Transcription
And then we are now at 11 o'clock and I assume that you can all hear me well, um, and can all see me. If you have questions during the webinar, you have the option to ask these questions in the chat here. You will find it down here in the Zoom bar, um, yes, in the chat area. You can click on it and then you will, um, and then you can enter something there, enter a question, and, um, I will try to answer it either during the webinar. Based on the experience of previous webinars, um, this usually doesn't work out. I have deliberately included a bit less content today, so that I can answer your questions accordingly at the end at the latest. I will also open the chat for myself right away. Um, then I will see what might come in. Yes, and now it is one minute past 11. Um, one of the most popular questions, very briefly in advance. Um, you will receive the recording of this webinar afterwards. The recording will also be put online on YouTube. This means you have the opportunity at any time to watch the recording again. Um, but you also have the opportunity at any time, um, with various AI tools, um, to simply go to the YouTube website and then, um, either copy the link, say, yes, what did Möenk say about this and that topic, give it to me exactly. Or you have, um, the option with an AI-based browser to extract the content. Um, just by the way, maybe there will be a webinar again at some point where I will present such things in more detail. But today it's about other things, namely AI tools for the acquisition of, um, properties and objects. There are one or two very slight overlaps with the webinar we did before, so in this, yes, in this practical webinar series, I would say. Um, this means that if you attended both webinars, then there are one or two small things that you have already seen, but I will, um, of course, ensure that we always have new variants, always new tools here, and that is then also really the main part of our webinar today. Um, briefly about me, just so you know who you are dealing with. I have been part of the real estate industry for 7 or 8 years. Before I came to Side, I was at BNP Paribus as a project manager for digital strategy. or startups Propex, and checked, um, yes, which tools, which, um, which programs we can possibly use well within our company to, yes, work more efficiently, work better. Um, I have been with Side since 2022, so in a week or two, a week and a half, um, I will celebrate my third anniversary at Side, and for about 2 years now, I have been, um, yes, a keynote speaker, lecturer, and also give workshops, um, at companies. Um, all on the topic of AI in the real estate world. So, these two combined, among other things, at ADI, at CIA. ADI has another webinar tomorrow, um, for real estate agents. Um, at BDB, there is another webinar for architects. That's the day after tomorrow. At iWS, I am represented as a speaker and lecturer. What I don't see here is, um, any techy developer or anything like that, or in that direction, I didn't study that. I can't develop, I can't program. Um, but I have realized, about two years ago, that I don't need to, to, um, make a lot of what I, um, what I want to do, to make my daily business here, yes, more efficient. Um, because I can now do a lot of that simply with AI, because AI enables all of us to program things, um, to automate, or even to build websites, without us having the slightest idea how to actually do it. And, um, that impressed me a lot, and, um, that's how I actually got into this whole topic, and can now show you very concretely, artificial intelligence in the real estate industry. What can we actually do? And, um, I have, um, put together a very short agenda. Um, but we will only go through the, um, yes, the larger points of it. Um, we will deal with how we can search for properties and for land right away. Um, for this, we will look at on-market acquisition possibilities. Um, but we will also look at off-market acquisition possibilities, and there, of course, um, also, um, Side will be mentioned briefly. Um, then it's about compiling data during property due diligence. Um, I have included an example here. Um, how can we actually, yes, check data and documents that, um, either the owner gives us, if I'm an agent, or the agent gives us, if I'm a potential investor, how can we compile them to quickly extract information? Then we'll deal with data research with artificial intelligence, and then, um, and this is honestly, um, the, I believe, the most exciting thing of today. Then it's about automating acquisition processes. Originally, I had another tool in here. But then, especially with this last one, I realized that with GenSpark, complete automation is possible, um, why should I show you the other tool if it's much more complex to build and if it's, yes, um, also more complicated, takes longer, and doesn't do as much as building it simply via GenSpark. So, this is the big, big trick, I would say, that we will look at today. But, um, we'll start first, and I want to show you a tool right at the beginning, which, um, honestly, is not an AI-based tool at all, but, um, a tool that is very, very relevant, and if I had known it personally before, I would have saved myself a lot of time in my life, and I think very few of you know this tool. And it's about the so-called Instant Data Scraper. I'll show it to you once. Um, I'll show it to you here. The Instant Data Scraper is a so-called extension for Google Chrome. This means, um, I see in Google Chrome, when I click on this extensions button here, I don't think you will see what I am opening here at this point. That's why I will now share my entire screen, and then you will see what I actually mean by that. I'll make it a bit bigger here. Um, I have to stop sharing for a moment and then I'll share my entire screen. That's always a risk. Um, possibly there will be a slight mess here, but you should now see my screen. Yes, share desktop. So, that should be correct now. And this Instant Data Scraper, it's actually not intended for the real estate industry, for the real estate sector, but it's primarily intended to simply pull everything from a website in a structured form, in tabular form, that you actually want. But we can also use this in on-market real estate acquisition, yes, you can also see it as on-market. Because if I now, for example, simply go to Google Maps and now select a city. I'll go to Bonn, um, and I say I'm interested in properties in Bonn, then maybe the first thing I want to do is contact all the agents there and submit my acquisition profile, just as an example. And then I go to Bonn, enter "Immobilienmakler" (real estate agents) in the top left, click enter, and what happens now? Of course, all real estate agents in Bonn will be displayed here. And if I now scroll further down on the left side of Google Maps, you will see this slight, this bar on the side, it gets smaller and smaller as I go down, because more and more real estate agents are displayed. And I tell you honestly, I think if a real estate agent is not listed on Google Maps, yes, here, then, um, this real estate agent doesn't exist. So, I claim that all real estate agents in Germany are listed here on Google Maps in some form. And, um, now you see all these real estate agents. You could now go to their websites and then say, I'll contact them to submit my acquisition profile there. Um, but that's also much easier, and that's with the Instant Data Scraper. This means I'll click on extensions up here. Um, after I have downloaded this free Google Chrome extension, I click on Instant Data Scraper, and now this field opens here. So, and you see, everything on the left side is now marked, and this field that opens has now automatically provided me with the company names of the agencies. It has provided me with the rating, the number of reviews we had for it. You can always tell a bit from that, whether it's a very big player on the market or a new, smaller one. Um, well, if we have an average rating of 2.5, then we should assume that maybe something is wrong there, I would say. Um, then we get the addresses here. We get the opening hours, the phone numbers, and, most importantly, the websites directly. Now, of course, you can use the websites to, for example, yes, with ChatGPT or with another artificial intelligence, enter some of these websites. We can do that once too. So, so, and so. And then say, um, give me the email addresses of these companies. Companies. One must, um, so one can, when talking to AI or entering something, um, make as many spelling