets to deliver that POC, see if that is actually what we expected, is it going the right direction. And that’s the moment, uh, after well probably couple of months to allocate bigger budgets and [say] “Yeah we’re going to progress that that way” or “Not not yet, that technology is not mature enough, it’s it costs too much for for the moment.” Uh, so step-by-step approach, that that’s what I recommend. 

Justyna Szymańska-Laskowska: Mhm. So Hilmar, TMS supports its clients in sales and Retail operations. Um, can you think of processes in your company or maybe wider the whole industry that could be the most successful with AI implementations? 

Hilmar Hohmann: Well within, um… I’m going to answer that question in a second. So I think first of all as I mentioned before, um, irrelevant of what, uh, industry you’re in, I think all the administrative processes can be supported, um, to some extent. Like translation, um, or translation services, um, like deep neural networks like GPT where you ask contextual questions and you get sensible answers like “write me a project plan” or “formulate that email.” So that is something that can be applied in virtually any, uh, industry, it doesn’t matter where you are. 

Um, in our industry, uh, we basically have two major use Justyna Szymańska-Laskowskaes. Uh, we’ve also talked about this before. Um, one is image recognition. Um, we have, um, or we support our customers, um, with in-store sales to, um, put it short. And in doing so we are taking a lot of pictures of shelf space. And the like… the “Holy Grail” within, um, the the salesman area or within that area is the the “perfect store” approach. So so you take a picture from the Shelf, you compare it to a sample picture, and then you see where do you have out-of-stock situations, where have articles been misplaced or products been misplaced, um, are the correct articles been placed. Um, that is like the Holy Grail, um, that we’re aspiring to, um, or where we’re working towards to. But this is rather difficult. There [are] whole companies that are purely set up to, um, realize this goal and they’re not doing this very successfully in all honesty. Um, so we’ve been looking into image recognition in terms of categories. Um, has… is that a picture of a category that is a shelf? Is is that a picture that is a category of, um, the the Justyna Szymańska-Laskowskah area or the store itself? Um, that is something that we that we have been looking into. 

And the second area that we, um, uh, approach is the “traveling salesman problem.” So if you have salesmen… so we have round about 600 salesmen that travel, um, the the German area and you want them to visit as much stores as possible right? So and you want to make sure that they’re as efficient as possible. And that that is it’s a well-known problem within within IT where, um, we’re using a heuristic approach to determine… with, um, it’s it’s like a a, um… how am I phrasing this… it’s like a drawback between, um, the best possible result and the lowest computation time. Because the the amount of computation, the the amount of calculations that are required if you hand in like 250 addresses that need to be visited is… it’s faculty [factorial] right? So it’s it’s a lot, um, of computations that need to take place. At some point, like, you don’t want to, uh, let the traveling salesman wait for 20 minutes until he has the perfect result. He wants to have a result after 5 seconds. So you need to take the best guess and then that’s a a very, um, direct approach or a very clear approach for AI in in that form—a heuristic approach. But that’s where where we’re actually applying it. 

And then keeping our eyes and ears open. So that’s basically, um, Talk to peers, um, going to conferences, um, seeing what’s what’s happening, um, on the market, what players are emerging, what use Justyna Szymańska-Laskowskaes are emerging and then, um, discussing internally what are the use Justyna Szymańska-Laskowskaes, what are the potential benefits that we could reap. 

Justyna Szymańska-Laskowska: And from a technology, uh, consultancy perspective, Kristof, can you share a specific Justyna Szymańska-Laskowskae study where AI significantly improved business outcomes? 

Krzysztof Szabelski: I would mention that documents. Wherever we have documents in the organization—and every organization have documents—uh, that’s the good area to look for use Justyna Szymańska-Laskowskaes or to start maybe… to start yeah Pro likely. Um, so I I will name two Justyna Szymańska-Laskowskaes just to give an example of what I was saying before about those cost and accuracy. 

