Dr Kate Jarvis & Johnny Morris 2025 Wrap-Up: AI Agents, Enterprise ROI, and What's Next for Real Estate AI in 2026

We kind of have to talk about that study that said 95% of enterprise businesses didn't see any actual returns from buying some AI over the past couple of years. I mean, they weren't working with us. It's a starting point. The kind of agent marketing would have you believe that the future is. A brokerage and owned by CBRE interacts with an investment agent owned by a big investment manager and they interact and make a decision and they may be the president who's the only person left in these businesses since the last of the year. And nobody wants to live in that world. I think it's time that we're all honest about that. Exactly, no one wants that. LLM's as their own sort of technology just aren't good enough. And we said that last year, listen everyone, it's still true. When we look at our customer base and we look at our most successful adopters of AI, the reason they've gotten to where they are to your point is they've made it one of three strategic priorities. If you want to win big, you have to play big. Whereas we've passed the era of experimentation with AI, right? The businesses that are going to come out the other side stronger, better and in fact survive the contraction of the marketplace around AI-nated businesses over the next several years are going to be those that took risks in using their financial resources and their team's time to find the tinkerers, leverage them, rethink the way they do business. Hello, I'm Dr. Kate Jarvis, the CEO of Fifth Dimension. And I'm here today for another episode of bricks, bytes, and bold moves, the future of real estate. And here I am with my co-founder, Johnny Morris, to do a wrap-up of the year 2025 and a look ahead to 2026. First of all, for those of you who might be new to our podcast and to our business, wanted to do a brief recap of what we're all about. So we are on a mission to power the world's investment decisions in real assets. So every time you're thinking about how do I optimize this portfolio to get a few more basis points out this year? Should I buy this new asset? Is it on thesis or off thesis? Am I going to be really regretting this decision based on foreign exchange rates in the year 2030? We want you to be thinking of Fifth Dimension as the platform that's helping you get behind the scenes underneath the data data that you own and data that exists in the world in real time to help you make those decisions and drive a better future for your business. Now what does that matter? Johnny and I, of course, as being both real estate nerds and data science nerds, care deeply about making the real assets industry a better place that can move faster to embrace the future of decision making with AI at its center. But we also care about making the world a more equitable place. So behind the scenes, a larger part of our vision is also to create more equitable access to real assets. And whether that means in 2026, we're going to take the Fifth Dimension team to a Habitat for Humanity location to help build a new home. Or whether that means in practice day by day, a lot of the businesses who use Fifth Dimension are kind of scrappy, startups, 10, 30 people, not just global asset managers who also use Fifth Dimension. We're creating a more level playing field for anyone who wants to get involved in the real assets industry to be competitive if they're willing to embrace cutting edge technologies. So that's who we are and what we do. Now, one of the first things I wanted to talk about here is some of our predictions from last year, Johnny. I know it's hard to go back in the annals at this point to the year 2024. But I remember one of the things that we predicted is that everyone would be hyping AI agents, whatever the definition of that was. And that in practice, probably few businesses, if any, would deliver on the hype of the agent. What do you think has that played out? Is that true? I must admit, I cheated somewhat and did check in on what our predictions were. I didn't just rely on memory. Well done. We could be prepared. I think that the word agent has been bandied around as wildly as we expected it to. I think in particular, what we were talking about was the sort of generalist agent concept was going to be the marketing hype from Microsoft and OpenAI and those other companies. And that was going to fall short of what people need. Because when you get into agent territory, which is all about how can I do multi-step processes? How can I do complex work? You suddenly get very, very deep into domain specific difficulties. So taking an agent design to work for everyone and trying to get it to write an underwriting model or understand cap rates. It's very easy. That's you need to build technology to do that. You can't just do it by sort of poking around with a very nicely branded agent. Look, I'm going to play a devil's advocate here and pretend for a second that I know nothing about technology, right? Because there are a lot of businesses out there that are selling higher and AI agent for developers. We have Dean who specializes in the development of commercial assets. And we've somehow trained Dean to be perfectly performing out of the box for all tasks that you might need in running a development business. Now, I can see just superficially a lot of issues with that in practice. But I do think most people are hearing, hey, hire this AI agent or this team of AI agents that a company can easily produce for any given function expertise. Why is that not true? It's almost like where to start. Because this hypothetical has such a sort of iceberg nature of like there are so many reasons why that doesn't just work. But the biggest one, and I think it's something that we understand deeply because we work with such a broad range of institutions in real estate. And so closely with them