mistakes as one wants, I feel. Um, but, um, that looks a bit silly when you do it in a webinar. Yes. Um, and now it searches the whole thing online, searches the websites online, and will, um, give me the corresponding, um, or the corresponding, um, the corresponding email addresses for it. This means that from this list that I see here, with one click, we can do it again, so, um, I have extracted a list of all, um, real estate agents. Um, in this case, I don't know, I think about 60, 50, 60 of them. It can go even further, or there's more. If I now, for example, if I click on "Start Crawling", then, um, "no element with scrollbar found", then it doesn't work in this case, but, um, "Start Crawling". Okay, normally, honestly, this is the first time I've encountered this. Normally, it's like this, that it recognizes this scrollbar here and then automatically shows you everything that you can find there, um, if you were to scroll down yourself. In other words, it's one click, and you have all real estate agents, um, yes, displayed directly. The webinar is for real estate agents and project developers, because acquisition is the same at first. Um, therefore, another solution for this would be to enter, um, "real estate agents" and "project developers", to say, yes, show me all of them. Then, again, use the Instant Data Scraper, and now let's try it with "Start Crawling". Shame, it didn't find it either. Then let's go further down here, and do it manually. But all real estate agents who are now in this area here, so they should be displayed now. Yes, that's how it looks. And, um, with this, we now have the websites of all real estate agents. There are lists like "Listen Champion" and so on, where you can buy lists of project developers, of property owners, etc. Honestly, you don't need all of that, um, if you use this and then, um, with ChatGPT, we'll see how far it is, um, have the email addresses displayed, I wouldn't make the list too long, because if you have like 100 properties, then, um, then it takes a while for it to finish. But this way, you have everything scraped directly from websites. Um, scraping is what it's called, and, um, with this, you can then, um, immediately continue working and have a corresponding list and can then, um, use it and approach it, whether it's real estate agents with acquisition profiles or, um, you send the AI directly to the websites and tell the AI to give me the acquisition profiles of the companies if available, then you have a quick, large customer list immediately. Yes, and, um, now I'll share only, um, now I'll share only Google Chrome again. So. And make it a bit smaller, because the screen is relatively wide here, and then it's possible, um, that you might not have seen everything, because as a tool, um, you can use it for everything. Yes, there are super, super many possibilities on a website. Even if you want a list of listed buildings, they are sometimes on Wikipedia, click on the button and you have this list in a CSV file. I didn't show you that just now. You can export the whole thing directly as CSV and work with it, right? And it's free. And if we want to take this to the next level, then there's a tool. This is also not an AI yet, sorry, but then there's a tool called Apify, and, um, we'll open that now too, the Apify console for it. Apify is basically a website. I'll show you what it looks like. You log in, go to this console, and Apify, um, offers in its API store, as it's called, many different possibilities for so-called pre-runs and scrapers for the internet. This means, um, you give, you say what you want, and you get a lot of information displayed and downloaded in a list. And you can, um, of course, do this, as we just did, for example, for Google Maps. Yes, this means that what I did here is use the Google Maps scraper. I then told it, um, I'm looking, I'll show you what it looks like here. Um, you might have to look a bit here, how does it work? Honestly, it took me 15 minutes or 10 minutes. And I understood how to use it. I simply enter a location. In which area am I looking now? Now for Google Maps information, for example. In which area am I looking? Um, what am I looking for? Yes, Real Estate Agents in this case. Um, then I can specify further details, what else it should download and pull. Um, what I find very, very exciting is Company Context Enrichment. Um, this essentially means that, um, it searches the internet and the website further, the website of the agent, and finds contact details that are on the website or are on the internet, for example, for the managing director with an email address, sometimes with a phone number and mobile number. Um, simply to approach them accordingly. Um, you can also use Business Lead Enrichment. It all costs, I don't know, the world, honestly, um, to simply, um, download it. And then you get to, um, the last run I did is this one, Scraping Finish. And now you get, um, the company names. You get the Total Score that it has on Google. Because we are also simply searching for Google information, the website, category names, and the phone number. And you see here in the contact information, um, yes, further information about it. You see here further social media data is not available for all, but you have, for example, Instagram data and so on. And Lead enrichment is also available. Um, you see here Name, First Name, um, yes, Sales at Spada in Hamburg. Um, here we have the CFO, um, also with departments, and if I go a bit further here, you see it's extensive with phone numbers. Um, and, and. Yes, with this, you can get further information very, very quickly, and what we just did with the Instant Data Scraper, um, for example, across all of Germany, and it takes you 6 minutes and 30 seconds. Um, and for, yes, 662 results, I paid just under $9, um, $8 in this case. Yes, and, um, you can then download these lists as, um, as a table. Um, as, yes, an Excel table, for example. I have already tidied them up a bit so that you can see what you have, and, um, you can then use them to initiate further steps and go to the corresponding websites, um, yes, or contact them or whatever. And this applies not only to agent contacts, if you want to enter your acquisition profile, but it also applies to, um, to project developers, it applies to property owners, it applies to almost anything you can think of, and now only for Google Maps. I would like to show you another possibility for this, um, because we also have the possibility to crawl on-market properties specifically with, um, API. Um, what does that mean? In this case, we use a crawler, a scraper, um, that is available at Apify for ImmobilienScout24 ads, to give them to us in a structured table. Yes, there are tools like Immometrika, for example. But these tools have little AI included. And, um, what you can do with it, download it like this with an AI, I'll show you that now too. The first step is, um, for this, we go, um, once more to Apify, look for this crawler that we have here. It's this one. Um, it costs $ for 1000 results that you get. That's negligible, and it works like this: you just enter a link, and I have this one as an example link, and this is basically just the link to, yes, it's not the absolute correct example right now, but this is, um, a link to 163 rental apartments in Osnabrück, for example. How do you get this link? Well, that's just the one that's up here. This means you simply go to Immoscout, um, enter, yes, Osnabrück, and click on "Wohnung kaufen" (buy apartment). If you now want to scrape all apartments, for example, and then analyze them, you now have an overview. If you now say, for me, um, new apartments are uninteresting, then remove them. Then you are at 99 results. Go to "Eigentumswohnung" (condominium) or copy this link, go with it again to, um, to Apify, enter a link type, click on "Save and Start", and then this tool starts working. Here you will first see the logs while it is working. So, what is happening now? What is it looking at? Um, and you will see the output. We'll wait a moment and you'll see a bit of what's happening here. But let's go to "Board" and I'll show you the finished result, of course. Um, the finished result in this case for these apartments, um, is this one. And you get an output. It looks like this at first. It's an incredibly long, incredibly large table, um, with a lot of basic information about each individual, um, apartment or each individual apartment that you get. This doesn't help you much yet, but you can download the whole thing as a table. The table, then, has to