One example is a project we delivered for a Career Spring. It’s a US organization helping people, uh, activate on a job market. And the big portion of the business is to, uh, move the, uh, job postings from one place to another. And as you can expect it, every job posting have different structure, uh, and so on. So that that’s that’s the thing that people been doing. Uh, we created… and that was relatively cheap solution created within a couple of weeks, uh, where the the content was just moved from one place to another, uh, automatically sort of automatically understand the structure and change it to the Target structure. And that’s the process that been eventually… that is eventually running in a “human in a loop” sort of sort of way. So that we measuring that about 75% of job is is is reduced but still people are seeing that. So so so that’s what the algorithm provided and I’m checking it, I’m correcting if something’s wrong. And that was relatively cheap as I as I said: couple weeks of work, uh, couple hundred dollars monthly cost of running. Uh, but the accuracy is we can call it low in that Justyna Szymańska-Laskowskae. It’s possible to organize that process wise. 

And there is another project that we’re developing for over two years right now and is about processing legal documents, uh, in one of the niche. Uh, and you can expect that legal documents you can really you you really want to avoid mistakes that you pointed at something which is not right or you not pointed at something which is critical in that Justyna Szymańska-Laskowskae. And, uh, taking that accuracy from that rather low to rather high is weeks to years. Uh, and that’s in that Justyna Szymańska-Laskowskae that that that product is at the moment, um, well looking for the attention on the market and testing in the first Justyna Szymańska-Laskowskaes whether this is enough and actually building the the trust, which is another another thing: Do I really trust that system, uh, that I can put my, uh, documentation through and Trust if it pointed to the right paragraph let’s say and I don’t need to read the entire documentation? So, uh, yeah those are those are really good Justyna Szymańska-Laskowskaes. There are a couple others that are related to text processing, documents processing, uh, and I think yeah that that’s a really good, uh, place to start thinking about optimizations. 

Hilmar Hohmann: Okay so yeah, what about development? I I completely forgot like this is is like the the most apparent use Justyna Szymańska-Laskowskae. I’m not sure like we we’ve been using within development, um, we actually have again Co-pilot integration. Um, we’ve been using it in interview processes and and people are sitting there like, “Wait I only need to hit tab? It’s doing my job?” Like, “Yeah but please explain what happened.” 

So, um, um, that’s really helpful and I… why I’m saying this… I um… because your question was with with regards to how can I use that within an IT consulting. And I’m happy that you haven’t said, “Oh my people can be 30% more efficient, I can sell more projects.” Cuz that that’s what I I actually heard on a conference from one of the biggest IT consultancies like, “Yeah now we can, um, sell our projects cheaper.” Like all right, your your employees are going to love that, you you need to be more productive. Um, yeah it’s not it’s not how we’re introducing it at least within the, um, IT departments. Um, it’s also a factor of happiness and, um, feeling well at your workplace. So if you have tools that assist you and make your life easier, that will make your employees happier and they will enjoy work more. And a happy worker is more likely to say and you he will create a, um, a better atmosphere within the within the team and people are more willing to help each other. And if you’re supported by technology that’s a it’s a a win-win situation. So happy that you didn’t say Effectiveness but it is some sort of, um, making life easier, um, within certain areas. 

Krzysztof Szabelski: Yeah that’s, uh, it’s a good point. And actually I can I can… our finding because that was an expectation of “Oh yeah now that there’s an AI writing code so the project going to be cheaper.” So we put a lot of effort into measuring the impact and stuff. And I could actually… if I’m… yeah we’re trying to be we are honest with everyone we talk with our audience. Uh, but I could if I don’t I don’t want to be honest I can say that yes AI is is is doing like 30% Improvement in in writing code with the little asteris[k] that this is about the new code in isolated scenario. But if we’re talking about the whole business day of developer who writes a new code, who fix it, who code review and other people work, who lead a team and talk to the business in the first place trying to figure out the best way to code… then it’s only about 5 to 10% of the Improvement that is coming out from what you said: that I don’t need to… I’m faser on a stuff which is repetitive and I’m not not enjoying it. I’m enjoying solving business problem and that not yet can be automated. 