is that the variety and complexity of the work that needs doing is just beyond what Dean could ever be programmed to do. So, for example, if you are creating a general agent and say development agent, giving it the ability to understand what's important, your domain knowledge, your expertise, alongside allowing it to understand the 2000 document data room, alongside understanding your processes. Just isn't really a solvable problem because there's so many different problems there. So, the sort of the reality of hiring a dean is like, well, you have a kind of, you have something that can help you with some tasks wrapped in an agent, essentially. Rather than something that's actually meaningfully going to deliver some ROI. Yeah, I think that's right. Taking my naive hat off for a second. I think one of the things that we've seen at Fifth Dimension is there are a couple different approaches here to get more optimal results than just saying you have an LLM that is, let's say, fine-tuned in particular ways to operate independently as an AI agent to do complex tasks is you can break those complex tasks into sub-tasks. And maybe you can send some of those sub-tasks to an LLM and maybe even two LLMs that compete each other to find a more optimal answer for just what is the least renewal date. If the end task that you're really looking to do is do I want to strategically offload this asset because it has too much turnover risk. And so threading together and understanding, as you say, deeply of what actually composes that complex workflow is a really important part of how you need to do product engineering to make an AI agent or what you might superficially interact with as an AI agent work. But I think also what we're saying is LLMs as their own sort of technology just aren't good enough. And we said that last year. Listen everyone, it's still true. It's a part of a system. The way I think about it is it's a bundle of intelligence, a raw spike intelligence that you have access to when you're building software. But pure IQ points or pure intelligence doesn't get you, doesn't underwrite a deal, right? Like that's a long involved process with expertise, accumulated knowledge, relational impacts. And so the kind of agent concept, I think, can be quite confusing if you haven't got your head stuck in the technology world. Because if you have a dean, for example, you're like, hey, well, if dean is an agent for developers, dean, how should I mask this site? And dean is going to be like, here's an answer. It's going to be terrible. It won't be calculated. It won't be considered. And the reason it won't be is because massing a site is a complex specialist piece of work that's pretty deterministic. So the actual way to solve a problem like that is to build a piece of technology that can work out how to arrange buildings on a site for the optimum configuration, to have some kind of understanding of the target market you're looking at, the assets, the use cases, the planning zones, and your own financial modeling. And then pull all of that together and come up with an answer. And the reality is is that you can't. There's no magical technology one shot button to do that, even with all of the magic we have available today. The answer to automating processes like that is working with a working with a technology like 5D's platform, where essentially the human is in the loop. You decompose those complex steps, you do them in the right order, you collect all of the contacts, the knowledge information, and it's all kind of played out. Yeah, I'm reminded of the business where we met, way home, which I love you, Nigel, great CEO, grateful to have met each other at that business. But even there, what did I join in 2019? We were talking about the idea of a one click home purchase, right? Well indeed, yeah. Which has been bandied about for decades in this industry, whether you're a broker. That's not the first time I've sat in a room or even worked on some elements of one click phone process, yeah. Exactly. And I think what's interesting to me to reflect on there is not only how far we might still be from that from a technical standpoint, which I think you and I and a number of people who are more technically minded are very ready to accept. But the fact that maybe it's not what people even want. Yeah, I mean, this is a big thing, right? Like, where's the oversight, the control, the direction, how do you express your opinion on that? I think that's why that's sort of higher an agent. That whole framing is just wrong because you're not hiring an agent. You're buying software to solve a problem. And you shouldn't be thinking about it as equivalent to a person in your team. It's a different tool site, right? Yeah, I think it both over and understaffs. Yeah, exactly. It's like, yeah, yeah, because software is very, has lots of benefits that people don't, of course, right? Like where transparency, for example, like understanding, well, there's lots of, let's not, that's a distraction, I think. But like, that's a whole rabbit hole to just quietly sides that. But yeah, I think the, I mean, I'm, I'm thinking about those kind of, it was more of a last year thing. Those kind of AI agent works place, you know, you had like a nice picture of an office and then you had Dean in one desk and then you'd have, you know, Alexa or another desk and Andy somewhere else. And you would see your agents walking around and you would distribute your work to them. And then maybe there's that like, oh, and then people put people in here and we'll have some kind of VR office space. And that's not how people want to work. And whilst it was a very entertaining, I think diversion of like, hey, we can, we can describe our software as people now, isn't this fun? It's not, it's not useful for, it's not useful for businesses, isn't it? Yeah, 100% and there are a lot of analogies like it used here, but I think