Praxis-Webinar | KI-Tools in der Immobilienwirtschaft

So, good, then I would say we'll start. Um, welcome to the webinar AI Tools in the Real Estate Industry. I am very pleased about the very large number of participants, of registrations. Um, I actually didn't expect that, or we didn't expect that, in total, in that amount. Um, so we have almost 600 registrations for this webinar and that is, I think, already, already, already special. Um, but also because the topic is, of course, extremely, extremely exciting and, uh, because we said we want to do a comprehensive overview of various AI tools that each of us can actually use in the real estate industry. And we'll do that right away. So, it will be very practice-oriented. Um, what you are seeing right now is a nice, uh, presentation. Shortly, I will be working in the browser. That means, um, yes, really in the workspace, more or less. Uh, then we won't be looking at the presentation the whole time. Um, there were already questions in advance. We will make the video of this presentation available at the end. Um, I will also, if I see that there is still time for it, um, perhaps go into one or two questions again, uh, that I see then, if you should have questions. The chat function is, uh, deactivated, but there is a question function. And if you should have questions, write your questions in there. Even if we see that it's not quite fitting time-wise anymore, um, then I will try in the next few days, depending on, uh, to answer all questions accordingly by email. Then, um, I hope that I'm not taking on too much, but, uh, yes, I will orient myself very strongly on that and we will manage it. Um, exactly. Practical webinar, what does that mean? Um, it means, uh, you don't have to participate, uh, you can primarily watch it, but, uh, everything I show with various tools, uh, I will demonstrate live here how it all works. In most cases, the tools always require a bit of processing time. That means, um, we'll take a look at how to enter this prompt, for example, so a prompt, uh, yes, the task for the AI tool, and then we'll see what happens. But in order not to have to wait 5 minutes, 10 minutes, 15 minutes, uh, we'll go directly to the results. That means I've prepared all of this to some extent, but, um, I've done it exactly as I will show you. And, um, a bit about me, uh, I've been part of the real estate industry myself for about 7 years. I was previously at BNP Paribus as a project manager for digital strategy. Uh, I've been at Site since the end of 2022. Uh, yes, about a month after I started at Site, ChatGPT was released and kind of sparked the AI hype, uh, in Germany and in the world. That means that was, uh, quite a fortunate coincidence. So. And, um, for some time now, I've also been a keynote speaker and lecturer for AI in the real estate industry, AI for real estate agents, AI for architects, including at ADI. Uh, there will be a webinar at CIA, uh, at BDB and at IRAPS. And what, uh, you don't see here in my personal introduction is, uh, the statement that I can, uh, program, am a developer, am a coder, uh, whatever you call it nowadays. Uh, I can't do any of that. I have no idea about any code. Um, but I always know who has an idea about it, and that's AI, and I also know who can build this code, execute it, and how I can adapt it. And that's with artificial intelligence and with prompts. That means, um, what you see, uh, is, um, something that, normally, only a few years ago, could only be developed with code. At least in some areas. So, I'll show you various websites, landing pages that you can, uh, build, but above all, uh, yes, real estate-related content. Um, and anyone can do that, that's the core message behind it. I am, uh, not special in that regard. I am certainly more digital for you, but I have, in principle, uh, no prior knowledge and, um, have taught myself everything because, as I said, um, if I have a question or something doesn't work, I ask AI how it works. And, uh, that's exactly the point. We have the opportunity to, um, automate everything we do, uh, to ask questions, why something doesn't work, and then move forward. And that's much, much faster than a few years ago, where we, uh, yes, had to learn things individually, etc. And I want to convey that to you very much. The focus of what we are doing today is therefore on real tools, on real application examples. We'll look at, uh, seven different tools. That is at least, these are at least the main tools that I want to show you. Two more will be added. And, um, we always look at these tools on the basis of: What does it bring us today in the real estate industry? What can we do with it? Uh, matching acquisition profiles against, uh, possible offers, uh, that we have received, is done within seconds. Um, zoning plans, according to the motto, what is actually the zoning plan here? What is the history of a plot of land? That also takes, uh, less than, uh, minutes, and all of that, right away, uh, with very concrete examples. What I will not do is, uh, share future outlooks, make promises about how AI will change our industry. Um, because it is already doing so today. So, you can do everything you can imagine today with artificial intelligence. And, uh, this all comes a bit from my own, uh, my own interest, when I'm at presentations and seminars and webinars, um, and I'm told, yes, you can build that somehow, then I don't really know how, uh, what the starting point is, or, uh, do I even have to build my own AI now? And, uh, I'm missing that a bit in the, uh, yes, in the fundamental, um, webinar, uh, presentation area, and therefore I say, um, I want to show you what you can use now. And these tools are all also extremely inexpensive, honestly. And, um, what you need to know about it is that AI is not a jack-of-all-trades. A jack-of-all-trades is, uh, my, uh, favorite term for this, for this, for this topic, actually. Um, you won't find an all-in-one solution. You won't find, uh, a tool, if you say, I am a real estate agent and I want, uh, an AI tool that does my job, then it won't be a tool with which you, uh, or that does exactly the job of a real estate agent. But, um, if you break down the work you have every day into individual process steps, uh, property analysis is, uh, yes, for example, analyzing the exposé. But it's also gathering further research on the plot of land, uh, or on the building, checking possible legal matters, uh, that, uh, relate to the plot of land or the building. Um, these are all individual process steps, and these individual process steps, you can automate them today with AI, and partly with free AI, which helps you incredibly. That means the possibilities that AI offers today are, uh, actually limitless. You just have to engage with it a little. Um, and then you can start. And the first step to engaging with it is, of course, to participate in this webinar and see, uh, what exists at