Justyna Szymańska-Laskowska: Yeah. So let’s focus on the perspective of a company that would like to introduce or Implement AI in their daily Solutions and operations. So based on your expertise, Hilmar, what is a… where is a good place to start? What would be the first step that such a company should take if we wanted to give them advice? 

Hilmar Hohmann: That’s pretty much the same answer that I gave five minutes ago: start with the administrative processes. So use, um, AI as as a service. Don’t try to, um, get ahead of yourself and do every everything from scratch. So introduce processes that have been, um, that have proven themselves. That’s pretty much it in a in a nutshell. 

Justyna Szymańska-Laskowska: And how do you think businesses can ensure that they have the right infrastructure and the right talent for AI projects? 

Hilmar Hohmann: That’s a bit more of a challenge. Yeah that is actually a challenging question. Um, with regard to infrastructure, it’s not only infrastructure, it’s processes again. Uh, we we’ve had a a short discussion beforehand and, um, it’s making sure that your the skill set of your internal teams reflect of reflect what you what you’re trying to achieve right? So your, um, uh, IT support team can probably enable those services that I just talked about within seconds. It’s basically just, um, expanding the the licenses and and you’re good to go. But if you’re talking about, um, developing your own large language model, that’s a whole different topic. You need… it’s it’s not only about infrastructure, it’s about processes, it’s about the skill set of your employees. If you don’t have the internal skill set, you need to invest a lot of time, a lot of money. Um, that is depending on the company size, can be difficult. If you don’t have the the time, um, you might want to look into externalizing it. So you [Kristof] might be the perfect partner for that, but then again that’s a lot of money as well. Um, so if you’re looking into developing something yourself, um, you need to open the the money back. 

Justyna Szymańska-Laskowska: So Kristof, what are some some common pitfalls to avoid when starting with AI? 

Krzysztof Szabelski: I would say, uh, starting too hard or too heavy. Uh, expecting too much. Uh, there’re multiple reasons that you can just miss your expectations. You expected everything will be automated—it won’t. Uh, you can push people too hard: “This is an AI, you’re now going to use it,” and people are “No no no no no no no, I don’t understand it, it’s going to replace me, it’s blah blah blah.” Another… people human ability to change is probably the huge limiting factor here. So we need to [understand] and it’s not AI powered it’s not AI powered yeah. So, uh, yeah I would say start slow… not slow is not… start small. And, uh, step by step increase efforts and budgets to monitoring the output that we’re getting. That’s the good approach. And starting too big and expecting too much is probably the most common [pitfall]. So “go big or go home” is not the not not good for AI. 

Justyna Szymańska-Laskowska: Well yeah. So Hilmar, to wrap up our discussion, let’s look into the future a little bit. How do you see AI evolving? 

Hilmar Hohmann: Let me take out my crystal ball. Uh, it’s hard to say. Um, we’ll look into the video in five years and then we’ll compare. Look into the video and and yeah… so so I need to be very careful what to say now. Um, it’s it’s really hard. I think some use Justyna Szymańska-Laskowskaes, um, will develop themselves. Um, take Vision AI as an example. Um, this is a use Justyna Szymańska-Laskowskae that it’s probably not very clear if you just say it out loud but pretty much everyone has been using it in their daily lives, which is their cars. So there’s image recognition: the the signs for the speed signs, other cars, distance. So that’s all it’s all AI that’s happening in in our daily lives and it’s basically improving our lives. So I think some of it will develop naturally without us interfering. It’s basically part of our our daily lives. 

Um, other parts, um, will to some extent also evolve naturally. Again we’re going back to the Integrations into common applications. Um, foreJustyna Szymańska-Laskowskating, um, and now we can discuss whether this is true AI or whether that it’s like linear regression or whatever, um, you’re going to use. So basically, um, usage of historical data to make predictions. Um, I think it’s going to take a bigger and bigger role. But, um, ultimately all it all goes back to what Kristof said in the beginning, that you need to make sure that your data is, um, tidied up. So that is probably something that you should be, um, doing within the next couple of years to be able to jump on the bandwagon the bandwagon, um, when you see that a certain mat maturity level has been reached. 