of it a lot more as each human being becomes in collaboration with an AI their own function, their own collective. Right. So as someone who has a long term interest in psychotherapy, I think about family systems therapy. And I'm just like me, plus Agent Ellie, a fifth dimension is like me having a conversation with myself, right, because it's bespoke to me, it's trained in the way that I think about our business, the way I make decisions, et cetera, but with different angles, right? It's like, oh, we'll give me the bear case on fifth dimensions growth in 2030. And you can have that conversation with the part of yourself, right, that indexes towards that bias and has that context. Or I can sit and say, okay, Ellie, I want to talk about how we recruit top tier talent in 2026. And specifically, I want you to think like someone in the B2B SaaS bubble of the year 2015, who's trying to hire 100 people and not 10. And it's all about that collaboration and creativity that creates a set of inputs to the human progress of a future of decision making, right? You make decisions differently in an AI native world, but it's not because you hired some AIs, it's because you had a really interesting and challenging conversation with more real time data inputs and more insights at your fingertips and ever before. Yeah, for sure. And whereas I think the kind of agent marketing would have you believe that the future is a broker agent owned by, you know, a formerly well brand, you know, a broker agent owned by CBRE interacts with an investment agent owned by a big investment manager. And they interact and make a decision and then maybe the president, who's the only person left in these businesses says yes at the end of it. And that's, yeah. Exactly. No one wants it. And it's not possible. It's neither possible nor desirable. So I think that delta, right, between the technology, how it's sold and what its actual capabilities are, let's make a three point triangle there. Is a big part of the reason why some businesses haven't seen ROI, some, you know, palpable return on an investment from adopting AI technologies in 2025. I think there are probably a host of other reasons, but I think, you know, we kind of have to talk about that study, that MIT study that said 95% of enterprise businesses didn't see any actual returns, at least in a P&L sense, in a concrete, financial sense, from buying some AI over the past couple of years. Can you break that down for me? Yeah, I guess the thing that immediately comes to mind when I saw that headline was like was buying some co-pilot licenses isn't an AI strategy. And the reason, well, firstly, I think survey based studies are a bit, it's always a little bit finger in the air, right? Okay, do the scientists let me just, let me just say quality is always an issue, yeah. But I think the reality here is the fundamentals about adopting technology well still apply. So whilst you get some shortcuts, right, you can buy co-pilot or HTTP for everyone, and some things will happen. Great, there might be an overall like small uplift in the productivity of your entire organization, whether you have access to the data that actually runs your organization, whether you're using legacy processes, legacy systems, whether they're actually blocking that, whether you have to gather every single person in a room to make a simple decision. Actually, still the things that say down your business. So the way I would have been surprised if it was more than 5% because how old is, how old are AI driven systems? Three years? 82 years, if you think enterprise, it was very, the first cut was very consumer focused. And as you say, how many people within those businesses had adopted the technology and in which functions before they started looking for returns? The fact that 5% could find value so quickly, given that things take longer when you have a large business, right, it just takes longer to change things as more moving parts, is more of a good news story than that. Hey, where's the rest of it? Just think of that adoption curve, that those are those early adopters that were in a place where they could drive value and measure it. And what we'll see over the next years, I think it will be a big thing for 2026, is the next 25% of the moving. I guess we should probably maybe still talk about the people just have their head stuck in the sand, that are very much because I think the thing that, for me, the thing that has changed on the kind of ROI argument on technology investment is the ceiling is much higher than it used to be. So let's say five years ago we're talking about, we're going to migrate to Microsoft Dynamics, let's say, and manage everything through Microsoft Dynamics. The ROI on that is capped at some kind of back office efficiency gain, with maybe there's a tiny, tiny opportunity of your data being better place, maybe you can make better decisions. But it's purely capped at a percentage of your cost base. Whereas when you start to look at the opportunity with more intelligence software like 5Ds, it's all focused on your both efficiency but also decision making quality. Your ROI is kind of uncapped, right? Like you can actually drive alpha, you can make better decisions, you can create better returns, you're not limited by your cost base as a kind of return. And for me, that makes the opportunity much bigger. And that's why there's going to be this really interesting mix of the most, the most existing businesses that can change fastest. And the new players that aren't kind of hindered by how things used to be done and can do things differently. Yeah, it's interesting. I think what we'll see maybe in 2026 is that that 5% becomes 15%. But that we also start to see it not just being confined to back office to your point. Yeah, exactly. And there's a widening of the gap between those who are doing it well and moving it pace in AI adoption and those who