be adjusted a bit. Ask ChatGPT or the AI, um, how you would do it, or how you should do it, or if the AI can do it for you, and then you have here, this looks a bit unstructured now, a list of all condominiums that are currently online, um, at, um, at ImmoScout. Um, and you can then use that and work with it. The point is, um, if you are now looking for condominiums, for example, that you want to acquire, then, um, many condominiums are also rented out, so that, um, with the filter you have set, so that ImmoScout always sends you the latest properties, so that they would not be displayed to you at all if you only search for unrented apartments. Now, however, you have downloaded all the data from ImmoScout that there is to know about these apartments. And you see, it's extremely extensive. So, you also get the pictures, links to the pictures, and, and you have to, um, a bit here, um, so it's confusing, of course, now as a human, right? You have to look a bit, um, how to manage it, I'll show you in a moment. Um, but you get all apartments out, and it could be that, um, that these apartments are actually rented out, but, um, in the description, for example, it says, yes, the apartment will be available on November 11th, then that would be sufficient for you. Um, and it would be an, yes, an apartment that not everyone finds directly and sees directly. Therefore, um, it is very useful to download this data like this and run it through an AI. Tell the AI, pay attention to this and that in the description, um, or, for example, forward the pictures directly to the AI and say, what, what would the, um, renovation of the whole thing cost me? That also works, and, um, then work with that, and it's relatively fast. What I did and want to show you once is, um, I took this data and put it into GenSpark, that's the tool we'll look at at the end today, um, into GenSpark, you can't see it anymore, these are older messages, and then communicated something with GenSpark. So, what you see here is always the answer from GenSpark, what it just did, and, and then I asked a question from time to time, um, as you do when dealing with AI. I'll go down a bit, so. Um, I always asked a question here. No, not here. Here, for example, and said, now include this and that. GenSpark understands the table we just saw. It understands it, and you can then, um, simply with this table. Yes, I have to do that again. No, luckily I saved it. Okay, give me the website again. So. Um, you can then simply tell GenSpark, this is the data. Please do this and that with it, or please calculate all, um, apartments according to my own, um, ideas or houses or whatever you upload here. Um, and please, um, please give me those that are most relevant, most interesting to me. And what James Spark does here, for example, is to create these mobile platforms for me, which you see here, and all of this with the data that I simply downloaded via Apify, then uploaded here, and then spoke a bit with GenSpark. Um, this here is my personal, um, personal overview, which, of course, at first is not much different from what we, um, from what we see on ImmoScout. Completely clear. But what I have done here is also to include current square meter prices, which I also had calculated by the AI, so that I can now see exactly which properties are below market, which properties are above market. Um, I also have an overview here, there is a need for renovation, um, there is no need for renovation, this is furnished, this is not furnished. All this information and details, they come from the texts, they come from the further details that we have in this table, which ImmoScout does not show you at all, or with which ImmoScout cannot filter at all. And in the end, I have an overview of these corresponding properties here. You can also build it so that you can click on it. Here, for example, it works. You see the pictures, and you have your own overview, of course, with the normal description, but we are not interested in that. We want to have it read out automatically, um, automatically, yes, AI-based. And this, this page and this overview, GenSpark builds it for you too. You can also enter how much agent commission you have. Honestly, that's already included in ImmoScout, so that the total investment is calculated, and, and, and it's very, very easy to set this up. These are a few prompts that you have to enter. You can experiment a bit and then search on-market for all properties that might be interesting and relevant to you. And that, um, works, so there are many different ways to do it. You can also simply send an AI to all the websites that we have previously extracted from the agents and say, give me an overview of all properties and then compare them with my own data. Um, these are all step 2, step 3. But all of that is possible, and all of that is possible, for example, via GenSpark. Yes. Um, we'll now leave the area of on-market acquisition. Let's go into off-market acquisition, and for this, I would like to briefly show you Side. Um, I think many of you have seen it before and know it. Side offers an overview of all development potentials, all existing data of all properties in Germany, and can then make them searchable in reverse. This means I'll just click on "Grundstücksuche" (land search) here, and the simplest variant is now to say, I'm looking for residential properties or residential building land. Um, I can either enter an area, a whole federal state, or I, um, yes, just select something where I say, that's interesting to me, and search for land, um, that has a development potential for residential use, so that it's already residential building land of at least, I'll take a bit more, 1500 square meters GFA, maximum 10,000 square meters GFA, and then, um, I'll also specify that the plot should be at least 30 x 30 meters. Um, it's always useful to do it this way. And I'll click through a bit faster here and now just say, the plot should be completely empty, and then click on "Weiter" (next). Now you see, you have 80 results, and for off-market acquisition, so, um, I look for plots and then approach them independently to, um, yes, find the owner and then, um, see if I can develop the whole plot or not. It's perfectly suited for off-market acquisition. Here you see in the cadastral, in these survey data, in these 3D points, that something used to be there. In the cadastral data, there's nothing anymore. This means it must have been demolished recently. I can imagine that something new is already happening here. You also have such examples, or such, um, situations. Here it's different. New buildings have already been constructed here. Yes, so you have to look past it a bit, um, that not all of these 80 plots shown are 100% accurate. This one here, again, seems to be free. Yes, this plot allows you, it's already residential building land and it allows you to build with 4000 square meters. And, um, if you find these plots and buildings, like this, first plots in this case, via Side, and then immediately go into acquisition, you will sometimes find plots, um, you might drive past them every day and not even know that it's residential building land. Yes. Um, well, that's also built on by now. Let's go back here, more into the inner part. Yes, here, for example, we have a plot, um, that is in the inner area, it's residential building land. Um, it probably looks like a garden area right now, but you actually have an entrance built through it, because it was planned here at some point to build on this plot as well. So, this plot is currently free. You could build on it with 1600 square meters, and, um, to go into acquisition and ask who owns this plot, um, should I or can I perhaps, um, make an offer for it? That saves you, of course, from a project developer's perspective, the agent's fee, and, um, it allows you, I believe, a completely different price negotiation for the plot. And you find these plots via, and what you can also do, and this is something I actually do privately, is to search for plots where you make owner research somewhat scalable. What does that mean? Um, let's go into an area again, like this one, for example, and say, in this very large area, you have to say, I'm not looking for free plots, but I'm looking for plots where, in addition to the existing building, in addition to the GFA, I can build at least 500 square meters of GFA additionally. And, um, these plots should also have a certain, yes, I'll take just 15 meters, a certain length and width. And, um, here I now want, the difference to the search we just had, is that