Transcription
Good morning. We actually had so many participants in the last two webinars that we had difficulties with a provider. We have now fixed that. Everything should work now. We are prepared for you and have upgraded for the rush, as it always is with Zeit, right? That's how it is. Yes. 10:01 AM Annika. Yes, no more people are coming, but everyone knows it starts at 10 AM. Therefore, we are happy to start. Please share your screen, and then we will make it official for everyone who has joined in the meantime. That's how it looks. Yes, perfect. Welcome to our, I believe, seventh demo, which we are doing, and today it's specifically about project development, namely the project development of the future. That's how we titled it. Perhaps a brief, um, two more notes on the technical side. You cannot use the chat, but it's only possible to ask questions in this Q&A section. Annika will answer these questions live, or for the most part live, I hope. And therefore, a very, very active call to action. Please, at any time during the webinar and during the presentation, immediately type your question into this Q&A. I will work through it in parallel and also pick it up in conversation with Matthias, so that we don't end up with questions at the end like we usually do, where you no longer know what it was about, but rather that we can answer the question in the second it arises, in the best case. Therefore, it is very important to me, please ensure that I have a lot to do today and use this Q&A function, and then we will clarify it and, above all, incorporate it. Yes. And you can find the audio settings here on the right. Yes, actually as usual, showing the slide, I think, always. Exactly. And briefly on the agenda, what do we have planned for today? We'll do a short intro. Who are we? What is Zeit at all? And then we'll get straight into the application example. We've come up with something cool. Um, it's, uh, really like in real life. There will be a nasty surprise in between. I hope there won't be any more that we don't know about yet. Uh, after that, when we've done that, when we've carried that out, um, then we'll give a small outlook on what's coming next, and then we'll still have a bit of buffer for questions and answers, but we actually want to answer them live, so to speak, during the project. Um, yes, who are we? Annika, who are you? Yes, thank you very much. My name is Annika Meisters. I hope many of you have seen me before, at least on LinkedIn. I am your contact person when it comes to determining if and how Zeit can help you in your daily work to achieve efficiency. I alone, and also my colleagues. We are now a team of a few people who are available to you as contact persons. We call ourselves Senior Key Account Managers, and everything that is not answered today will of course be answered by me afterwards, and I look forward to the exchange also in one-on-one conversations afterwards. Today is meant to serve as inspiration, as a kind of overview, and we will do the in-depth part later in the respective individual discussions. I'm looking forward to that and looking forward to doing this today with you, Matthias. Would you like to introduce yourself? Yes, I will. Exactly. Matthias Zücke, I am a co-founder of Zeit. Together with David, we founded it. Um, I am, uh, an architect by training, um, have or I am also a partner at Mars and Partner, a large architectural firm, which does a lot of acquisition, and that's honestly where this, um, the idea of founding Zeit came from, because we always faced the big challenge of how to quickly assess a plot of land, how to accelerate your workflows, make them more efficient, and that's what we want to show you today with a whole workflow. I'll come back to that shortly. And, um, this is not made up either. We actually work exactly as we are showing it now. So, it's proven, it works, and, um, yes, I'd say we start with the case, right? Absolutely. Let's get down to business, that's what it's all about today. Yes. I just said it. Um, we are the tool for the first mile of project development, um, to work time and cost-saving. But perhaps a look back, or perhaps also, uh, how it still works sometimes, the normal way. We did it differently before. Um, we're going to look at a plot of land shortly, what happens when you get the plot of land. You, uh, first have to go into, um, yes, a CAD program. You have to collect the cadastral data somehow. You have to upload a PDF somewhere. Uh, you have to go to the geoportal. You will then create a draft, um, to see what the building mass is on such a plot. You want to know what fits on this plot. That's always the core question we get asked. Um, then the areas are calculated, then the subsidies might also be calculated, the costs are calculated, so that in the end, the project developer has to decide with the data of the architect or if you have your own architect in-house, um, then decide, okay, is it worth it at all, or is it not? That's already quite a long way. It costs money, it costs time, and perhaps it's already too late, or you have to create another concept if the first one doesn't fit quite right and the residual value, which is then the decisive value, um, is perhaps off and not high enough. Um, especially, this is the phase that now, for you as an architect, but also as a project developer in-house, only costs money, doesn't generate any money yet, because we're not that far yet. So very often, and this is the pain, um, that you know from your professional life as an architect, that you invest time in projects and properties that don't pay off in the end, and then you've sometimes sat for many days or weeks on such a process for nothing in the end, and you could have, or probably would have, liked to have known that much earlier. That's what I always take away from my conversations in everyday life, this increase in efficiency, especially in this first phase. Is it worth investing personnel and economic resources here, or to quickly recognize at the end that it's not worth it, in order to conserve these resources, right? That was ultimately what you also found. Absolutely. Absolutely. That's how it is. And a lot of money is really burned there. You also work, uh, a lot for nothing, especially in the current times, I think it's particularly bad. Um, and exactly, we want to enable you to make these decisions more efficiently and faster. For that, we have, um, built Zeit and will continue to optimize it, logically. And we want to present a workflow today, how we, uh, work here at Zeit, but also in architectural firms, or in the architectural firms of our partners, and how our customers work with it, where we know that it, um, really works well. And perhaps Annika, could you tell us again what we have on the platform now, and then we'll get started. Very quickly for those who, uh, perhaps haven't looked at the application themselves yet, who haven't seen it in use. We have fundamentally combined only publicly available data. That was already unique in itself, to bring different data formats into harmony. And we speak of a platform at Zeit. So we have a platform that includes various building