all and what is possible. Um, we'll start relatively easily, I would say. Gamma, gamma.app, is the tool with which I myself have just created this presentation. So, you should be seeing my screen. Yes, um, gamma. You see it here, it's a very, very fast, individually customizable tool for creating, uh, presentations. Uh, personally, I can't really use Microsoft, Microsoft, because I'm just very slow. I can't, uh, and I honestly don't feel like it. Gamma is an AI tool. Click on generate. We are now really building a presentation from scratch. I click on generate and then simply say, uh, Gamma, listen, create a presentation for a webinar on the topic of AI Tools in the Real Estate Industry with an agenda, an introduction to the topic of AI, and with many practical tools that can already be used in the German real estate industry today and are strongly based on artificial intelligence. Um, then I click on create outline, then it finds the content itself. In this case, you can also do this with your own content, and you should actually do that too. Um, it then says, I would create these eight cards. We click on generate, and when we click on generate, then in the next step, the question will be asked, how should it all look? And, um, a quick first suggestion is made. It sometimes takes a while. So it will be with other AI tools that we'll look at shortly. Uh, no, exactly. It goes directly into the creation of this presentation and has already come up with a design for it and is looking at, uh, which layouts fit. And, uh, this presentation is now being put together. Um, what should be said is that this presentation that I'm giving you today, uh, is, of course, visually created with Gamma, but the content is completely, uh, self-generated, and this content that we are looking at today with the corresponding ideas on how to use the tools, uh, does not come from any AI, um, because it simply wasn't available on the internet yet. You would have to prompt the AI very, very precisely. So, you would have to give a very precise task so that the AI can really do something with it and really work with it. Um, to come up with these ideas that I'm giving you today. Yes, we will also put the webinar online. That means afterwards, AI will have the opportunity to, uh, yes, look at these things, and then it will probably appear in some AI, what we are doing here. But, um, currently, that's not the case. That means the content is all self-made. What you see here are AI-generated images that generally fit the style. Um, you have the option to adjust everything accordingly. If I now say, um, listen, I want this slide, for example, to be more visual, then I click on the button. Um, and this AI will probably create the slide for me with an image, uh, and build a corresponding visualization for it. And, um, I can adjust everything very, very quickly. That means, for example, here I say, I want the layout to be symbols with text, and it looks okay or good right away. Uh, and to adjust this completely in a few minutes to get a complete presentation, uh, that is very, very easy. Um, as I said, you can upload your own old or already used presentations here, uh, and continue working with them directly. But you also have the option to create websites. Uh, you also have the option to, uh, So, now I have to move myself away, sorry. You have the option to create websites. Uh, you have the option to, uh, create documents. That means you have a very long text and say, yes, it should be a bit more visually appealing, perhaps with your own logo. Uh, you just have to copy it in, press the button, and then something good comes out that you can use. And that's the first tip, which comes less from the real estate world, but presentations, I think most of us have to create from time to time. It's incredibly fast with this. I can use my own design. Uh, I can also say, I want everything in the style, for example. Uh, you see that here, how I've done it. Um, and the whole thing is even free. So, you can test Gamma completely for free. If you want to use your own style, you'll be in the 20 € per month variant, 20 € per month per person. Um, so in a company, uh, it might become a bit more expensive, but test it first. Um, and there's also the 10 € per month variant. So for 10 € a month, you save yourself hours of, uh, presentations that I, uh, have to create with Google Sheets, with, uh, with other tools. Yes, um, that's for the introduction, because my presentation, uh, that you see here, is also created with Gamma, and that's something I would definitely recommend. We'll move on to real estate-specific applications and, above all, to a Large Language Model, uh, which probably most of you know, ChatGPT. Uh, ChatGPT is, uh, the best-known, first, actually larger, uh, Large Language Model that many of you and I, uh, actually work with by default. Um, ChatGPT has a functionality that is, uh, also in conversations that I, uh, regularly have, um, that is a bit misunderstood or that very few people know. And this functionality is essentially called Deep Research. What is ChatGPT Deep Research? And for that, I'll go again, no, sorry. So, I'll go back into the presentation and open ChatGPT deep first. ChatGPT. The view or the interface that you have here, you've probably seen before. Uh, this is what we can work with. Here at the top, we can change the model we are working with. This is probably set to 4 by default for you. Uh, that is, uh, yes, ideal for most tasks, it's quite good. One or the other may have already noticed, uh, Large Language Models and especially ChatGPT, they hallucinate. That means sometimes they output content that, uh, is not true, that has been made up. Just as a tip, uh, you can deal with this very, very well, not completely, but you can make the probability very, very low that it happens, uh, by including in every prompt, um, please, yes, support your statements with, uh, with sources. So search the internet, search for sources, uh, and based on that, see what you get in the end. And, um, I would recommend that. I would also recommend, uh, working in these other variants. O4 Mini High, for example, uh, they have a certain, a certain reasoning, that's what it's called. Reasoning essentially means they think along. Um, about a year and three-quarters ago, I would say, it was like this: you enter a prompt in ChatGPT, um, in which, so a prompt, as I said, is always this task that I, that I enter somewhere. I will probably use the word more and more often later. Uh, you enter something and say, I want this and that. Um, and the prompting, so how precisely I enter it. Generally, the more context I give, the more extensively I write about something, where I want the solution for, the better the results will be. But, um, that was, I would say, with a slight, but perhaps still. That was, uh, until these newer models like, for example, Jet GVT O4 Mini High or also O3, uh, came