we are now looking for plots that already have main buildings in existence. This means we are looking for, um, single-family houses. So, for, um, we'll also enter that in a moment. We are looking for single-family houses, um, from certain construction years. Yes, somewhere around here. We hope that there is no monument protection on it. So, um, from these construction years, we are looking for single-family houses that are on a plot that currently has a very low floor area ratio. This means that these 500 square meters of additional development potential, we can probably achieve by building a second building on the plot. And to make it a single-family house, let's say 60 to 140 square meters of footprint, the existing building should have, and 80 to 300 square meters of GFA, the existing building should have, and it should have a maximum of two full floors, then we can be very sure it's a single-family house, and we are simultaneously looking for those with poor energy efficiency class. Why? Yes, we now have 1200 results. But we have also requested a very, very large area. And these 1200 results are all plots, such as, um, let's take this one, such as this plot. We have a single-family house on it, yes? We have a slight increase, you should know that, but we have a single-family house, but the plot would offer much more possibilities in terms of building regulations. Namely, in terms of building regulations, it has the possibility of 918 square meters of GFA.
to grasp, so about 767 m² more. We could therefore build a second uh house on it. This could possibly also be a multi-family house at this location. I would be quite sure of that based on the data at least. This means we have a single-family house with the possibility of building another multi-family house on it. And you now have a list of single-family houses. With single-family houses, it's usually the case that the resident is also the owner. And uh if you now take this list here, perhaps look at each building again and look at each plot of land again and say, yes, does this really fit here or not? The plot of land here is, for example, very well uh suited, if you disregard the slope, because we simply, although even with the slope honestly, because we simply see, we already have two buildings here, there is a street, there is a street, here too, only not here yet. So we could definitely densify here. And now you get, now you have a list of many of these properties. And because it is a single-family house, I don't think an owner search is really necessary anymore. Because if you now send an individually designed letter, stating, uh, we know that on your property there are so and so many square meters. We would like a part of that, would like to build a new single-family house here and offer you, while we are building the single-family house, we will renovate your house thoroughly, because that will probably be necessary at some point with energy class H. Then you have created a win-win situation for both parties. For you, you get the land uh and for the uh for the owner and resident, because he gets a free, almost free, one would have to discuss, renovation of the building. This win-win situation, you create this for all these uh addresses and properties here. And you can, of course, then contact them all one by one with a letter, for example, and see if there is corresponding interest. And this is one way to acquire plots of land completely uh completely off-market and uh to see, yes, what uh what what comes of it. And I can tell you, it works. As I said, I do it myself privately in this way or in a comparable way. Uhm, and uh very, very good uh results come out of it. At the same time, you also have the opportunity to search for plots of land over time. Uh, for example, industrial and commercial plots of land that are located in a residential area for redevelopment. Uhm, you can do this search for all asset classes. For now, we are very much in the residential area here, but this naturally works for all asset classes. Uh, you can then directly take over a renovation plan over time and a residual value calculation. So, uhm, all of this goes very, very quickly and uh with it you can scale this off-market acquisition, I misspelled, sorry, uh at the end. Yes, uhm, that is possible with it. So, and now we go into uhm, into the next area, data extraction. Uhm, data extraction, that is something when I as a broker or when I as a project developer, existing investor am sent a property with either only an exposé or uh many, many more uh further documents with it, many, many more uh topics that need to be included, then I have to read them all first. And AI can do this extremely well. AI is very, very good at really saying, uhm, these are the, these are the documents and I will now give you exactly what you want from me, what you need, for example, to check it. Uh, example, I took this uh Immoscout exposé. This is an offered multi-family house in Düsseldorf. Yes, it is basically just saved once from IMO Scout and uh given out as a printable version. And uh I have also created further uh documents for myself. These are all completely fictitious for now, I must say. But uh I wanted to show you that you can read in many documents at once and thus extract data. This means I have created a document with craftsman and modernization invoices. As I said, fictitious, but it states exactly uh what has been done in recent years and when. Yes, uh then we have uh the transcript of a conversation with the owner, where he explains a few more things about it. So, so also unstructured data that we might need in the end. Then we have a rent statement uh that we have received here, with corresponding invoices. So these are just the example properties or these are the example documents in the end. And uhm, and now we want to have the really relevant data for us from this. And for that, we can simply go to ChatGBT and provide ChatGBT with a corresponding prompt, i.e., with what we are entering here, what ChatBT should do in the end. And this prompt, you see, is a bit more extensive. By the way, I create all prompts directly via Chat GBT. This means I say, "Create the perfect prompt for this use case," then the perfect prompt comes out. Uhm, you don't have to write all of this yourself here. I didn't write it myself here either, of course, but uh this prompt first says, you are an experienced real estate analyst and investment expert for multi-family houses. You have a deep understanding of project development, portfolio management, and real estate valuation, and analyze a series of documents for an existing property to be checked as an investment. Extract all relevant economic, technical, and legal data from the texts that are important for an existing investor or project developer to make a prioritized purchasing decision. And with such a prompt, you then only have to upload the corresponding data and documents uh, which we have here. And uh, let's upload this one here. Uhm, these are not all of the documents we saw earlier. I'll just go back to all documents very briefly and take them. Sorry, it's somehow uh not open. Yes, but that's relatively quick now. So, and now I have it. So, I'm now uploading all these fictitious documents. It might be that uh there are too many. Sorry. So, I'm now uploading all these documents that we saw earlier and uhm, wait a moment until they are uploaded and then just click on Let's go on Start. And what ChatGBT now does is it reads the documents and that was very fast. Uh, it now gives me all the data that I actually need from it in a well-founded, structured way. Year of construction, construction method, floors, and so on. And that, I think, is the first step for uh as a project developer, as a portfolio holder, to extract and analyze an exposé and all the data available with it from a data room, for example. Simply upload it, say, I need this information, please give it to me, and let's go. And this data extraction is extremely easy and fast, and you can, of course, also have the whole thing automated by the AI if you build a uhm, for example, a custom GPT. Uh, how the whole thing works, you can also see in the recording of the webinar that we have provided here, for example, uhm, you build this yourself, you upload the documents there every time and then get this overview. You can also continue to work with the overview, everything is possible, but be very, very sure that the data that is in here is all exactly correct and exactly represents what you saw in the data or what the AI saw in the data you