Kate Jarvis, CEO of Fifth Dimension on PropTech, Vertical AI and Building the Bloomberg Terminal for Real Estate

0:00This episode is sponsored by Cpoint, the business account built for startups. If you're a startup founder fed up with finance admin, you need Cpoint. Cpoint is different to neobanks because it connects all your bank accounts and your Stripe account to see your cash position, burn and runway instantly. It automates bookkeeping by pulling invoices and receipts from your whole team's inboxes, so you just sync everything to zero and pay outstanding bills with a click. It has a 3.49 % yield on treasury and real human customer support. Find out more at seapoint.co. That's S-E-A-P-O-I-N-T dot co. Use code UNICORN for a free month.

0:35Seapoint Treasury is a money market fund. Rate recorded at 1st of October 2026. Rates are variable and subject to change. Capital at risk. Hello and welcome back to another episode of Riding Unicorns. Today our guest is Dr Kate Jarvis, CEO at Fifth Dimension. Fifth Dimension is bringing a long overdue AI revolution to the stagnant prop tech industry by challenging conventional approaches and siloed data bottlenecks. So Kate, thank you so much for joining us on the show. Maybe we could start by just hearing a little bit about your background and what led you to starting the business. Yeah, of course.

1:11So long-term machine learning nerd, I should probably get that out there at the outset. I got my PhD from Stanford in linguistics and symbolic systems. And this was back in the early to mid 2010s, 2012, 2013. And at the time, I had mentors who were sitting around doing NLP analyses of Amazon reviews to predict whether if you use this particular adjective, you were really happy or really mad about what you had purchased and using those data, obviously, in business to inform what people care about and how to market products. And so I was always fascinated by essentially how you can leverage large amounts of data, data at volume to filter noise from signals.

1:56And I wanted to apply that kind of understanding of statistics, of language and so on to help really improve the lives of everyday people. So when I graduated, I was in educational technology and, you know, use machine learning algorithms to essentially build math products that could infinitely generate new math problems for eight-year-olds that are tagged to California's curriculum in order to get them to understand particular concepts. And these are great applications, right, of machine learning-based technologies. But it's tough because education is highly regulated. And also, you're selling really to people's parents rather than to kids themselves, which is a weird-ass dynamic, I have to say.

2:42And so I wanted to break free of that. And really, if I care about using machine learning to leverage, kind of create socioeconomic change, why not just go into financial services, right? Like how could I be in fintech and what problems could I solve in fintech at scale? And I'm particularly interested having lived through the educational technology experience as well in digitizing laggard industries. So I lived through at least two boom and bust cycles while I was living in San Francisco through my PhD in startups and scale-ups of new technical innovations that everyone talked a big game about.

3:16And then they turned out to be absolute dog shit when they were in the hands of consumers. So I was also highly aware that tech innovation, as we know, the pace picks up year on year as we go further through history. But that doesn't actually change human behavior. Human behavior isn't transforming any quicker than it ever has. And so I wanted to take that insight as well into financial services. And that's when I really got deep into understanding real assets. When I went into a private shared home ownership scheme business called Way Home, which is where I met my now co-founder, Johnny Morris.

3:52And back then, I was just fascinated essentially about how the built world, A, operates on an enormous amount of debt. You look around at these skyscrapers and you're like, how does this exist? How can we have an office in this building? It's because there's an enormous amount of debt that's funded it. And how complex these capsecs are against properties that you hold for really long times. They take a lot of time to develop. They're enormously capital intensive versus capital markets trading, right? I'm a retail investor. I can sit around, I can short the S &P 500 because I think something interesting is going to happen and I can make money off of it.

4:28But there are tons of people in the world and the large piece of financial services is tied up in the built world. It's like the hundreds of millions, the billions of dollars that you need to put together and the parcels of lands you need to put together to fund a wind farm and manage a wind farm. And so I got really interested in that problem and sort of the data angles of that problem while working in a business where we're doing residential transactions, and it would still take us weeks to months to try and underwrite residential transactions with a large team of people, because we would sit around updating Excel models all day, trying to extract data from PDFs, and then taking photographs of comparable properties on Rightmove and trying to figure out how to make a transaction work.