onto the market. Um, with these models, you can actually keep the prompt shorter, say what you want, because these models develop their own prompt, to, um, so to understand the context first, and then in the second step, with this larger prompt, which is developed by themselves, which you don't see, to logically derive the, uh, solution results. And an AI thinks exactly as we do. It thinks in three directions and sees, okay, I can't get any further here. Uh, therefore, I go back and go in the other direction and think further there, and if it doesn't get any further, uh, or says that the answer is nonsense, then, uh, it would go back there. And, um, that's generally about Large Language Models, and now we want to talk primarily about ChatGPT Deep Research. And Deep Research is a function that you find here at the bottom under Tools. That means, here you click on "Perform a Deep Research". And I'll do that now in parallel directly and then read you the prompt. Um, unfortunately, I have to, I want to show you how it works, but it can take 15 minutes for something to come out. That's why I have to do this in parallel. Um, I enter a prompt and it gives me questions. So, in rare cases, there's no question, no follow-up question. Um, in most cases, you will receive three to five questions, uh, and, um, that you have to answer once. And when you have answered it, then ChatGPT Deep Deep Research starts. The whole thing is a, uh, is a tool that performs a so-called deep research. Deep research means, uh, it searches all the information that the tool has on what you asked. And all information means what is on the one hand practically in the, uh, not in the intelligence, but in what the AI knows, and on the other hand, and that's the very important thing, um, it searches the entire internet and all sources that can be found. And if I now, for example, enter, "Perform a feasibility analysis for the plots Oststraße 91 and Oststraße 93 in Düsseldorf." Um, so you've heard that, these are two vacant lots right next to each other. They are extremely central in Düsseldorf and have been completely empty for years. Um, I live in Düsseldorf myself and they've been bothering me for years, if you're in the, uh, real estate industry, then you want to, uh, see such vacant lots be developed, at least that's my feeling. Um, they have been completely empty for several years, and I want to get the complete history of the development, possible contamination, and above all, a statement on current building law, what was planned there, and information on the ownership structure of the plot. Now I've answered these questions relatively quickly, uh, so that we don't sit here too long. Um, and now the following happens. You can view the activity here, what ChatGPT Deep Research is currently researching. Where it's researching. Currently active in the, uh, German Architecture Forum, and it finds all information about this plot and sometimes it's very, very wild what's written here. Um, but in most cases, uh, it actually fits. And, um, now I'll show you, for that I have to, I just realized, copy a link. Now I'll show you, uh, so that we don't have to wait these 15 to 18 minutes, depending on the time. Um, I've already done this, uh, this analysis, and I did it three or four months ago, when ChatGPT Deep Research wasn't as good as it is today. And you get a twelve-page report in my case. I'll go through it very quickly. You see here the sources are listed for everything the AI has found. And I've marked some things, uh, so that I can show you what comes out. With this small prompt, with these two or three sentences that I just had, I get after, let's say 12 to 18 minutes, uh, a twelve-page report that tells me, among other things, that there is no parcel-specific development plan. We have to check according to 34. Um, size, layout of the, uh, plot, everything, everything is included. It also tells me that a building permit inquiry was submitted in 2017, where about two-thirds of the area was to be built. That wouldn't be told to us, of course, if it wasn't publicly available somewhere, you have to be clear about that. But, um, this information apparently is on the Green Party's website. Uh, and honestly, I could have searched for a very long time. I probably wouldn't have found it myself, and I wouldn't have searched on that website. Um, 54 apartments were supposed to be built here. That's what it tells me. We have building restrictions. Um, five protected trees were felled instead of four. Uh, yes. I didn't ask for that, but, um, it could be interesting. What I find particularly interesting is that the area was historically used as a residential and commercial property. There is no industrial prior use, so residual contamination is unlikely. Um, I got that out of a prompt where I didn't even ask about residual contamination. So, it understood itself what I need or what I want to know for this plot and could give me that accordingly. Before 1945, a normal residential house stood there, after the, uh, war, a single-story shop building, and, um, I could go on forever here. It knows itself that a subway was built there. Uh, but it also says somewhere here, I don't think I've marked it, that, um, this subway, uh, does not run directly under the house and that it is likely, um, that we can actually build an underground garage there. I think that was in the building permit inquiry. Um, but it knows that the subway was built there and therefore we assume that the, uh, area has been investigated for possible bombs from the Second World War, which we no longer have there. It tells me that in this building, a normal bourgeois house with shops on the ground floor stood. It tells me that in 1975, the city kitchen, Finecost Catering by Theo Platzer Panzer, was there. If I click on the source here, you see that it's simply a scanned newspaper from 1957 with an advertisement with this address. And, uh, that's an incredible, uh, depth that is output here in the entire, uh, yes, in the entire analysis, and you get that for every plot. So, of course, there must be sources for it, but you always find some sources, and at the latest, the check whether there is a development plan or not, ChatGPT Deep Research does that completely for you. It's an absolutely powerful tool, uh, if I want to find out something about plots or buildings. And, um, there is also information about the owner, to some extent. Um, ChatGPT generally has data protection issues. That means, if it would find the owner because they are listed somewhere online, then it probably wouldn't output it. In this case, the plot is privately owned. That, uh, can be said, in the 2000s, Oststraße belonged to Kommenwert Immobilien, and then the next thought process happened. What about Kommenwert Immobilien today? The AI asked itself and could say that in 2017, uh, it was taken over by Volovia. So. And, uh, I could continue like this, various