blocks, and what we are best known for is this second building block on the right, the development potentials. Because we have built and trained our own AI, which is not a large language model like ChatGPT, but an AI that can read and understand urban planning and calculates densification potentials down to the square meter. Exactly, which for every plot of land decides, can't more fit on it, and where is the living space of the future located in the respective city. Of course, it's not just this development potential that is relevant, but nowadays other topics are also included, such as, above all, energy-related topics, renovation potentials have led us to expand our platform accordingly, to the point where we can now calculate renovation potentials for every residential building and create a renovation plan. And this is always very important, always considering economic viability. What is economically worthwhile at this location? In terms of development potentials, in terms of energy potentials, and in the end, we have made it searchable backwards and built the first search engine for plots of land where I can proactively search for projects and properties. That's the theory, and now we go into practice. Yes. One more thing beforehand. We don't just offer this software; we also have an expert team of architects here who can help our clients live, who can do special in-depth analyses, 3D designs, and visualizations. And we will also offer in the future, there will be another webinar on this, very new, an API interface with a developer platform where you can build software for yourself with various AI tools using our data. This whole remote working is just starting. We'll do a separate webinar on that, it's not relevant today, but we will include it. I think it's a very, very exciting, exciting area to build highly customized software. Who are our customers? The main customer group is project developers and brokers. However, we also have, um, uh, more and more banks, property owners, and planners, but the main group is brokers and project developers, and today we are focusing on the project developer case. And, um, we'll get straight to our application case. Of course, everything is fictional. Uh, Annika, let's assume we are project developers and we get a, uh, a hot tip. That's how it sometimes works. Yes, um, maybe not maybe, but it happens very often. And we are on the, it's called the Deutzer side in Cologne, the wrong, uh, Rhine side, and it's about a gas station. We honestly don't know anything more than that right now. Um, this is what it looks like, and, uh, we know that this gas station operator wants to sell, let's assume, and he wants to sell to the highest bidder. Said, effective case, and don't take everything too seriously, what we're doing, but the workflow, that's what it's about today, uh, it works pretty well, and it's about finding a design idea, a price for this plot of land, and then approaching this operator or the, uh, seller relatively quickly. That's, that's the thing, and, you know, it's a, a shell gas station. Um, and you can already see, there's a roof structure here that's quite interesting, but otherwise, it's a pretty large plot of land. To the left and right, um, there doesn't seem to be much. We'll just take a closer look at it shortly. And the special thing about this webinar is that we're going to look a bit, um, uh, yes, I'd say outside the box of Zeit, because we really want to present the entire workflow. We'll start with a property search, which we'll do with ChatGPT. You'll get the prompts from us that we use internally, and you'll get them all sent to you afterwards. It will probably be in the email. We'll do that next week, I think. Um, then, of course, we'll start with Zeit to check it. We'll look at the concept. We'll do a 3D design, we'll do a residual value calculation. We'll do that relatively quickly because today is about the workflow and not about the in-depth understanding of how to do it with Zeit. There are, uh, other webinars for that, and always new webinars. Exactly. It's important to understand that, uh, with the large language models, with the normal AI of ChatGPT, the built environment is not understood. There simply aren't enough training data for that. That's where our, um, sweet spot is. Therefore, we cover the lion's share, of course, but, um, for example, something like a visualization, which also belongs in a pitch deck, we want to impress the client or the buyer or the seller afterwards with the right concept, so visualization is part of it, and, um, we'll also do a presentation, which we'll do with Gamma, so these are all proven other software, um, technologies that are really good. We'll do all of that live now. That's why it's a bit exciting today, uh, for me at least, because I have to do two or three things at once. Trying to do it live is a bit like a cooking show by, uh, Johan Lafer and Horst Lichter. That's how I feel today, at least. Um, yes, we will definitely have some things prepared in the oven. Some will be cooked live. Uh, very briefly at this point, perhaps for temporal orientation, because we have until 11 AM today, and of course, it was necessary to prepare one or two things, because something like the ChatGPT research function also takes 15, 20 minutes or similar until it's ready. We have everything prepared. But one thing is important for me to tell you. Matthias has prepared it, but it didn't take much longer than the time we have today. With a little practice, you too can achieve the result that we will show here today at 11 AM, perhaps a little earlier in the best case, within an hour, at most an hour and a half. That's very, very important. It's not rocket science, and we want to offer this webinar today to give you a look behind the scenes and to share the prompts with you so that you can replicate it in your daily work with the help of Zeit and the other applications. That is our goal, that is our expectation of ourselves for today. Therefore, we will shorten one or two things a bit. We can go into more detail in a one-on-one conversation, but please today, really, um, the topic that it didn't take much longer in preparation, and you can achieve that too. Perhaps with our help, we'll help you with that, that's what we're here for, that's why we want to share this initial information today, but it's just as achievable for you, and that's important to me again, it's not that Matthias sat there for three days and we're doing it in 45 minutes, but it really works the same way in our daily work. Yes. So, we are in Deutz, uh, on the Deutzer side, and it's about this plot of land. We've just looked it up, Deutzer Freiheit, Deutzer Freiheit 103, 103A, I think. And, um, we know now it's a gas station. You can already see the building's appearance through these satellite points. You can also recognize this roof, and, um, we would first and foremost look at it, probably via the satellite map. What's in the surroundings at all? You can see a parking area here. I always like to look at the actual use. Um, is this a public area? Ah, that's not bad at all at first, because I already know as an architect, public