are doing it poorly. Yeah, especially those that go faster, right? I think that's the thing. I so rarely see any any technologist talking about really. But when speaking to our partners, our customers, like speed is a really important thing, you know, like, what if you could make an underwriting decision in a day and pace, I think makes that huge difference, like speeding up this. Yeah, I think what precedes speed or pace is prioritization, right? When we look at our customer base and we look at our most successful adopters of AI, the reason they've gotten to where they are to your point is they've made it, you know, one of three strategic priorities. They haven't talked about, you know, we have an AI strategy. We have three pillars for that AI strategy. It's, you know, AI ready by 2030, which is literally still where some of these global asset managers are in terms of adoption. It's they are committed to developing a roll out adoption program with us, right? Where they leverage internal advocates who are already tinkering, right? All night with AI. They're like, okay, those people, right? Like, they're going to sit in a room with fifth dimension and they're going to figure out the next 30 ways, how by in the next 12 months, right? Somehow the work of those people, the interest curiosity and productivity of those people will profligate throughout the rest of the business. And I think that us watching that happen, then really informed our product decisions for how we were going to build and move the product forward. Can you talk a little bit more about that? Yeah, absolutely. I think the tinkerers of AI is a, is probably almost any business as big as asset, right? The group, the group of people that understand your domain, understand your business and have that growth mindset where they're like, I want to explore, I want to understand, I want things to change, I want this organization to learn so it can do things better. And what we saw is working, it's the, in each customer, maybe one in 10, one in 10 of our users, I would think of as one as a kind of tinker reminds them. And we saw what they really wanted was to scratch that itch of kind of getting deeper. So rather than say, hey, five, fifth dimension provides me this great rent role workflow where I can kind of automate my rent while they can like, well, what if I just wanted to tweak this field here and understand how this works and kind of pick at this and change this. So actually what we've done is lent in really hard for that because we've seen that kind of impulse and realized that we can add a huge amount of value by empowering those tinkerers by essentially saying to them, hey, here's a platform that understood that it's connected to your data. Understand your domain and has many of these kind of workflows templated we call them skills and allows those tinkerers to really actually then say, right, well, I'm going to, I'm going to take ownership of the skill that automates our initial underlying. And I'm going to really, really kind of go deep in that in the same way I would when I'm building a spreadsheet and get it just right for current thesis, current business, and then it's going to be reliable and then continue then I'm going to share it. I'm going to share it with probably every once I'm really excited about it, but maybe maybe maybe maybe our maybe our IT team has controls over who's actually sharing it, but you know, I got very excited when I'm sharing it. And then you kind of see the it's almost like these nexus points in business, you see this kind of like automation flowing out. And that drives so much faster and higher value change. And also, I guess answers that question of like, well, you know, I have thousands of mini processes or tens of thousands of mini processes in my business. How am I going to write them all down, work them out, and then change them to be a native. And the answer is, actually, you need to lean into, lean into an empower those kind of growth mindset tinkerers in your business. And that's been very core to our product, all of our product work this year and our product philosophy for for 26. I wonder if one thing that might be coming sort of hard and fast for a lot of the businesses that we work with is the official or unofficial appointing of AI engineers within their business, right. And you and I have talked about, you know, sending people jerseys, you know, honorary QB for the investment team, AI engineer. But I think that that might become part of the very incentivization structure of a lot of these businesses, because how could it not be, right. These people are not only driving value in terms of bringing their domain expertise about how underwriting of retail assets is done to the building of technology. They're rapidly becoming more informed about AI technologies, not only more than most of their peers, but more than a lot of people working in AI businesses. Yeah, for sure, because it's so applied, right. The thing that the advantage that that kind of that AI engineer, let's say, has is that they're working on real world problems with a life feedback loop. And they understand that, you know, they see when you chuck a complex kind of bundle of lease amendments into, let's say it's co-pilot for argument. And then you finally worked out how kind of the agent builder worked and you kind of work that. But it just breaks because there's too much, it's too hard, it's too difficult, and it's brittle. And they sort of start understanding the limits, I think, faster than others might. Yeah, there's a real power of learning by inference, through tinkering, that I think is generally undersold in terms of AI, but it's why people say things like, you know, start small start today, because it is true. It doesn't really matter whether you started with chat GPT or co-pilot, and then you graduate to something like fifth dimension, the more you use it, the