entered. This means it's very fast. You no longer have to read it and have everything you need for your purchasing decision at a glance. Yes, and this also works with the free chat version. Just saying, yes, and then the next step would be to say, well, uh now I have all the data about the property, right? And now I also want to know uh what other information is available that is publicly available. And for that, uh we use ChatGBT again, but we use this so-called Deep Research variant, and I'll show you how the whole thing looks. Uhm, for that, we go back to ChatGB Team, open a new chat and enter a prompt. I have also prepared this one in advance. Unfortunately, it doesn't look as nicely structured as before. Uh, this prompt basically says, uhm, I want for a plot of land, so now a different example, I want for a plot of land in Cologne, Deutsch Kalerstraße, or 3 Optimus plot of land we have already dealt with in another webinar. Uhm, I want all information for this plot of land that could be relevant for me if I want to rezone the plot of land for residential development. I want the tool to research completely, output everything it knows. Uh, and then uhm, I want a complete report on it. And Deep Research is designed for this. So you click on plus, select Deep Research, and then click Start. Then you will be asked one or two more questions. Deep Research is designed to really search the entire uhm, the entire internet for all information that exists, to output it to you in a structured way, so that while the internet is being searched, sometimes 7 minutes, sometimes 25 minutes, you can do something else, and in the end you get an eleven-page, fourteen-page report uh that outputs everything that is relevant for your examination of the property, of the plot of land. You would have to answer a few more questions here now. I won't do that now. I have, of course, prepared this in advance. Uhm, and have used this prompt so that you can also see a result. You see, this is already the result and very, very extensive. It's even more than I uh uh thought. The tool took 28 minutes, looked at 19 sources, performed 112 searches, and uh current use and structural condition are indicated, even with a picture, history of use. Uh, so everything that is found online for this gas station, it was built on behalf of Deutsche Shell AG according to plans by architect Herbert Baumann. Uhm, it is super super extensive. Overall legal situation with development plan, monument protection, we have here uh planning law feasibility and and uh usability of building rights. It outputs everything that is available. Yes, and I would recommend doing this for every plot of land and every property that you look at more closely, because sometimes you find information uh about it that you wouldn't have searched for otherwise, because you wouldn't have known that this information could actually be relevant. And this Chat GBT Deep Research, I would now expand on your behalf with the use of agent mode. The agent mode, for example, searches for development plans and uh and I'll show you that in the finished result. Uh, this is here, this is the prompt. You are a professional research assistant for urban development and building regulations. Your task is to find the valid development plans for the street Am Pickenhof in Neuss. And the agent mode is basically also something like ChatGBT Deep Research, but the agent mode clicks through, you see it here, that's how it works, it clicks through the web activity, through the websites, then searches further with it. And what the agent mode can do, which Deep Research cannot, is that it can navigate completely freely on websites and click on things, understands the websites. This means that while Deep Research only outputs text information and therefore reaches its limits with websites where, for example, development plans are shown on a map, the agent mode can understand this website, clicks on the corresponding things on the website, enters uh the addresses uh in this case, and then outputs, okay, this and this development plan should apply here, because I saw that on the website where I clicked. And this clicking is what makes the difference between agent mode and Deep Research. And now we have here, I haven't entered a specific address here, but now we have here various development plans uh for this area in Neuss, yes, which I could now check. You can also enter a specific address. It doesn't work in all cases, not in 100% of cases. Therefore, I would recommend it like this. It's actually a bit better. It's a bit more manual work, but it works very, very well. Uhm, but it also saves you a lot of time searching for a development plan, and the tool also looks at further information. Should there perhaps be a new development plan or something else? All of this is also checked automatically. And the agent mode is not only relevant for this, but also for your micro-macro analyses. Uh, while you are examining these properties. These micro-macro analyses are basically something very, very similar. Here too, we go directly to JGBT, because with micro-macro analyses, it can also happen that uh that a tool has to click through Google Maps or other maps to see exactly what kind of area we are in. How is it here, for example, yes, noise maps. Yes, these are usually not texts, but maps. And uh that should uh then also be taken into account and found by the AI. And you achieve this when you enter a prompt here, for example, I'll go further up, when you enter a prompt for this location analysis with the micro-macro location for residential at this address in this example. And then we get here uh, so in the prompt we have the working method uh exactly uh named. We say we want absolute data. Uhm, we can also enter a scorecard, let's say, we'll come back to that later, so that you say, it's particularly important to me that it's very quiet, because what I want to build or the properties I want to invest in are residential properties in a quiet location. You could enter all of that, as you wish. And then the agent mode goes on the search, on the journey, gives you an executive summary. Uh, with our scorecard, we get uh 78 out of 100 points in this case uhm, and get a profile and get a micro-location. It has now checked exactly, yes, the next square, Hoffmeierplatz, is 200 to 300 m uh away. So, and through this, there is the possibility to carry out this micro-macro analysis here. And you can use this well as a broker to present it in an exposé at the end. And you can also use it well as a project developer to simply see if it fits with what you have in mind. This works really, really well. Of course, I would always look at the location on Google Maps myself briefly, but it's perfectly suited just to have these factual data. Yes. So, and now, now we come to the exciting, very exciting, very, very exciting area. uhm, namely, uh yes, developing the scorecard, which I've already talked about, uh to then fully automate real estate acquisition to some extent, or at least fully automate all checks. Uhm, if you invest yourself, then you have a purchase profile. I also had a purchase profile generated by the AI from Möllenkamp GmbH, Königsallee in Düsseldorf, with Managing Director Dipl.-Ing. Martin Möllenkamp, which could be used. I have a buy-and-hold strategy, but I also have a value-add approach. Uhm, I want to uh minimize risk and have the following property criteria uh like I search for residential and multi-family houses, residential multi-family houses with at least four units, in which locations uh I search, what purchase prices I have, occupancy rates, and so on. This is already a very, very uh yes, very extensive uh very extensive purchase profile. Uhm, if you give this purchase profile or a purchase profile to a broker, then the broker understands very well uh what you are actually looking for. And the broker can then also very, very well uh yes, give you exactly the properties he is looking for, but also only because he knows the properties exactly, because he has all the data and information about them, and because he can match them with his own intelligence, not with artificial intelligence, to what you are actually looking for, and then certain things fit, yes, 28th year net cold rent or annual uhm, cold rent etc., then these things fit, but in principle you always have certain things that are a dealbreaker for you or not a dealbreaker. And these things, you might have them in mind. Perhaps the broker also knows them gradually. Perhaps it becomes clearer with the purchase profile, because