area always means setback distance is not a big problem. That means, theoretically, if you were to use it up here, I already know that you can go relatively close to the property line. Next door, there seems to be residential, um, and mixed-use. So, we have a row of residential buildings coming out here. That's the situation we have here for now. And since we're live, what do we need to do first? We want to learn more about the place, right? Annika and I are not from Deutz, I don't know anything about the place. So, I'm going to do a deep research. I want to know what's going on in Deutz at all. And for that, we have prepared the following prompt. You'll get it later. It's all very, very small now, but I'll go into the most important things that are in this prompt. So, it says, for example, that we want a property and building analysis, that ChatGPT should do a property and building analysis. It should find out the legal situation. Yes, does it find a development plan? Does it find history about this property? Um, what about this gas station? Was there something else there before? Um, what are the building law requirements? What are the building technical situations? We'll also get a market analysis. We'll do that ourselves with our tool shortly, but we'll still get it in parallel. Traffic law situation, also always interesting. Are there any plans? Yes, everything that is public, council resolutions will be found, and of course, the topic of exposé creation is also included. But we'll come back to that separately shortly. This prompt is pre-made, it's always the same for us. We only enter the address here, and, um, then, if I already have a specific idea for a use, then I also put a use here, if I know that only this one is possible. Um, then it's entered here accordingly. But we know that it's currently an industrial and commercial, industrial and commercial area, and honestly, we want to develop housing there because we know that's what works best. Um, we don't know yet if that's really possible, but I'll just let ChatGPT run now. To do that, I go into my ChatGPT, um, and basically put the prompt that I just showed you in there. I'll do that now. I'll do that here next to me. It's the same prompt, only the address is in there, and I click, this is important here, deep research. So, that means I send it off. Same prompt. Now, ChatGPT asks me if there are any specific usage scenarios, so it understands the prompt, asks about usage scenarios, etc. And I always say, um, just, because I don't want to influence it in any way, just do everything as you see fit and send it off. So. And in the meantime, we'll put that in the oven, it's simmering, and, um, it's starting now. Yes, so it has the address and, uh, considers all of that, and the research starts. This usually takes 15 to 20 minutes. In the meantime, we'll do a few other things. We'll shorten it shortly because it's simply taking too long in real-time, but we'll look back at the end to see what it has actually found. Um, but we'll continue with our case now. So, research has been initiated, and now, very briefly, you have of course also worked with ChatGPT and entered prompts and worked with the results. The prompt that we will provide you with is, I repeat, shaped by the daily work at Zeit and maximally optimized. We will send it to you, and you can then go through it and process it further and use it for yourself. Yes. Um, what you also have to do now, we do that manually as well, but I can assure you. This is one of the biggest customers. What about the development plan? Yes, of course, it's important. Development plan, land use plan. We are in the process of integrating it. However, it is very, very complicated because we are dependent on external data. We also have cities where it has already been integrated. If the data is in X-Plan format, it's easier for us. We are working on it. Right now, it still has to be done manually. This is done via the city's geoportal. Um, and we also did that once. I looked at the development plan, and you can already see that it's a relatively old thing. Um, still before 1960, I believe. It is currently still valid. So, we are up here at the corner. You can see that the entire development is already completely different. Uh, and with that, for me as a, yes, professional, it's practically written off, and a new development plan is already being drawn up. And, um, as an architectural firm, Maß und Partner, we are set up to say that we very often build against the development plan if it's not very new, accordingly, using section 34. That means we look at what's actually there, how high the buildings are, what the utilization is, and work with that, and Zeit was also programmed to provide the right answers for exactly that. That's important to understand. Um, there are so many old development plans that can now be overridden by the, perhaps upcoming, building boom. The chance is very, very good that you can achieve that. Um, therefore, section 34 is a very, very important component for me, um, yes, of, um, yes, building law, and perhaps even more important. So, um, but we want to look again at how the surroundings are. So, I know there's a development plan, it's old, there's a resolution to draw it up, I can't look into it quickly, it's not public. Um, we want to build housing there. Yes, that's simply our vision. Um, we don't want to create a commercial plot of land. That simply doesn't make sense. We see commercial and residential coming from the, uh, line here. But we want to look at the market now. Yes, while ChatGPT is still researching, we'll take a look at the platform with Zeit, let's look at the plot of land more closely. Now you can see exactly this issue that you have in Deutz. Yes, Cologne has high rents. Here, it suddenly becomes significantly lower. Um, comparable properties are selected by us to see exactly what the prices are. They will also be automatically inserted shortly. We see that, and let's take a look at who lives there. So, if we want to build something, what's the right case? So, we see most people live here for rent. Um, houses and apartments are about fifty-fifty. The market is picking up, well, very slowly again, I would say. Most buildings are older. That means new construction will, I think, go over quite well at first. That's always important for demographic positioning. Age groups 30 to 45, that means we're here in, uh, perhaps more of a family area. Um, middle income, high level of education, uh, perhaps also single or two-person households. Then I look at, um, how is the, um, how is the neighborhood? Is it loud? Yes, it's quite loud, honestly, here. Sound insulation plays a role. How is the location? Shopping is good, of course, education is good, we have a university right around the corner. Uh, I see it up here again, right? That's why these single-person households that appeared here, this proximity to the university, that will be important again. The concept of the gas station. Um, then we have leisure here. Yes, there's a park nearby, but this is, we're in the middle of Deutz, honestly, not, not bad. Uh, good. We also have very good accessibility, you