more you get a really intuitive understanding of, oh, if I ask a question in this way, this is the type of answer I will get. And the combination, right, the multiplicative effect of that kind of knowledge with deep domain expertise in how your business works is the future, right? Like that is the artistry of the AI human being collaboration, in my opinion. I'm reminded yesterday of giving RCTHN some prompting advice on how to make something look better. As an example, I guess the one way, one, one thing we've seen much more this year, and I expect to see more in 26 is we've seen more customers or potential customers coming to us with kind of demos and mockups and prototypes that they're AI tinkerers have done. They've been using loveable, but in the right way. We're really excited by this. Look at what this could do. Look, we've got this kind of janky deal cloud API bill. We can kind of in theory push data. It doesn't quite work, but we can see the potential here. Can that can that work? Is that possible? Can you make that work for us? To which the answer is, of course, yes, that's what we do. Yeah, I mean, it's really exciting, obviously, for both sides of that equation, because it shows a different degree of buy-in and commitment to an AI native future, right, to be in that headspace. But one of the things they're also talking about is the importance of the data that's being brought into the AI systems layer, right, of your business. And I think something we've also both seen this year is a lot more businesses having been with us for a short period of time, or for the past three years, saying, like, we now, having tinkered, realized how much better the output of Agent Ellie could be from a future facing investment decisioning perspective, if we were to allow access to this system, that system, this system. And not only in a read and in ingestion sense, but in a right sense, because if you are going to open up your world to AI, you also want that AI to be ultimately in control of the source of truth. For a lot of your fundamental transactions data, and the businesses, again, that we've seen that are moving fastest towards that return on investment in AI, have been opening up more and more of their worlds of data to that AI agent. Do you think that's correct? Yeah, absolutely. We label this crawl walk run, don't we, that you start with the tinkers and the users getting value out of automating parts of the exploring automating parts of the process, making things faster, better, starting to accumulate intelligence. And then as that starts to mature, then you're into the space of actually, let's start integrating our kind of unstructured data share point box. But also, let's integrate with our data lake house in fabric and the core transaction tables, and let's put those at which is the kind of walking stage. And running is the fall. Actually, I want, you know, I want the system to write back to those tables. I want the system to update when there's a change in kind of when a new lease is extracted and we have that information when there's a new decision made. And we've one of the, I guess, one of the sort of one of the specialisms of our business is taking, helping people on that journey and helping them deliver the kind of walk and run phase as well. And it is quite incredible when you start, when you start connecting the data sets that matter that previously been locked behind kind of proprietary, very difficult to work with kind of BI systems or reporting systems or just, you know, the sort of data extraction of the RPA system. It's a, like, people's eyes light up because they see a report that would have taken days, maybe even for an external consultant to run, they can just ask for their and that and it's run instantly and they can watch it kind of, they can watch it, generate, they can understand how it's formed or our sources and then get straight to the value at, you know, the decisions or what they're doing. So I have my favorite examples, but give me some examples of use cases here where that extra bit of data engineering is really paying dividends with with fifth dimension and then how you think that's going to play out as we move into 2026. I think the, if I answer this and reverse, one thing that we've seen a lot of is that there's a drive to start actually centralizing data more in the industry. But the sort of historic skill or us as an industry haven't historically been great at data pipelining. It's, it's not something that many organizations have done well in the past. And actually we, that's something that we're helping a lot of companies with is kind of, okay, if you're thinking about building out a data pipeline because you need to transform your financial data, you know, your accounting data. And it needs to go through this process. We can help you with that deterministic process because we know that it will unlock so much, are you being plugged into your data lake analysis and then into 5D system and carrying building those pipelines and connecting the core information in your system is a massive step to unlocking value and moving forward. And so I guess one of the things we're seeing is maybe, maybe the first integration is no longer into a kind of line of business systems, I would think of it like a CRM or an RPA. The first integration is often going to be Microsoft Fabrics or Snowflake, which of course is more flexible and easier to work with and you can kind of once that's established, you can just grow fast. And we, that we see that will kind of become, I think, more of the standard operating for a lot of businesses, which means that it puts some much further along on that kind of data journey for whilst getting by because I think the other thing that a lot of people, we've spoken quite a bit about this on the podcast is not to get too lost in the, but we've historically been quite bad at data pipelining. And