one can imagine it. But if you want to check properties using artificial intelligence, then you must, you must always think when working with AI, you are now giving the AI something, and it has absolutely no idea who you are, what language you even speak, uhm, and what you actually want from the AI. And therefore, when working with AI, it is necessary to build a scorecard and can also call it a knowledge database, where it is precisely stated how you evaluate certain locations, when we are talking about locations now, when is a micro-location very good for you? Yes, one can of course simply say, I'm looking for A-location and the micro-location should also be very good, and that has to fit. And that is then something that the broker who sends it to you describes individually or individually uhm, yes, decides for himself. And what you also decide individually for yourself. But if you want to do the whole thing via AI, then that's no longer possible. Then there is this, this emotional perspective or these, these perspectives, they no longer exist, but you have to say exactly what you mean by that. And that's why I have now developed a sample scorecard and a scorecard means that an investment is relevant for you if it reaches a certain score. You could build it like this, you can also do it differently. This is just an example. The scorecard was of course given to me by the AI. I just told it what is particularly important to me and what is less important to me. And then uh the scorecard or the AI uh built this scorecard so that I have different categories here. Uh return and cash flow gives uh a maximum of 300 points, but requires at least 180 points for the object to be relevant to me. Location and market uh yes, for 250 points, requires at least 125. uh structural integrity and technology and rental and tenants. Uhm, this is all something that is output in points here. And then we also need the exact statement. Uh, yes, how much uh gross return leads to 100 points, in this case over 5.5%. Uhm, and 60 points are given for 4 to 4.4%. And then the AI can use this basis, this point allocation, to match it very well with what is in all the data. It can also calculate the data itself if the return is not included, for example, and then make this point distribution. And if you have such a scorecard, and as I said, you can create it via AI, it all looks very extensive here. You can simply upload your purchase profile, then add a lot more information in natural language about what all this means, and then this scorecard will be output to you. If you have this scorecard, then it is practically the knowledge database, knowledge base is what it's called, uhm, to completely draw the purchasing conditions with AI and uhm, uhm, yes, to set it up for the AI so that the AI truly evaluates and thinks about the properties in your sense, and that is, I think, the best time you can invest in building this score workout for yourself, for brokers, but especially for the AI. And you should do that now, because now is the point, uhm, yes, now uh uhm, now more and more people are using AI for such checks and uhm, just do it, spend half an hour on your own purchase profile and build it together, whether for existing properties, whether for residential properties, for plots of land or for anything else. What exactly are you looking for? How exactly would you evaluate certain things in the purchasing process? What is a dealbreaker, what is not, what can be compensated for by what? Put all of that into a document. That's the first step. And the second step is then to uh optimize this purchasing process completely and have it done by AI. And for that, there is now a uhm, a tool even from ChatGBT. I'll show you that. You can find it via uhm, so it's called Chat GPT Agent Builder. Google Agent Builder uhm from JGBT from Open AI, then you'll get to this uh page of the platform. This is basically something that looks sensible right now. Uhm, but is currently still in the uh developer platform, but even as a non-developer, you quickly understand how it works and can build it yourself. Uhm, if you don't understand something, just ask Chat BT, just ask the AI, take screenshots of it, and it will give you the answer, and then you can build it yourself accordingly. And the point is, these automations that you already see built here, you can of course also build them yourself with Zapier, with Make, with N8N, these are tools for building automations. Yes, this is uhm, for someone who has never built anything with these tools, much faster and much easier. You also have to learn a bit at the beginning. Yes, what does that mean? Do it with AI or watch videos on YouTube about it. Uhm, that's very, very quick. Yes, and if I now have the start button here, then it works in principle like this. I enter the prompt that I would otherwise enter in Chatvt, simply here in these instructions. And in these uh instructions, you can of course also get the prompt from the AI again. In these instructions, there is basically the statement, uhm, I upload, I upload objects uh or documents, analyze the documents for me first and extract the data for you first. Yes, what we did once before via Chatvt, that's in here. And then uhm, it goes to the next step. And in this next step, uhm, our purchase profile and the scorecard are stored here. And it says, pay attention, you take the data that you extracted before and overlay it with this scorecard and uhm, and think with the help of the scorecard, how many points does the whole object get and uhm, is it suitable for me or not? I wanted over 650 points in this case. Uhm, and then you can tell me exactly, is the object suitable or not. That's what's in here in the end. And then we go into this area here, if the word Suitable comes out here, so it's built like that or the prompt is built like that so that this word is included. If it is suitable, if it has over 650 points, then it goes into this area, and in this area, ChatGBT then has the task of looking at the object very carefully again and thinking about what measures. Uhm, what optimization levers do I have here? With what measures can I possibly make the object suitable for me after all? Can I adjust the rents? So, to adjust the rents, it would research how high the rents are in the area. If it hasn't done that before, of course, but can I adjust the rents? Uh, if yes, uhm, how is it or could I carry out a vacancy? For example, uh if yes, how much would that cost me, so that I can agree with the uhm, with the tenants? Could I build an attic extension? Uh, can I densify here, can I rezone, and so on, and then you will be given a solution for it. And I have already done the whole thing in advance. Yes, and uhm, it looks like this. Uh, here you can't see what the whole, what CHGBT thought or did, but here we have the short diagnosis. Why is this object not suitable for uhm, for the scorecard we had? Well, we have a cash flow gap, uh we want gross or we have gross 2.8 to 3%, which is less than we want. Uh, and the energy efficiency is extremely weak. We didn't want that according to the scorecard either. And now it gives these corresponding optimization levers. We can adjust the rents. Uh, potential added value 3 to 6% of the current rent. Uh, main risks, formal errors, contradictions, energy class can reduce possible tensions, political environment. Uh, then we could include index-linked rents for new contracts. We would have a modernization package uhm, where uhm, which apparently uhm, yes, a subsidy here, corresponds to a subsidy here. Uh, we can renew the windows, light variant, add balconies uh and and yes, so very, very many possibilities to simply see if this object is interesting for us after all. And all these topics uhm, are output to me here by the AI. Uhm, and I could therefore build them myself. And now I have already announced, we are coming to a so-called super agent and uhm, yes, we have now, I will slightly exceed, I'm sorry, but uhm, this is very, very relevant and super exciting, uhm, the AI Jens Spark, this is a so-called super agent. A super agent. Uhm, the word sounds a bit silly, but a super agent is basically uhm, a mix of many, many different tools. You can say here at Jenspark, just like at Manus, for example, uhm, I want you to uh research these things, then output them to me as a table, and when you have output them as a table, then expand this table with these and these displays and announcements, and then also output the whole thing as a website and build a podcast and a video for me, and you can say all