probably don't need to look much. Um, we have a stop right in front of the door. So, um, that's definitely a good, a good location for now. We also have a bit of competition to the left and right, of course. Something is being built, all residential, it seems. So, I've already got a slight overview, um, of the, of the location. Let's look here. Ah, he's definitely still, uh, there. And, um, we've already had a feasibility study running in the background. Yes, let's pretend ChatGPT is finished. We'll look at it again later to see what it says, and I'll show you what it looks like when it's finished. Yes, we've now used our software to look at housing, what does it look like, what's the market situation? ChatGPT does the same. There are eleven pages of text here, and I'll give you a brief overview of what's, um, yes, what's the topic here in what was found. Yes, we know now that the development plan is, honestly, initially, uh, obsolete, I would boldly claim. And let's look at the research. What comes out, or what has ChatGPT found? Two, two important things I want to mention. Oh dear, it seems that on this plot of land, this gas station is a listed monument. It has been classified as a monument since 1982. The problem is this roof, which we've already seen, and that's, of course, a considerable problem that we have now, Annika. Yes, that's not something we like to hear, is it? And I have, we have rebuilt the Dom in Cologne, so that's definitely not something to be trifled with with the authorities, I would say. Otherwise, um, planning law is invoked here, that it is currently still commercial, um. Uh, infrastructure and traffic are looked at, the surroundings, the opportunities and risks, and what is also included in the prompt. Usage proposal for, uh, precisely this, um, yes, area. And we see here variant A, which is always the highest ranked, student housing and campus services, so, uh, yes, apartments for students. Why the campus nearby? We've also just seen it. ChatGPT has also seen it. So, that helps to verify it. There are two other variants. Hotel, boarding house, and office, of course, but we want to, honestly, we want to approach this with the idea of an office. Okay, now we know what you mean, right? With the micro-apartment. Yes, exactly. Sorry, right. So, now we actually know, um, that there seems to be a problem on this plot of land with the, uh, with this monument. And what we do now is, of course, to look at what building potential we have on it. And, um, I'll do that as follows. I'll turn on the surroundings here, or we'll start again with, um, with a different view, because we know that not only this gas station plot belongs to it, but more plots belong to it. That means we now have to combine plots, and we'll just do that. And we do that very simply by clicking on "Combine Plots" here at the top. And you can now see that I can select a second plot here. I can select another plot, and the building potential is updated live here at the top for this, for this plot, for these three, four plots. Um, now we don't want to keep the gas station. We know there's an issue with the, uh, with the roof, but the gas station itself is not the problem. That means I'll accept the selection shortly, but first I'll tear it down again. That's super well programmed by our, um, uh, tech department. I just take the buildings here and build them back, so to speak, and you see the calculation of the building potential being re-initiated live here at the top. When I've done that. The plot is now cleared for the AI, I click on "Accept Selection." A new function for those who are watching and say, I already use Zeit, I already know Zeit. This function has been added recently because it was also one of the biggest wishes, to be able to combine cadastral parcels to truly represent actual projects. We always looked at one cadastral parcel before, now the combination possibility is there, and also this live calculation of the potential, including demolition, so that I can remove the buildings, as MAS has just done. That is also new. I just wanted to add that briefly at this point because it will be important again shortly, uh, what our AI has calculated exactly. Yes. So, um, then it's like this, that I now have my plot here, so to speak. I know that in the middle here is this roof, and, um, we now want to know what the building potential is. I click on it here at the top. Zeit immediately tells me the building potential, but it's currently an industrial and commercial area, and we want to rezone it for housing. Yes, we know there's housing nearby, it's coming down there. So, for me, as a developer, a bold, um, decision, we want to offer housing there, and now I just click on "Residential Area" here at the top. And now this happens. The AI now retrieves the data from the surroundings and calculates, based on the type of use we just clicked, the number of full storeys, the GZ, the GFZ, the BGF, and so on. And let's look, it says 6800, 1350 square meters of ground area. That's what the AI calculates, and I'll now turn on the surroundings here. Zack, the surroundings are now being switched on, so to speak, right? Exactly. And now we have the issue with the roof. We know that this roof is in the middle here, this is this gable roof. Now we have to make sure that the, uh, building that we want to draw here works around this roof, perhaps. That would be my idea as an architect. I believe that the roof cannot be demolished, but, um, we will keep it. My idea is therefore to build a building that incorporates this roof in the middle. Yes, that's the, uh, the take that I'm, uh, uh, taking here. Now we have to see if this volume of 6800, which the AI previously stated, works here. And for that, I'll go to a completely new tool. It's called the measuring tool here at the top. And I use this measuring tool. We are on a public road. So, we are allowed to go right up to the street here in front. Um, I am also allowed to go over my plots here. I'll just do something like this here. We are also on a public area. I am also allowed to go right up to the property lines there, and I am also allowed to go right up to the street here in front. Then I also see the distance here. So, I see about 10 meters right now. And then I go diagonally in here, go around here in front, and spare this roof, so to speak, and then close the building here accordingly. So, then I already see the area here directly. I