building permit inquiries, etc. We'll leave it at that. Uh, let's check again how far it is here. You see now, it has just looked at 20 sources. You also see which sources, uh, the tool has looked at accordingly, and with that, you can continue working here. And one thing I want to show you, however. Um, the possibilities of ChatGPT are even more far-reaching. If I enter, for example, as a prompt, I've written it down, I have to read it once. Uh, seven residential units are to be built on the plot. Um, there is a building permit inquiry. I want to sell the plot, including this positive building permit, yes, these are two different things, let's assume. Find me project developers, uh, who have acquired comparable plots for comparable construction projects in comparable locations in the last four years. I only want to get those who have implemented these projects. For that, I also need the source information. The project developers should have built between 5 and 15 residential units. And so, with this, I can do a very, very precise customer search. Um, who might be interested in this plot that we have. And, uh, I am active in the Düsseldorf real estate, uh, scene, let's say. Some of those named here, I haven't heard of them, honestly, but I always get the source. What have they built? Henning Bauträger, Hennck Bauträger has built a ten-family house in Gerresheim or also 14 condominiums in Unterbach. It's a comparable location, it has checked everything, and these could now be interesting for the development of this plot. And that is, I believe, a milestone for any kind of, uh, any kind of acquisition and, uh, any kind of finding information. You don't have to research yourself anymore. Absolutely not. The AI does all of that, and the AI is excellent for it. What I have to say about that, however, is that this is only possible in the Plus version of ChatGPT, which costs you a mere 23 € per month. That is very little for what it can do. I absolutely have to recommend that to you. And now, um, a brief outlook or a brief look at other tools that also perform this deep research. Actually, every Large Language Model, uh, currently has such a deep research function. I want to briefly point out one. Grok is from XAI by Elon Musk. You can find everything about it, but, uh, Elon Musk is known for classifying data protection issues, uh, somewhat differently than, uh, than others or than we do. Um, if Grok finds online information about the owner, then Grok will output it, to my knowledge. Um, but that's just a general note. Yes, uh, you generally have very many, very many of these LMs, of these tools, which all have their particular, uh, yes, specialty. Claude, for example, is very, very good at understanding very large text documents, working with them, and telling you something about them. Perplexity is very good at researching information. Fairly, they all do that. Um, yes, but that's the core of it. Deepseek is, uh, especially interesting for this deep research function, which is a bit faster than with others. Um, Deepseek is a Chinese company, you should know that. Uh, and but it is currently completely free. And Google Gemini, we'll go into that again shortly, they also have this function. So. Um, now, however, ChatGPT Deep Research is actually not the best, uh, deep research function or the best AI for it. Um, a better one is Manus, and Manus AI. And I'll show you a concrete example now. So, Manus AI is, uh, responsible for going into the individual topics very intensively and giving you exactly what you, what you want. I have now brought an example prompt. I'll give you an overview of what Manus can do shortly, but the example prompt for Manus is: "Analyze whether the property from the exposé fits my acquisition profile and give me an investment strategy, things to watch out for, and documents to request. Tell me whether, depending on my acquisition profile, buying the property is a high-yield deal, whatever that may mean. Also, give me a short executive summary to explain the purchase of the property in a few points." And, uh, what I created, I can show it briefly here, perhaps. By the way, also with Manus, these are initially fictional, uh, property exposés. Um, yes, this is initially just text-based fictional property exposés that I can use here to input it accordingly. And, uh, this is the exposé, but this is also the mentioned acquisition profile, and the two files, uh, so if you have that in reality, then, uh, you can do it exactly the same way. That's the core message, actually. I can upload them here. They are uploaded, and I say start. And now it needs, yes, I would say about 15 minutes, which is why I've also prepared them accordingly, and it looks closely at the exposés. Um, you can also upload multiple, looks at my acquisition profile, and tells me what is possible, what is not possible. And what you see here is, uh, yes, the final, or is what Manus has done. Um, here it has individually addressed all eventualities, all details, has, uh, actually used the browser itself, to, uh, not mine, but a browser, to, uh, extract further information from the websites, and has put it all together here. And in the end, I get this promised executive summary with the sales price, with the total investment including renovation costs, which it has initially calculated without me giving it my renovation costs. If you do that, uh, then it calculates it under, uh, your renovation costs. Um, and then it says what matches the acquisition profile, uh, and what doesn't, and what the investment strategy is. But this here, what you see here, is just the executive summary. Manus is actually the overachiever in school, who always, uh, does more than you actually told him or you actually want. Manus has created all these documents for me. It has created the core data extraction, for example. Yes, well, that's still relatively clear, uh, and little, but it has analyzed or found market data, uh, also accordingly, uh, with sources, which are probably in the other file, uh, from where it got this market data. Um, and it can do that for any location. Uh, we have a financial analysis with results. Um, yes, what are the, what are the financing scenarios? We have a sensitivity analysis. So. Uh, we have a cash flow forecast, critical success factors. Uh, I get a complete analysis from it, and the more I input beforehand, uh, and the more I say, yes, pay attention to this, pay attention to that, and please use these values or research, uh, the market prices here and there, uh, the better the results we get. And it has also, this doesn't look nice, I have to say, uh, presented an investment analysis in color, with which I could work further. It has presented a cash flow forecast. It's all, honestly, a bit unasked for, as I said, the overachiever, yes. Um, and with that, you can feel like you can do anything with it. Yes, Manus also creates, uh, websites for you. Manus creates, uh, creates, uh, creates slides for you, creates videos for you. See it again. Um, and it also has access to Google. WO3, which is what you probably see a lot on social media right now. That's Google's video generation tool. With it, it's possible to make very, very good videos. You can still tell a bit that it's AI, but to make very good videos, and for that, you normally need a VPN to America, but through Manus, it's actually already possible. And, um, the whole thing is therefore an autonomous AI agent that really handles complex tasks and thinks along, and it's much better than Chatbam. You get all of this for 19 € per month, and this 19 € per month variant is sufficient for starters. Actually, I am unfortunately on the one above. I would have to downgrade. I would manage with this 19 € variant per month if I have one question a day and need one analysis a day.