of that in one prompt here, and it will know itself which individual tools it should and must use to do all of that for you. This means that this tool is also a deep research tool, Deep Research, as we had with ChatBT earlier. Uhm, and can do many, many things at once for you. And meanwhile, this is new, uhm, I honestly only used it for the first time on the weekend uh and then immediately incorporated it into this presentation. Uhm, there is the possibility to build such a super agent yourself. This is something like a custom GPT at Chat GPT, as I said, we had that in the uh in the previous presentation. Feel free to take a look at what a custom GPT can do there. Uh, this can do this too. And the point is, it can then simply uhm, with this scorecard that I have, uh and with the information, build a uhm, a custom GPT as a super agent. I click on it once. Learns briefly. So, and now I'm building a new agent, click on create new, and now I just have to tell it what to do when something happens. So, I'm basically just telling it uhm, if you receive uhm, the information from an exposé uh or everything that I upload to you, then please match it with my purchase profile and my scorecard, uhm, which I either give you individually or which is already stored. You can do all of that, as you wish. And then I want you to immediately fill out the entire scorecard, that you immediately uhm, check what levers I actually have to uhm, to get even more out of it. And I can also tell it, then create an executive summary that you send to yourself by email and send to yourself. Jenspark can access various tools. I have of course already prepared this super agent here. I can also show it. You see, it's very, very extensive. First, the answer. Yes, I have only uploaded these uhm, yes, data in this super agent that concern the uhm, object that we have already seen for the object in Düsseldorf, and then it says, I have received these documents. Before I begin the professional purchase analysis, I need the purchase profile and the scorecard. Then I uploaded that too. I can upload all of that to Jenspark so that I don't have to upload it individually anymore. Yes, easy, no problem at all. And then uhm, it looks at these six phases of purchase analysis with the scoring system and now comes to an executive summary with the key findings, yes, one purchasing criterion was violated, the purchase price is below the minimum requirement of 2 million euros. Object does not fit the purchasing strategy. And then you see here how it used this scorecard. So the object is not relevant for now. It is just above the minimum score, but it has violated the uh purchasing criterion. Yes. Uh, but now you see here how the AI actually checked this. The KO criteria musts, we have one that is not fulfilled. Otherwise, all are fulfilled. Then we have here in category A returns and cash flows uhm, which have been checked very precisely here. So this is always the section from our scorecard. Legal consequences seen. Uhm, where we then say, here we want the return of XY. Then it says, uhm, how many points can be given here at maximum? 50 pieces. But we are in the range of a return of 2.5 to 2.9%, which is below 4%. So we give a maximum of 20 points for it.
The cash flow would even be negative here. That means there are zero points, and so it goes through every single area of the scorecard. Here also warehouse and market. He has checked everything, so everything that the agent mode also did for us beforehand, that was all checked again accordingly. Um, and this was all here, so in Jamspark this agent mode is practically already included and this public transport connection and everything you see here was automatically checked here. Then we go into category C, how is the building substance? For this, he checks the invoices, uh, the craftsman invoices and such, everything that was in there. The overall result is now, uh, we have not met the criteria in one area. We already had a NOG exclusion criterion beforehand anyway. Um, now we go into external research again. What other data, uh, is relevant for us in the macroeconomic situation? This is now phase 4. Uh, then he looks at it, he goes into the research of the rental prices for the apartments, uh, the purchase prices for the houses. So, are we here in an area, uh, that is, uh, relevant? The average industrial is much, much higher than what we have here, for example. Um, so on this basis, it would be attractive for now. Uh, then he searches for all the financial capital, the modeling. In principle, I could go through all of this individually now. Uh, in principle, um, he takes the scorecard and can research everything automatically for you. Everything. In the cash flow analysis, it now turns out that this doesn't fit here. And then he automatically does a sensitivity analysis again. So, in scenario 1, if we could increase the rents to market level, the cash flow would still be negative. If we optimistically increase the rents, the cash flow would still be negative. And if we reduce the purchase price by 10%, the cash flow would still be negative. He does all of this independently. So all these, um, possibilities that you would have to, uh, then perhaps convert the property into something relevant for you. All these possibilities are first checked automatically here. So, fillers, gap protocol, and risks. So, what do we not have statements about yet? Um, no, you could also ask the real estate agent about that again. That comes in here. Then we get a risk heatmap automatically issued. Here, uh, things are very critical. Some things are significant, and some are okay, yes, are moderate risk. Uh, and we get a market situation report again, have a decision template, yes, with which we can continue to work. And all of this, all of this you get out, just because you uploaded your scorecard, your acquisition profile, and the data about the property. By the way, everything with sources, so everything, everything is checked. Yes, and it goes incredibly fast. Jamber is incredibly good at this. He researches everything and only stops when he is really at the end and has a good, good result for you, a good result for you. And you can, of course, also continue to work with this and say, yes, you have overlooked one thing. Um, think about it, we can still do this and that, or you can upload further data and then say, now do the whole thing again. This is a perfect data room for this, uh, to check this property. And you see, I also briefly hinted at this earlier. Um, we have here, uh, I mean, I also did this once before, yes, exactly, I asked him again, which solutions could lead to me generating a positive cash flow, um, for example, eviction or something. He says purchase price reduction or optimized financing. Uh, for that, you would have to do all of this here. Um, then he says about the eviction, uh, value turnaround eviction calculated over 12 to 24 months, uh, with a termination agreement and and, um, it is very fast, very simple, and with this you can automate everything. I then built a second agent again. Uh, I would like to show you that again very briefly, uh, because with, um, with Jenspark you can deposit certain, uh, certain, yes, um, tools, such as, uh, your email, your email tool, such as, um, a Notion page, for example, where where further information is, how you actually acquire. You could also have your scorecard on your Notion page. Um, you can connect very, very many documents with Jenspark in principle. And I have now told him here, please, uh, do the following, uh, do this whole analysis that we have just seen, and then build an executive summary and send it to me automatically as an email. And I said that in the first prompt. So, if I have this in the first prompt, then I just upload all the, um, all the data. So, I have this tool, this system, I upload all the data, and then what we saw before happens. It is completely calculated once, and then, uh, I also automatically receive an email with this executive summary. Why is the building interesting and why is it not interesting? We have, uh, status is KO, category A missed. Yes, unfortunately doesn't fit. Critical gaps, solution approaches, price reductions, and next steps, which are indicated to me here. This means you get a property, get all the documents, just put them in here, get an email as an example with the perfect calculation of the whole property, with the executive summary, and you know, is it interesting for me or not. Very simple, and that is enabled in Jens Spark. And this is now an example of many. Everything