I see the meter display, so 10 m, 10 m. Um, a hallway definitely fits in there. Probably two access cores. I can rotate the whole thing here again and then I'll just say 3D and then a building will be erected, so to speak. This is new, you haven't seen it like this before. And this building, uh, fits in here, so to speak, into the Yes, I'm perhaps a bit too close to the building here, honestly, those must be setback areas. Uh, I need to move it a bit. Um, and I see now, and this is the, the big clue. I'll make it a bit bigger for a moment so you can see it better. While I'm drawing here, I see the GZ and GFZ calculation below, I'm moving it again live. Yes, I see the masses, I see the large floor area that is being created here. So, I can also do this with multiple buildings. We have one here now, and I can now simply estimate the height that we were told beforehand by the AI. What do we have here? Uh, five full stories, 19 m, 1350 m² ground area, 6800 m² GFA. Then I see here, um, from the height 17 m, we can even go a bit higher. Um, ground area 1500, a bit too much, honestly, not so badly estimated. The roof is free. Um, and I get 7700 m². So we are even a bit below that. Could even allow a bit more, actually. Honestly, I've been a bit sloppy with the setback area just now. Um, so I would go in with 7000 m² here, I would claim, you can definitely get that onto this plot. Uh, whether it's architecturally a great success, I don't know yet, but it's already quite important that we know this building, we now know it fits, we have a GFA, and with that, you can proceed accordingly. Very briefly, due to the question that also came up, this is not the so-called expert analysis, this is software Site, you do this independently, you can do it on any plot, with any of your projects. Um, the expert analysis, these services that we offer in addition. These are manual services from our architects, which usually come in at a later stage of the process. What we are doing now is really meant to simulate. You receive this offer. The gas station is not easy. We have an issue with monument protection regarding the roof. It's not easy. It's a challenge. That's why this project might be attractive to you, because it can perhaps be purchased economically, um, yes, relatively cheaply, because this challenge comes with the plot. And yet you want to check if this could be something for you. And we are now really on this first mile, as Matthias called it at the beginning, very early in the process. We are still in your internal processes. Are we going for it? Is this a relevant object for us or not? The expert analyses from our architects usually come at a somewhat later point. What Matthias is doing now is what you can use with the software Site every day and to its full extent. Exactly, this is not a hack, it works just like that. You can also do this with the software. Um, and this is not an expert service. Exactly right. Um, so what we know now, we have been given the construction potential from time here at the top. That fits, we can get it on there architecturally too. Yes, I would say we can definitely approach the city with this building. And, um, but I want to get something else out now, because we now know, this is the building, this is how much mass it is. Now we want to know, what is the price we can set for it? And for that, I'll leave the 3D model there, I've just opened the window again and, um, I'll go to accept selection again here, because we now want to know what price we have for it, and for that we simply go into this project calculator. You already know it if you've attended a webinar with us before. With this, we calculate a purchase price, the residual value. And now, for us, only the new construction topic is interesting, because we are demolishing anyway. And we have just seen, 7000 m² gross floor area we have, um, I just forgot to include the change of use here, but 7000 m² um square meters GFA we have seen, fit onto this plot. We said, 1350 m² ground area we have. Um, the AI now automatically calculates whether we need an underground garage or not. We simply take the recommendation and yes, oh wonder, we need an underground garage. Yes, so the AI even says how many apartments we, um, need here, or probably need. Now we have 70 m². If I say, we make it student housing, I would say, I'm at around 35 or something like that. Uh, this is also a very, very relevant economic lever later on, isn't it, especially in the discussion with the stakeholders of the city and the respective authorities. Creating housing for students here. That has also been shown by the ChatGPT research, and it will of course also be a somewhat larger topic shortly. Perhaps also regarding this calculation table, our residual value calculation, it can be overwhelming when you see it for the first time, please don't be afraid of it, we will go into detail. We'll fly through it once to get to our goal as quickly as possible, to be able to generate this pitch at the end. We'll go into detail on that in further discussions. Exactly. It's about the workflow. That's important. Um, now we are already instructed to adjust the prices, because we have made the apartments smaller. Everyone knows, smaller apartments, higher prices. We'll just take the recommendation. Um, the parking spaces adjust themselves. We have a preview here for the rental space that will be created there. Here above, acquisition costs are included, construction costs. We get a suggestion for the rent. We have subsidies, there are none for new construction here at the moment. We have a purchase price for or a selling price for the square meter of living space. Below here is our return. Annika and I, we say 10% is set as standard for construction. You can set it as you wish. Um, I don't know if ten is enough for us. It could be tight. We'll take 15% as an example, so that something is left over. Um, so we come to 5.5 million residual value. You can adjust all of this as you wish. You can enter your construction price there, if you have system construction, whatever. Everything can be adjusted. It's always just a suggestion there. You can reorder it for yourself as a standard. The important thing is, we now know, okay, there seems to be, um, um, 5.5 million in play. If we dare to do it and say, we look two years into the future, go to €6800. So not what we recommend, but even a bit more. Then, of course, we have a huge lever here, um, with the residual value. So you can play with that. This is daily business, you know that. Um, it's a normal project developer calculation in here. Um, cost calculation, um, always pre-filled. Very, very simple. So I know, construction volume is going up. I now have information about the plot price. We are over 5.5 to 6 million, I would say very, very roughly. And, um, now it continues. We naturally want to, um, somehow, from all this data we have, which we can also export, we can print everything, what we have seen here, we naturally have to, um, somehow generate visualizations. It's part of it for us. And, um, perhaps very briefly here again, a small dramatic pause at this point. So, what have we actually achieved so far, or rather, you also partly in advance, you worked with ChatGPT, did the research, we have the information about the location, we checked everything in Site, we also created a PDF with all this information. What Matthias is now showing live in application, you can download as a so-called Site Report as a PDF to work with it further. It can be processed internally and externally, of course. The data is available outside of Site in PDF form, and we will use that again later. But what we have achieved so far is within, yes, if you do it regularly, not even 30 minutes, the location analysis. We have looked at an economic assessment, we have looked at the construction masses, we have the legal assessment, and we already know that what