maybe we have to be ready for AI in 2026 or 2027 and we're going to wait and we're going to hold off and you can, you can get there now, but you do have to progress, right? Yeah, of course, for me, if you want to run, you have to put some effort into your system, yeah, you have to try that. But there's still the rocky montage, right? Yeah, I mean, I think about, I have a lot of thoughts and feels as I know you do about sort of the promise of enterprise AI search. Oh, gosh, I've been trying to avoid saying that. I know, it's a note. It's like an anti use case, isn't it? It's a black hole, but I actually think when people, if you scratch the surface of what people mean by that, all of the work we've now done with our, our sort of more successful customers are most advanced customers in helping them devise a proper source of truth that an AI agent can have access to all of those data. It actually delivers on the promise of enterprise search, i.e. what you get out of that in the short term is to ask any question of any data no matter who you are in the business. So you can say things like, what are the last 10 deals I did like this deal, and then make an inference from those deals about what my business thesis is and then save that business thesis as a skill to Agent Ali and then apply that business thesis to assessing the next 10 deals, right? Because that amount, it's always a piece of the puzzle, what people are asking for in enterprise search and what's required in order to make that happen is a variety of technologies and also access to the right data at the right time, right? But I feel like we have to contrast that with what how enterprise searches sold as a general product. Which is, we do, we do, you plug in your SharePoint and type questions and everything will be known. Which is just, I mean, imagine asking a human to do that. It's simply not how people or technology work. SharePoint is part of it, but it's also integrate your transaction data and integrate your core business metrics and start building kind of institutional knowledge and then you can then you have enough context and not information to pull that kind of information together. Plugging in a generic kind of enterprise search tool, hoping it will also be able to understand that 15,000 page PDF you've got an environmental report on the last asset you bought is too much magic. It's not true. Yeah, it's several degrees of magic wand, but if I'm to again embrace techno optimism for a second, I think one thing that will be unlocked in 2026 then for those businesses that have made this investment is that ability to see around corners, right? So today, you know, get any answer of a set of data that allows you to make an investment decision tomorrow. It's going to be, hey, we know what your goals are for this year. You're, you know, tracking 30% behind those goals this quarter, we recommend you make these four changes to your portfolio and actually, you know, drop retail from your platform, etc. And that will be what helps you achieve that goal and some of that work being done for you in the background, which I think is really exciting because again, when we started talking about this stuff with OpenAI's initial launch several years ago, this is what people were really hungry for and excited about and I do think we're actually closer to that today. We're so much closer. Yeah, absolutely. The combination of better understanding how to connect these disparate debt resources and the actual intelligence we have access to puts us on the path where we can start, as you say, seeing round corners, start which really then drives better decisions, better returns and alpha. I mean, I think the future is bright, but do you have any final predictions for 2026 or advice for our listeners? I have a very boring one that I think Google will be the technology star of 2026, but that's a very me kind of our listeners are probably less interested in that prediction. I think they will, they've been playing a long game and I think they'll come out ahead as a result. I think for me, the core of 26 is that we'll see the impact of that 15%, we talked about earlier, the sort of 95% they've got ROI. We'll see the impact of the sort of next 15%, which will drive a lot more best practice, performance, interest and start to really differentiate the companies that are on this path. That are walking around by our technology versus those that are watching, let's say, not even crawling. And for me, you'll see the first businesses actually driving better returns the result. So as this kind of, as we start to unlock more of that intelligence and seeing round corners, you'll see businesses that are doing that, making better, you know, literally delivering higher returns. Nice, I love that. I think I'll go for a two for here in that vein, which is if you want to win big, you have to play big. That's also true. So we've passed the era of experimentation with AI, right? The businesses that are going to come out the other side stronger, better and in fact survive the contraction of the marketplace around AI native businesses over the next several years are going to be those that took risks in using their financial resources and their team's time to find the tinkerers, leverage them, rethink the way they do business. I feel like that's a whole, we could do a whole podcast on that, like learning how to change, because that's it, right? Go to think big and then learn to move fast, because what is happening in the wider world with this technology and just the whole world is there's more uncertainty, more change, more chaos. And how do you deal with that? You deal with it by adapting to it faster. That's right. So I guess two back in for some future episodes on these topics. Teasing future topics, maybe. Thanks so much for joining me, Johnny, and here's to another year at Fifth Dimension. Thank you, Kate, and goodbye, everyone.