or even twice a day. This is an incredibly powerful tool, which I can only recommend to you. And um, now we'll move on, we'll stay in this topic area. When we look at the next tool, the whole thing is called Jenspark. M DB is a tool that has a few more functions. So, I'll show it to you. Jensbike is a tool, it's a super agent that brings even more functionalities. It can incorporate a knowledge database. This means that if you have to upload your own purchase profiles and um and files in Manus that you need to consider, calculation tables, etc., then this is no longer the case with Genspark, because you can upload it once. So with Manus, you have to upload it individually every time. Um, with Jenspark, you can build up a so-called knowledge database and um, just upload it and then tell it to work with it. And that's simply possible. With Manus, I can link my Gmail account, my Google Drive, my calendar. I have the possibility, um, so for me, for example, it's important to know in the morning, um, what calls I actually have today and what the past conversations were about. Um, then I say: "Manus, access Jenspark, sorry, access my emails." Um, and first access my calendar. Look, which appointments I have today, then access my emails and see how the current status is in this appointment with the conversations, with the people and then um, for all where there was no real, real email correspondence, please research the companies and see how it fits with what we do, for example. Um, this is a prompt, it might take 5 minutes, and I have a perfect briefing. Jenspark does this with this super agent, and this super agent, you can already see it on the page, it can do many other things, it can create slides, um, so create slides and presentations. I would still always, um, simply because they can't adapt them that well. Therefore, I would always recommend Gamma for this. You have to do a bit more yourself initially, but it's fine. Um, in my opinion, that's the better option. And you have many other agents here. Um, the super agent, it's like Manus but faster and, in my opinion, less good in terms of content. Um, you have AI tables, and I'll show you an example of that shortly. Um, you also have this deep research. Yes, well, that's more comparable to what Manus does. Um, and in some countries, not yet in Germany, there is also an AI agent that takes over calls for you. And um, at least in America, it's the case that you don't really notice that it's an AI agent. Um, but that's a completely different topic, um, which we don't want to talk about today. This Genspark tool, in my opinion, is something that should be part of the basic equipment, I'll tell you honestly. And now I'll use it once, because we've already looked at other things with other tools. That's why I'll now use the opportunity here to develop these AI sheets and these AI tables. And, um, I've created a prompt for myself again and will then have three AI-generated, I must say. Well, sorry, where were we here? Um, 3 AI-generated, um, we have this, 3 AI-generated exposé. Um, they have just opened in parallel, so I can just show you. Um, I had three fictional properties generated and had exposé created for them. Um, and then the point is that I say, these three properties are part of the portfolio, and that's also the prompt that I give. I am a real estate investor and have been offered a multi-family house in my portfolio in North Rhine-Westphalia. It consists of these three properties. Create a table for the portfolio scan from this and include all details relevant to my investment to make the buildings comparable. Also output average values and total values of these buildings. Um, classify the content into three categories according to importance for my investment decision and structure the table accordingly. And, um, if I click on start now, it will do that first. Um, the point behind these AI sheets is not really the automatic reading into a table. Um, you see, the table is now being incorporated here, and then the information will be added step by step. That's not the, um, not the biggest point at all, but I can also work with individual columns within the table and then, at the end, tell it, I have also prepared that, of course, so that we can look at it here. So, this is what comes out. Yes, um, here I get an overview of the individual three properties. In this case, it made sense to do it in columns and not in rows, because we have more rows than columns. It has output everything for me. Um, it has given me the portfolio average. Um, I could also have a median calculated, no problem at all. Um, it has output the property data in category B and in C, the location and the features. And, um, now I would also have the possibility here to say, um, create another column on the right, um, and, um, research information about, ah, no, okay, that would be wrong. Create another row in which you store information about the district. And I can do that, and then Jenspark would automatically research on the internet and also provide me with this information. And it can also be more complex information. This is a very, very simple way. We use it, for example, um, to find out from a list of potential customers, um, what exactly they do and what their focus is. And, um, I could also, for example, say, this is the name, this is a company. Um, now search in the next column for the company's domain, and in the next column, search for, um, email addresses that you find online for these people or for people in general, not for this person. Um, research that and then give me a, a list of, um, this email address structure of first name. Last name or first initial. Last name. Um, it can always be different. And, um, then in the next column, I would say again, I would recommend doing it in individual columns. You could also do it together, but it's not quite as good. Um, and in the next column, you could then say, now insert the email address of, um, or now insert the name of this person into the email addresses. Um, yes, and then, um, you