initially perhaps didn't look so attractive, this old gas station, can still be profitable under certain conditions that we have checked, and are now actually in a position to say, yes, we will continue with it, we want this object, and it's not the only one that lands on your desk. There are ideally 10, 15, 20 plots per day, per week, depending on their size, that you can check in this way and quickly sort out. And now you can say, okay, this seems to pay off, this seems to be worthwhile, it corresponds to our plan, and now I want to present it internally. In the next Jof Fix, next Monday, I will be asked, what do we currently have on the table for acquisition? What looks promising? And then I don't just need it for myself, so not just you, who see the plot and know what's possible, but you also need the information to be accessible to others. And that's where we come to this next point, this, we always say at Site, we are all visual creatures, we need visualizations, we need a vision on paper that someone else can understand, who hasn't dealt with it as intensively as you have done in advance. And that would then be the next step, and perhaps the transition for you, Matthias, just again with the approach of what we have actually achieved in a few minutes, and where we already stand, and how long it would have taken. Open question to you. Uh, you can answer it yourself, how long would it have taken you otherwise, um, yes, to get to this point? That is the challenge we want to take on, that we are much, much more efficient with the, yes, data we provide, but also in combination with other systems. Absolutely. Exactly. And for that, we also have a special prompt. And, um, I'll scroll down a bit here. You will also receive this prompt. It generates an urban development model, because we also want to talk to the city here. Yes, a development plan is being drawn up, an exciting plot, a good corner, we need to act a bit here, and we're not creating a photorealistic model at first, we'll do that in a moment too, but first we're creating an urban development model. And it's very simple, you just take a screenshot from Site. It can be like this, there's also a screenshot function. I'll just do this whole thing and put it into Gemini. So, can you say what Gemini is, what it does, and Gemini is Google's AI, the license costs. So maybe you can say something about that too. Exactly. Gemini is Google's AI. It's currently using a really great tool, namely Nano Banana. This is activated by clicking on Image at the bottom. But you can also do it with ChatGPT. That also works very well, and I hope it works well now. I will now incorporate this prompt, which says to generate a quasi urban development model photo. You will get the prompt later, and I'll just send it off. Let's see what happens. It doesn't take that long, we can probably just watch it here now. Um, it's now pulling from the data we just entered into Site. Um, and, um, we'll have to say again briefly that it should create the image. Sometimes it takes a while, um, until it's generated. And very briefly, so, and Gemini is being used here in the free version, right? Um, we have a Google account, you need a Google account, but you can definitely do it for free with ChatGPT too. It's very, very similar. You can see here how this image looks. Uh, it's pulling the number of stories from the small height window that we included here. That's also quite interesting again. And honestly, uh, an urban development model, you can look at it a bit. The prompt says that it should represent our new construction, our new construction as a wooden model. Um, and this could already be a very simple presentation, it's not too over the top. Um, but, uh, it's of course something different than if I just take a screenshot from the software. But I can already tell you, we are working on integrating it directly into the software. It's just a simple workflow right now. It's really nothing more. You just saw it. One prompt, the image, zap, done, efficient work. Um, I'll go back to the presentation, because, um, we've of course done a bit more here. Um, that was my preparation. This is what it looks like here. I also said, add a few more trees. That's also done with Gemini. And of course, I can also do the same photorealistically. Yes, it's just a different prompt. And now I have to convince our, um, yes, stakeholder again, and for that I simply said, let's do a photorealistic prompt, and please show the roof. But I've switched back to the points on the map. I said, please show the roof as well, and then it looks like this. Yes, it's just a different prompt. I said, give me a wooden facade, a student dormitory on the roof, a running track and a café, and the clue here is that this roof is already built in. I have to position myself somehow. Um, Annika and I, we have to somehow see that the gas station seller somehow pays attention to us or keeps us in mind. And I can naturally do the same for a view from the street area. Yes, I can simply go down from Site to the street, there's just this purple model. It pulls the heights from the screenshot again, and then here's this gas station roof. Okay, honestly, it's propped up here, but I didn't prompt that out. There's now a café downstairs, so to speak. I wrote, make it a café on the ground floor, done. So, you don't need to do any fancy facade designs at this point, because we know the height, we know there's a GFA calculation behind it, that's enough. And, um, you will also receive the photorealistic prompt accordingly. It was just asked, exactly, how to achieve that look, you will get it afterwards. It's also a bit of a work in progress, of course, and sometimes you just have to try things out, but we are happy to help and, as I said, share the content. And you said earlier again, um, at this point it's not necessary to use other systems. I think that's very important to say again, because at the moment we've only worked with ChatGPT, Site, and Gemini, so really with as little software as possible, it's not necessary to create this model to transfer it to CAD systems or similar. Which otherwise costs you architects a lot of time, and you as project developers in-house. Um, and again, very briefly in terms of time, we are very early in the process here. It's not about creating a finished model yet, but about challenging and creating a vision with which you can then proceed to further discussions. You can also read that from the questions. Um, perhaps again, as I said, a tool for the first mile, a very early phase, to quickly reach a result, to quickly decide, is this something for us, do we want to bid on it, do we want to go into further discussions, or do we leave it alone, and that very efficiently. Exactly. And the numbers, yes, that's what Site is for, to understand the environment. We are not building this 3D model in an empty environment on white paper with a CAD tool, reloading everything again, no, we are doing it inside, in the real world. Um, you basically build this building, you only need a screenshot, and the rest of the emotion, um, can now arise here. And Annika, you just said it, now we naturally have to