that goes in this area, everything that, um, yes, everything that requires a lot of data and a lot of research and a lot of extraction, you can do with such a Jenspark user-defined super agent that you can build. And we started building it once before, it is very, very simple. I could now display the super agent. I could also create it further by simply entering what I want. Very simple. And then you have created it, pull all the information into it, and at the end, for example, receive this email. But you can also say, please access this and that table, enter all of that into the table, uh, so that you simply have an overview of all the properties that have been sent to you. Also super easy and super fast. Yes, I definitely wanted to show you this because I used it for the first time in this form last weekend, and it impressed me very much. Um, and I know a lot about AI. I am no longer easily impressed, but this is really very, very, very, very, very, very good. Um, I can only recommend it. And, um, yes, in this sense, we have reached the end of the webinar with a slight, slight overrun, I apologize, um. I would like to point out that on November 13th, you can register directly via this, yes, via this QR code. We have the next practical webinar where we will look at, uh, AI tools from the German real estate industry. So, it's less about large language models, so about ChatGPT or about Jenspark or about Manus, um, which we have covered a lot today, but it's mainly about, uh, yes, that we, uh, for example, look at Alpha Prompt, that we look at Quirepad, and indeed tools that really use AI from the ground up, from the base, to, uh, improve something in the German real estate industry, and to get an overview of what is actually available. Yes, scan the QR code, then you are, um, then you are at the registration page for it. Um, and if you still need or want my contact details, then I would actually have to, uh, show you the corresponding page for that now. I'll look that up again very briefly. I seem to have overwritten one page. I'm sorry. That was not the goal of the whole thing. Um, I'll look it up again in parallel and add it in. But otherwise, we have reached the end of this webinar, and, uh, should you have any questions about it, I really enjoy this topic. So, this whole thing was also created with an AI, with Gamma, that's what it's called. Uh, yes, well, now I've overwritten it. That also seems to have happened before. So, you can find my contact details here. You can find my contact here. If you have questions about anything, if you have an idea how you would like to build the whole thing, if you want to build an automation for it or something. Um, yes, add me, so the easiest thing is, add me on LinkedIn. Uh, I really have super, super fun with it. Nilas Möllenk is the name, actually. Nilas Niklas Niklas without a K. I have super fun with the whole topic. Uh, I am very happy about questions. I am very happy, uh, to think along and, uh, and to consider how one can build the whole thing. Um, and, um, I publish on LinkedIn sometimes more, sometimes less regularly, but I have, uh, I have built something with which I now publish more regularly, also, uh, articles on how to, uh, yes, how to, uh, uh, use AI for us. So, where things that I have shown today, uh, are then broken down again very concretely with very concrete steps, so that one can, um, so, uh, replicate it oneself. I do that too. I am completely caught up in this whole topic, I have to honestly admit. And, um, therefore, feel free to ask me, feel free to write to me, feel free to contact me, um, and of course, above all, contact me if there is, uh, interest on your side in time. Uh, then I will show you that again and we will see, uh, how you can use the whole thing. By the way, Zeit also has an interface, a so-called API. An API can be used very well to, um, yes, to connect it, for example, with Jenspark and with other tools, meaning that pulling data from Zeit directly into another tool and then working with AI is comparatively easy. Um, coincidentally, I know, I know how it works. So, we can also talk about that. Uh, yes, and now I have said a lot more here at the end, and I think now everyone has had the opportunity to, uh, scan this code accordingly. You will receive, um, a recording of this webinar sent to you quite soon, and, uh, yes, then I would say at this point, thank you very, very much for your, uh, for your participation. I have, uh, now again, okay, I have opened the questions again. Yes, there are two or three things briefly included. Um, owner data via Site is not available. Was a question. There is, uh, currently no student version of Site. That was also a question. Uh, we do not offer that at the moment. And a question for the topic Apify. Um, uh, and there is another question for the topic Apify, whether there is a subscription or whether it is per. Uh, it is different. I think you need a standard subscription, which costs something like $10 a month. Uh, and with this standard subscription, you get, uh, access to almost all tools. In some cases, you need a monthly subscription for the tools again, uh, which can then cost $ or more or less. Uh, and then you pay for the data or for the query per se, but you pay very little money for a lot of data. I was just told that the registration page for the next webinar is also, sorry, for the next webinar, the link to Jenspark. That is of course also not necessarily intended. Uh, I will have to change that again briefly. Uh, or I will send you the registration with the registration link with the, uh, with the recording that you will receive at the end. Um, and we still have, uh, two questions. Is it possible to receive the displayed prompts? Um, I can do that. But it will also take some time for you to receive the recording. That will come together. Um, I still have to transfer something. That is not yet automated for us. But I will do it. Uh, you spoke earlier about being able to contact the property owners. Uh, for that, I need the data. Exactly. I spoke about the fact that you can contact the property owners when you are looking for single-family homes, because then you are not contacting people, but the house on the property. That's why this search that I showed from Mide for single-family homes with a lot of space on the property where you can build something else, that's something, uh, you could theoretically, I'm honestly not 100% up to speed on data protection, but theoretically, you could write a letter to the house and through the house you would inevitably reach the owner, because I estimate that 85-90% of single-family homes are actually inhabited by the owner. Um, you could do that through that. Uh, what is the data basis of Zeit? And how is it ensured that it corresponds to reality? The data basis of Zeit are cadastral data, so 2D accurate building and property outlines, and 3D LOD models are placed on top of that. And these LOD models come from the state surveying offices, um, and, yes, describe the buildings with an accuracy of 10 to 15 cm for length, width, height, depth, whatever. And, uh, that placed on top of the, um, cadastral data leads to you having very precise information and very precise data for each individual, uh, property, because we take these 3D models, which are a very precise representation or a very precise 3D model for each individual property in all of Germany. And this information is then, um, translated, yes, so that we measure the facade, so that we measure the walls, the exterior walls, and we also know how many full stories or stories each building has. And that, combined, means that we can calculate how much GFA the buildings have, and from that, GRZ, GFZ, GFA of the properties or existing buildings can be calculated, and from that, again with an AI, uh, the maximum possible buildability of a property can be calculated. But we can, please feel free to contact me, then we can have a call specifically about Zeit. I will show you how it all works, and then we will see, uh, whether it might also make sense for you to integrate it into your acquisition basis. Uh, there are super many possibilities and super many solutions, and I think I am relatively creative in that regard. So, we actually still have many who are still in the call, even with a 16-minute overrun, um, thank you. Uh, I hope, uh, you enjoyed it as much as I did. Um, and I would say, see you at the next webinar. As I said, I will send you the link, and if you have any questions, please contact me. Um, thank you very much and until then, bye. Oops.