have a very quick list of email addresses. This can also apply to 200 rows or more. So, the whole thing does all of this for you automatically, researches on the internet, works with it further. And this works for real estate topics, it works for projects. Above all, um, with all the things that I'm telling and showing you now, I want to give you ideas about what is actually possible. That's the reason why we're really going into practical applications in the end. And, um, while this is still working, maybe we'll have time to go into it later. Um, but I can show you something else. Um, and that is the following. Um, the whole thing also creates very good landing pages, and I would see landing pages in this case as the possibility to analyze data in a structured, visual way. What have I uploaded here? This is a list. Honestly, I scraped it from Immobilienscout with 176 apartments that were sold as condominiums in Münster at the time, um, that were on Scout. This list, um, I probably can't show it to you quickly now. Um, the list has several, several columns, and by several I mean over 200 columns. And, um, now the prompt is, use the uploaded table, this list, and create a real estate market overview of condominiums in Münster regarding sales prices, including a classification of conditions and districts, and return it to me as an interactive website, in which I get a special, a quick overview, yes, a quick overview of the market. There are 176 apartments, each row is an apartment, and, um, here too, it needs some time. That's why I'm taking the link that I have here. So. Um, to the finished overview. Um, and the finished overview, um, or what Jenspark does, is to give me a brief overview of, um, yes, what it has fundamentally found. But then there's also this interactive website that I wanted. I click on it, open this interactive website. Whether it looks nice or not, um, is entirely up to you, no question. But, um, I have this overview. I see the apartment sizes, I see the construction years, I see the features, and I can sort everything by that. And it did that with one prompt, two sentences. And this is also a tool where I say, this is something you should definitely look at. And I myself am also in this basic version for €24.99 per month. Um, and that, um, that makes everything easier for me. So, there's also a separate Jenspark browser, which, um, accesses your browser and your own websites. Um, you have many, many more possibilities with it than I've shown you so far, but no, it's supposed to give you an impression. So, and now we come, we stay with these LLMs. I said at the beginning that LLMs are the AI tools with which we can automate everything we have, as long as there isn't an all-in-one AI solution. And I don't think there will be one in the near future. The most intelligent and best large language model currently available is Gemini. Google Gemini has an IQ of 170. Um, there are corresponding IQ tests for large language models, and it's currently free. If you're not yet caught up in this whole ChatGPT thing like I am, I would strongly recommend you to see if you let yourself be captured by Gemini, because, um, it's truly unbelievable what's possible with it. And I'll show you that with a very concrete example. By the way, an example that I've also already, um, published in a LinkedIn newsletter article. So, I've built up a LinkedIn newsletter, AI tools for the real estate industry. Feel free to look up Nilas Möllenkamp or on my page, it's linked accordingly, and, um, there I always show exactly how to use such tools. And the last article is about Google Gemini. What did I do with Google Gemini? I took a condominium from an Imos Scout advertisement. I saved all the images you see here individually, and, um, then enter a prompt. This is not exactly the prompt from the article, but a slightly simplified one. And if I upload the images to Google Gemini now, and this exposé, um, then I say in the prompt that I'm interested in the apartment for purchase as a capital investment and I want to know if it's worthwhile. I would have to, or I absolutely want to renovate and refurbish it, and I want the tool to tell me exactly, based on the square meterage, um, how big each room is, but above all, to look at the images and tell me exactly, this is possible. Um, these are the costs I have. Do I need to redo a floor? Yes, no. Um, and how much will it all cost? And, um, it's thinking now and thinking. This can take a minute or two. And, um, that's why I'm copying the link to the finished evaluation. So. And showing it to you. And, um, what it outputs now is, um, incredibly good, I must say. Yes, I get tables for each, for every single room. I say, or it tells me, if I want to remove the floor covering, it's 7 square meters, €10 per square meter, so I'm at about €70. Then I want to remove the wallpaper and it adds it all up. I get good renovation costs here. If you input your own prices for renovation or refurbishment beforehand, you'll get the whole thing calculated with your values, without you needing to know anything about the property. This means you can take an exposé from Imcout, upload it here, use the same prompt, and at the end you get the statement that you have a total investment of €63,000. Yes, that's always a buffer of 10 to 15%. Yes. Um, and then you have a statement about how expensive it is together with the purchase price. Total investment €458,000. Then it also researches the corresponding rents and says at the end, well, it's not really worth it. And otherwise, Gemini as an LM can do exactly the same as other tools, but it's much better at image analysis, and it's free. So, an absolute, absolute recommendation, um, for what we're looking at there. Then there's a tool, Prop of the Year, we became that two weeks ago in Berlin from the CIA. We're also very proud of that, which I'd like to briefly mention and show here. After that, there will be more tools. So, um, that means we'll also spend as much time on this as on the other, um, as on the other tools. It's also a bit the purpose of the webinar, and the whole thing is Site. Site is the platform in Germany where I only have to enter an address, and by entering an address, I can also zoom into Düsseldorf again, for example, and look for any building here, any, um, any plot of land, and I can simply click on it and get all the information I initially need for a possible property check. That means I'm looking for something where I see a bit of potential. So, we can take a quick look at this. Wait, this is the only thing I've prepared less than the other things, because I use it a lot, of course. Um, I enter the address and get the square footage. I get the building height. I get GRZ and GFZ. I get the number of full floors, the number of or the floor areas, the facade areas, the roof areas. Site knows all this data for every building in Germany. And what did we do then? We