somehow, from all this data we have generated, we have the deep research, these Site pages that you just saw, we have the report from Site. You can simply click everything together. The, um, the plot, the plot report, maybe we can show what it looks like for a moment. Yes, gladly. This report function, for example, is also included in the Basic package. You don't need the Pro package for that. The residual value method, location analysis via Price, that's the Pro package. This means that what we have shown today, in its entirety, is covered by the Pro package. But for example, for this report function. Um, exactly Matthias, you're sharing it now, you can also access this in the Basic package. Exactly. And this is now exactly this, these are the hard facts, these are the plot data, PV potential is also included. That's not really relevant for this plot, honestly, although perhaps for the roof, but I think technically that would also be difficult. You have an environmental analysis, the AI potential is of course included here again, aerial photo, we even have here for Diger, just so it's in there, um, a renovation plan for the gas station, and so you can see what it looks like. It's completely irrelevant here, we're demolishing it, but we can do it for any existing building, if you want to densify, it would also be included there. We have maps where you can enrich something. We have the project calculation included. So all these hard facts are already there. A few explanations and of course the dossier, this market analysis that you just saw or a while ago, that is also included, of course. All these hard facts plus the feasibility study, which we have now created, so to speak. We also created the image, and now it's about how we can put all of this into a good presentation. So, I can naturally say now, well, then I'll get two working students, who, um, can incorporate it into our template, or we do it all ourselves, and that is now briefly live, and that is, we'll do a ChatGPT prompt. We first need a, let's say, structure for a presentation. It's a pitch to our boss, Annika, right? We need to somehow say, this is a good plot, say so or boss, of course. Um, and there's also a prompt for that. Um, which, um, works as follows. You simply take the whole bunch of PDFs that we just had, PDF files, and put them into ChatGPT and then say, with this prompt, um, that it should make a pitch. Um, no, for whom is it? For, um, yes, the stakeholder. It states the structure. Um, and it then extracts the most important information for this pitch from this data. And the whole thing looks like this. So, we can do it live briefly, perhaps. Ah, the feasibility study is finished by the way, and I'll do it now. I'll put the pitch in here briefly. So, a pitch is more about sharing an initial vision and not sharing all or every detail of these 60, 70, 80 pages that you've worked on beforehand, but really short and concise and, um, concise, and preferably visually supported. That's the advantage we'll have from it shortly, um, to process it, in order to be able to present it to others, either within the company or, of course, to external stakeholders. And we also share it afterwards, of course. Exactly. What have I just done? I've put in the, um, the PDFs, which you've just received or seen, so existing, feasibility study, and, um, this, um, yes, the Site Report. I'll send the prompt off. And what does ChatGPT do? It looks at all the data. It's relatively quick and develops a presentation structure from it. And, um, we'll use this presentation structure to make the pitch now. So the final presentation. Oh, it's already thinking for a longer time for a better answer. I mean, yes, not bad. Thank God we've already, um, prepared the, um, the, um, outcome from these topics. Namely, I also included in this prompt that it should come up with a project title. It extracts the, um, information from, um, yes, the PDFs we have. It looks at the residual value. It looks at the costs, what is the idea from the feasibility study, everything is now coming together and will be shown here again. We also have to fast-forward a bit here, because I've already done it. We'll look back in here in a moment to see what it looks like. Ah, it's finished. Uh, fits, right. So, it's now extracting the data here, so to speak. Exactly, here comes the project title, Arena Pavilion Deutz. Why? Because it's close to the Arena, to the Deutz Arena. Uh, preserve icon, um, enliven location, increase returns. You can change all of this a bit later. It's now extracting the most important values from our documents, and, um, yes, this text can then simply be copied, and we go into a new software that I can really only recommend. Um, it's called Gamma, with which I can create presentations. And we'll do that live now. I'll go to, um, create presentation, say, insert text, and take the text here, this is what ChatGPT just outputted, and then go down here and say, generate it from the notes, zap, next. It now understands what all this text means and, um, configures it. Gamma is also basically available in the free version. So I also use it with the free version, it works up to a certain point, then it doesn't anymore. Gamma works with credits that are eventually used up. Basically, in everyday life, you can get by with the free version at first, and Gamma also contains an AI that generates novel things like these suggestions that we'll see shortly, and you then have the option to adjust this presentation, choose different focal points, or simply rework it. But the point is that you don't have to sit there for 4 hours creating a PowerPoint slide by slide, line by line, word by word, but Gamma does that, and that's this topic of efficiency, where we can quickly get to this result based on the Site data. And exactly, Matthias, you can perhaps briefly show what I can rework there, so that I maintain it a bit in my CI. Exactly. So, the important thing is, everything is automatically divided. Um, because we've already told ChatGPT that this is page so and so, page so and so. We have a template from Site in here, and we go to generate, and then, in real-time, what we've entered as information, it's already been condensed by ChatGPT. Yes, the PDFs we have, we still have them as, um, yes, um, um, supplementary material, but now the presentation is being generated live. Um, and, um, the images, Annika, you said it, they are now generated via an AI image tool, and we'll just swap them out. It's very simple because we've already generated the images with ChatGPT, and you can see the story is now unfolding. We have market environment and demand, that's being unfolded again. All the information from the PDFs, I haven't changed anything specifically. Everything is just taken from it, the AI understands what's important and what's not, and the, um, let's say, architectural concept is also brought along accordingly. It's quite impressive every time I see what's created based on the template. It's nice to watch, isn't it? Yes, absolutely. Exactly, it continues like this. It doesn't take long, honestly. It's probably only a few pages. I think it's almost finished, and, um, I can