So right now if you take a look at the Gartner hype cycle, we’re still at the… what was it called… the “plateau of, um, productivity”? Is it’s like the first the first slope that that is basically, um, the this disillusionment or or like you’re falling into the disillusionment. I think GenAI is the, um, at the slope of the, uh, absolute hype and now we’re being disillusioned on on what is actually being capable of. Am I only going to use it to translate emails or to set up a project plan? To repeat myself. Or what’s the bigger use Justyna Szymańska-Laskowskae later on? How am I going to, um, to to actually use it? 

And we had that that example yesterday where I was basically mocking the, um, the introduction or the, um, inclusion of a Co-pilot within PowerPoint. So I’ve presented a presentation that I’m going to show her later and it’s eight slides. It’s not that long. And and, um, I’ve never clicked the Co-pilot button within PowerPoint. So I clicked it two days ago. I was like, “What is it going to do? Is it going to arrange my…” like I’m a former consultant I I’m very picky when it comes to the formatting of of slides right? So I click that button and I wanted to see what is it going to do. And it was basically just, um, showing me pre-written, uh, uh, questions that I could ask the, uh, the the AI. And one of the questions was “Show me my important slides.” As if I wouldn’t know! Like I’ve created that slide myself, I should know what the important slides are, right? And then it said… so eight eight slide slide deck… and it said the important slides are 2, 3, 4, 5, 6, 7. So the first slide is an introduction slide, there’s nothing on there. And I don’t really… I don’t remember what’s on the last slide. So basically everything is important! Well is that the usage of AI? Is that going to be the the [killer] feature for a [presentation]? Probably not. 

Um, so it needs to stretch itself, um, it needs to be developed, but it’s still in in fairly early stages. And we’re going to see that, um, a lot of a lot of the, um, potential use Justyna Szymańska-Laskowskaes are going to be debunked. Um, they’re going to be not relevant in the next couple of years. And they’re probably going to be some use Justyna Szymańska-Laskowskaes that we haven’t even thought of. 

Um, another example where, um, I thought this remarkably good as a as a use Justyna Szymańska-Laskowskae: So my my wife is working, um, in a large FMCG company and they create thousands of physical products each year, um, and they ship all over the world. So where they are going to use AI as product descriptions, product photos, um, translations of those descriptions all automatically. And that is lots of people that have been doing that, that have been reviewing those descriptions in the different languages, have been working with translation studios. Um, that is all automated. Are there going to be errors? Most probably, right? But if you can automate that and you can then, um, reduce the time to Market from weeks to days or from months to weeks, that’s going to be a huge, um, Improvement in ter in terms of processes. 

Justyna Szymańska-Laskowska: And what is the time perspective [on] that? Is it 5 years, 10 years? 

Hilmar Hohmann: Oh that’s more like within the next year. So you can you can already use that. Um, the challenge is more finding the right startup that is going to be existing in the next 5 years still, um, where the product is still existing within the next five years, um, or being able to actually buy the product that has been produced. 

Justyna Szymańska-Laskowska: So Kristof, any final remarks, uh, from you or maybe a final piece of advice that you would give businesses how to leverage AI? 

Krzysztof Szabelski: So I will get back to the first question that we that I answered here is that: if you see a good Justyna Szymańska-Laskowskaes for AI to be deployed, um, build a POC. Make sure sure that is actually going to improve on anything and then invest more. And if you don’t see that many Justyna Szymańska-Laskowskaes, you tried something and it failed—invest in data quality. 

Justyna Szymańska-Laskowska: Thank you so much and thank you to both both of you gentlemen for, uh, this amazing exciting discussion. Uh, so thank you to our amazing speakers Hilmar Hohmann and Krzysztof Szabelski for sharing their valuable insights today. And to our audience as always, thank you for tuning in and, uh, for being with us. We appreciate your participation and we look forward to seeing you next time. Until then take care, stay healthy and curious.