said, we want to, um, with these, um, with these details, you see for example GZ and building heights directly in the overview on the map, and we used all this information to build an artificial intelligence that we developed ourselves, actually. And this artificial intelligence now checks all buildings in the immediate vicinity at the push of a button, and the buildability, or the development, that we have in the existing stock, and transfers the buildability, the schema F, to the plot of land that we are currently looking at very specifically. This means that we also get the buildability potential for every plot of land in Germany through this. That's not all. We then, um, invited building age classes, and through the building age classes, we calculated the energy efficiency classes for every residential building. That's only residential, um, in all of Germany. This means that I can now, for this building, in the next step, after we've determined the energy efficiency classes, because we know so much about the building, also create a renovation plan. And creating a renovation plan, that's also only possible with the address. And now, for example, the property would be in an energy efficiency class of A+, and I immediately see that I need to install insulation, insulate the facade, insulate the basement ceiling, and all of this is associated with prices. I get an overview of the subsidies I receive for this, and I get an overview of the ancillary cost savings I have per year. Final energy savings, I can adjust everything accordingly. I can also say now, I want to insulate the facade to KW40. Yes, but it costs me €280 per square meter. Then it calculates everything accordingly. And, um, so at the end, I have a renovation plan with subsidies, with an economic analysis, with ancillary cost savings, and all of this at the push of a button. I only need the address. That's exactly the point of Site, and that's why we became Prop of the Year. By the way, we can also make all the data we have searchable in reverse. This means that in acquisition, site is also a very exciting tool. If I say, for example, in this part of Berlin, I could also just enter Berlin, in this part of Berlin I want to search for residential building plots. I'm doing this relatively quickly, where I want to generate at least 500 square meters and a maximum of 1000 square meters, oh no, let's take a bit more, so at least 1500 and a maximum of 3000 square meters of GFA in addition to the existing stock. Um, and the existing stock itself should be from these construction years, and the existing stock should be at a GRZ of a maximum of 0.37, for whatever reason, and should itself have a ground area of at least 400 square meters and a poor energy efficiency class. Then this is the finished search. Yes. We have 304 properties, and we find exactly the properties in every city that fit this search. And of course, this also works for vacant plots, for infill development plots, for rooftop extension possibilities. You find all buildings that are relevant to you and see here, for example, yes, we still have considerable potential here. This is potentially an exciting plot of land. That's what Site does. Um, we also offer architect analyses. Um, our product range is still somewhat more extensive. Um, but, um, I believe this is a milestone in our industry, and I'm very happy to use and show it every day. And another tool in the real estate world is Alpha Prompt, also a German startup in the German real estate world. And Alpha Prompt, um, basically does data room analysis and the analysis of transaction data, of rental agreements, etc., on a big scale, um, with AI, and is therefore incredibly fast. I can upload my data on the Alpha Prompt platform. In the Document Intelligence section, I click on New Extraction, would then have to create things here, we're not doing that now. Um, and then I can upload data, and I can link my SharePoint folder to it. Um, I can simply upload the files, I can upload 8000 rental agreements. Um, and then say, now analyze these for me, that will take a few minutes. Especially with 8000, with 8000 rental agreements, it will take several hours. Um, but then within a few hours, you'll have a classification of the documents. You could have set this beforehand. So, it gives you a suggestion. What kind of document is this? What does it belong to? This is how I would classify it. Then you say: "Yes, it fits or it doesn't fit." In most cases, it actually fits very, very well. And, um, then you get an overview of these properties via this Extract section, and everything that is contained in the, um, in the respective documents that you have uploaded. I'll make it more concrete. You upload these 8000 rental agreements and get an overview here of, um, yes, all rental agreements that apply to a building that has somehow 100 units. Or you get an overview of all rental agreements that are currently still valid, because there might be some among them that are no longer active. Um, or you get an overview of all rental agreements that have an automatic rent increase included, or, or analysis of this data here, Alpha Prompt does this with AI. It still takes a few hours with very large amounts of data, but, um, this is something that you can't do faster manually, and definitely not cheaper in the end. Alpha Prompt is, um, in my opinion, alongside Site, one of the AI Proptech, AI startups in Germany. Um, I would recommend you to take a closer look at it and, um, do some deeper research into it. Now, I also have, um, with an eye on the time, um, now I have a small alternative to Alpha Prompt, an alternative that is not intended to handle really large transaction volumes. Um, the whole thing is called Notebook LM, and Notebook LM is a free tool from Google. Notebook LM allows you to add as many sources as you want, as long as it's not more than 300. Yes, um, you can upload up to 300, up to 300 sources, and sources can be anything. It can be a PDF of something handwritten, it can also be a video, it can be, yes, a normal document. Um, for this example here, um, or for this example, I first uploaded all the properties or all the documents for the apartment that we looked at in Stuttgart earlier. Um, these are not the original documents, by the way, but they were created for me by Manus AI with a three-line prompt. Um, but they are all things that are sensible. I told it to also give me two craftsman invoices on the same topic. Um, so the same thing was done, but a year later. Um, so that Notebook LM now has to check, um, whether there is any, or which craftsman invoice or which condition of what the craftsman has done, is the current one, for example. But I have uploaded everything for the property here. This also applies, of course, to portfolios, if you don't exceed 3800 files. Um, you can simply upload everything here. Also for internal projects, you upload everything that