then rework it. Of course, you can also insert your logo later. You see our logo is always included here at the bottom. We also work like this to make it fast. You don't want to lose time with any shuffling around in presentations. It's about the real important topic, namely the project, and the project needs to be pitched. And we've also prepared it here. It's called here in this, um, in the first prompt, Arena Cubes, so this student housing. Vision meets reality. I've reinserted the urban development model here. Here I love the information from Site Profit. The report is also included. Yes, if I want to look something up, then I impress with a 60-page report plus these 12 pages of presentation. What have we achieved in this short time? So, it was almost real-time. Um, from this gas station, which has a difficulty, namely this roof, we have created a really cool added value. We've built student housing out of it. The bet that you can build housing on it. Of course, we have to remove the soil and so on. Um, you might have had to invest another 2-3 thousand euros for that. Um, but we have indeed created a, yes, really very, very cool pitch that will stick in people's minds. Um, and all of this took no longer than an hour. Um, with the preliminary pleasantries, I'd say you can do it even faster. Um, and you can leave out all the unimportant things. Yes, I don't need to draw walls at this stage. And I have a basic assessment. I have, um, the target expectations, target conflicts reconciled. I have market research included. I have preliminary planning, a graphic representation. I've done a cost estimate. I have documented the events anyway, and with that, a part of the architectural work and project development performance has been very, very efficiently covered, and, um, yes, we want to bring exactly this efficiency to this first mile, and, um, I'd say, if you want to create affordable housing, you have to plan more efficiently. That starts very early, and you have a big lever in the early design phases. Um, and if you then decide that it doesn't add up, we know we can offer 6 million. Um, then not too much work was wasted. I honestly think it's a super exciting project. Um, and you can go to the city with it, you can go to the investor with it, and, um, we actually found this plot through our own search. It wasn't offered, of course, but we can also search for such special plots with Site. But that's not the topic today. Um, and we're almost at the end with this. And before we go into the last questions, perhaps a small advertisement. Um, next week, I think Tuesday, right, I'm doing it with Daniel, it's about brokers, so the other large target group of ours, and it's a lot about searching, how do I find such objects, how can I then go in with a, um, not quite as extensive pitch, how can I create added value there? Perhaps that's just a small plug. Yes, what comes next, and perhaps also what complements Matthias, perhaps you want to share the vision a bit again, you've hinted at it in the meantime. We saw this process track at the beginning and the yellow bar, the share of Site at that point. Feedback is also coming in. Thank you very much for the content today, and it's perceived as a great added value, what we've seen today, also in combination with the other tools and how efficiently you can work together. Exactly. And you've also hinted at what will come next in one place or another. Of course, our long-term goal is to extend this process track further. We work on that every day. That is also our vision, which we share with you at this moment. We've mentioned development plans, the topic of visualization, and so on. Um, and exactly, but what is also coming up is a question, namely the topic of AI hallucination. How is it dealt with, how high is the error rate? From my side, I can say, the better the input, the better the output. And that's exactly what we see as an advantage, that with this report from Site, I'm giving the AI very precise instructions on what I actually want and what the status quo is. And thereby, of course, these hallucinations are somewhat, yes, I'll put it bluntly, eliminated, because it already knows where I want to go based on the foundation that I've already checked in Site. Basically, however, it always applies, and I'm speaking for Matthias here too, the thinking person is always in front of it. AI is also a tool, the user of AI, but please always with seeing and thinking eyes and brain, and it must of course be validated and checked. And of course, there are also inaccuracies when using AI. Um, that's your responsibility to find them and to rework them accordingly. Therefore, um, yes, exactly. So, the important thing is really to understand that, of course, a large part of this report or presentation comes from Site with the data that you, um, have worked out yourself. So, no hallucination is possible there. Um, in the research, that's still perhaps the largest part, but we also did deep research there. That means it's a source-based analysis, and, um, we do this very, very often, several times a day, and the data that might not fit somewhere because of a wrong reference is vanishingly small. Um, and, um, the rest, so the visualization you've seen, you have that under control yourself, the presentation at the end, there's nothing like that. So, if anything, it's in the research, and, um, you usually read through it completely for a project like this. Sources are always provided. Um, we do it too, it works very, very well, and it will only get better. More and more work is being done to eliminate hallucinations. It's no longer a big problem, I would say, with the newer models. Um, yes, and exactly, we really want to offer this added value. That's why we're now showing other tools too. We'll probably continue to do so. It's just something that, um, yes, is very well received. Nilas Möllenkamp has also shown many other AI tools where Site is not involved. We're integrating it now to really do a workflow, and we'll probably do it like this again and again because it's very realistic, and we want you to get through this first mile quickly and be as efficient as possible. Um, I think, I think we're done, right? Exactly, so actually a perfect landing, I would say. Uh, all questions have also been answered by me in the meantime, but if you think afterwards, man, I'd like to look at this again in more detail, then please, please feel free to contact us directly. Um, you will receive the information reworked at the beginning of next week. We'll polish it up a bit and, um, yes, prepare it in a well-processable way for you, and then you'll receive it automatically, and can rework it. The recording will also be published on YouTube. Then you can gladly share it with colleagues or others in your network where you say, this was inspiring, we're happy about that, and in that sense, we wish you a nice rest of Friday, and a good start to the weekend, and look forward to all discussions with you in the future. That's how it is. Bunker, thank you Matthias for your time. Thank you Annika for your time. Very, very welcome. See you soon. Okay, then have a nice weekend, I wish you.