you, um, that you need. And, um, then you have a searchable data room here. You can simply ask the tool all the questions you have about the property. I also recommend architects, for example, um, to create such a data room, always at the start of a project, and then always add everything accordingly, and keep these data rooms. Why would I keep the data rooms? Yes, if you have a question about the property after 5 years and you don't want to leaf through your huge main folder, you ask this question about the property, and you get the correct answer, because the tool simply searches through all your data here. And working with that, that is, in my opinion, a very, very big win in daily business. Um, and I have now, for example, said that I'm considering buying this condominium, give me the current renovation status and point out possible challenges in the purchase process for financing in the coming years that I should include in my calculations. Um, historical maintenance on the apartment that has already been carried out is shown to me. Modernization of the building. Planned and decided maintenance measures. Yes. Um, the exterior facade is to be renovated, costs €15,000. Commissioning an energy consultant, possible challenges in the purchase process. Um, it is also shown to me in terms of content, this is very important, with everything I show you, it is very, very good in terms of content. Not always 100% accurate, but it is very, very good in terms of content, and, um, you can, um, especially if you enter your own prompts, go into more depth and enter your own costs, etc., then it will become better and better. But all these tools provide good results in terms of content, and, um, as I said, you have a searchable data room that you can use completely. You can create a work summary here, FAQs about the property, a timeline, so a mind map, all at the push of a button, and you could, this is a bit of a gimmick, also create a podcast where two people discuss this property. Honestly, the podcasts are quite good. I sometimes use them for personal development. Um, if I load things in here that I don't necessarily want to read, then I create a podcast, it takes 7 minutes, and I know everything important, and it works, and it's very, very good. Yes, and this applies to all topics in the real estate sector. So, this is a foundation of mine, and so, um, that was a very quick run-through of various AI tools, um, that you can use with application cases from your area. Um, I showed Gamma at the beginning, and now at the end I'll show you Lovable. Um, these are two tools that you can use just as well for everything you do. Um, but they don't fundamentally say, yes, I can create something good for the real estate sector. Um, but I would strongly recommend you to take a very close look at Loveable in particular, because Loveable creates complete websites for me. It's a startup that was founded in Sweden in October 2024, and, um, they generated €50 million in revenue last month. So, they are very, very good, and I can create anything with it. I can say, I am a real estate agent from Munich, specializing in high-priced luxury condominiums, create a website for me. And, um, it will, um, it will output a website for me shortly. Sometimes it takes 2 minutes, sometimes it takes five. Um, and I can adapt this website. I can also implement functionalities simply by entering text, what I have already done with Loverable, um, I can show you that here, for example. Ah, no, I'll show you the complete website. I built the following tool with Loveable within an hour. You see, it looks good at first glance. Yes, that was a prompt to get it to look like that. Um, I can upload an image of a house here, a single-family house. Um, then I can tell it which renovation I want to carry out visually on this single-family house. Then I click on generate image, it will be displayed here, and it generates, there's another AI behind it, of course, with an API. But, um, it generates a corresponding exterior image for it. And that is, um, it is a milestone. You can create any website, if you say you don't have a website, um, that's no longer an excuse. If you've just placed an order for €10,000, I'd almost say, forget it, leave it, because, or try to cancel it. Yes, because what comes out here now, I can't find what I just entered, we'll have to check that in a moment. Um, it creates websites here in no time at all, and each of you can create a website. It's extremely easy, and also with various functionalities. For reasons of time, I won't show them anymore, but I will, um, include the link to the tool in the documents, um, that I created there. Then you can try it out yourself. Yes, so don't be shocked that it says, um, a website costs €1000. Just say, I'll do it myself. And believe me, after 30 minutes, you'll have a website that you're initially satisfied with. Believe me, really. And the whole thing is available for free. Um, the $25 per month version, I would recommend. It's, um, extremely good, and with it, you'll definitely get your website finished, even within a month. Yes, so, um, we've reached the end. I'd like to ask you, if you say that all of this made a lot of sense for you, um, then I hope so. Um, then I'd like to ask you to, um, fill out the survey that will come at the end. We're considering whether to do something like this more often. In this case, we've quickly gone through many tools. We can imagine that we'll look at just one tool for 30 minutes, that we'll, um, that we'll include even more tools, yes, that we'll do exactly one use case for a, um, for a customer group or for a group. All of that is possible, yes. Um, but we want to know what you find interesting, and please fill out the survey. If you want my contact details, then I have this QR code here. Um, you can scan it. You just need to enter your email address, and you'll automatically receive an email with my contact details. Um, as I said earlier, if you have questions, and you've written them down, then I'll take care of them. Um, I don't know how many there are. We'll see, um, that I answer them accordingly. You can also write me an email with questions about not only Site, but about many things. Um, and, um, in the hope that I haven't created too much work for myself, we'll manage. Um, feel free to check out what I sometimes post on LinkedIn. You can also write to me there. Um, my goal is that you can all integrate artificial intelligence into your work, into your daily work. Um, I want to give you the tools for that. Um, and with that, I think we've taken a good first step today. Um, I'd be very happy if we did this more often. Um, yes, and for that, we need your feedback. So, and then I'd say we've reached the end at this point. Thank you very, very much for your participation. As I said, we had so many registrations. Um, I'm very enthusiastic about that, and I'm very happy that you were all there.