5:08So that really exposed me to the problem. And then you multiply that by like 100 ,000 when you think about businesses like Blackstone or Graystar, you're like, okay, if you're doing anything that isn't just residential, it has a higher price point, the underwriting is enormously complex, right? Some of the stuff I was just saying about the number of stakeholders, different types of credit and debt that can be against any property, etc. You have so much data you need to deal with in the built world environment when you're talking about the financial services aspect, and it's still all processed entirely manually.

5:41Today, even though they have brilliant machine learning algorithms, Blackstone is running hundreds of millions of transactions off of spreadsheets for valuation. I knew that there had to be a better way, essentially, based on my training over time, and that that better way really had to draw together unstructured, semi-structured, structured data into a single kind of decision-making engine that gives you the right information at the right time. And so that's really how Fifth Dimension was born, is we were like, we need to make something like Bloomberg Terminal for people who do this thing in real assets.

6:14And it's not just market data, because everyone can have access to market data. What you're thinking about are the 70 years, 90 years, 120 years of transactions history that sit against property that you're interested in. And how does that inform the decisions that you make? So yeah, it's a long thread of nerdery being exposed to how undigitized data are in the real assets and infrastructure industry. And being like, wait a second, that's also where most of the world's money is? How is it possible, right, that in this like weird area of financial services, you look under the hood, and like the literal only technology, the innovation that's happened in the past 70 years is they started using Excel.

6:53Amazing. I mean, there's definitely a lot of lessons in there about insights and inflections and noticing both and pulling that together to come up with the business. And what is the least appreciated data source for property? Like what what goes under the radar that you guys provide that you think is really valuable? It's a good question. I would say our customers bring the data to us, we provide them with insights, right? So like, what is interesting data to you is essentially a big part of your special sauce as a business, whether you believe like CBRE's report or Newmark's report, right, on market trends in this quarter of this year, actually is really fundamental to your buy box, what is your investment thesis and how much of this is just marketing, who is over or under indexing on inflation and all this other stuff.

7:43So we rely on having each enterprise customer we work with has their own set of data points that they care about across a whole bunch of disparate sources. We just help thread that together and serve it up to you in real time. So like, let's say you're managing a portfolio of 367 hospitality assets around the world, you essentially have a quarterly reporting cycle that you work to. And it takes the time between each quarter, right, for you to go to each of your property managers to try and understand what is your P &L? What is your CapEx? What is your OpEx? Oh, this thing you coded is like a utility cost.

8:19I thought this was a triple net lease. Like, was I supposed to pay for this? Was the tenant supposed to pay for it? Why am I losing$50 million over here? And it takes four months, right, to get any of those data collated and understood asset by asset, and then roll them up into a portfolio level. And by that time, right, you've already lost that money. There isn't a chance to react and raise rents in advance, for example, of a fuel crisis like the one we're expecting today. But you can't even do better loss recovery against tenants who have high credit risk and are maybe about to default if you have no insight into those data.

8:54And so for us, what we think is missing is not just more data. And we don't need you to necessarily invest in digitizing all 100 years of your past transactions data. It's about what you're trying to achieve. If you can just get insights, intelligence surfaced to you at the right time, which is why we call ourselves a decision intelligence engine, because I actually think we're changing how people make decisions rather than like a data intelligence platform. You said at the start, property is like the biggest asset class in the world. And we've seen the enterprise value created in the fintech space.

9:27And it hasn't really been replicated in PropTech to date. Do you think that's because the products haven't been that innovative? Or do you think there's some structural challenges around getting people to behave in a slightly different way? And that links there to what you just said about a decision intelligence because often people don't want to necessarily change their workflow. They just want to make better decisions. So is that being important to how you think about positioning the product based on kind of what's happened previously in the prop tech space? Like, look, the market is really disparate, really fractious, really complicated.

10:06And that goes back to like what's different about building a skyscraper in downtown Tokyo, funding that and following it through its lifecycle, managing that as an asset from, you know, retail trading. It's just enormously diverse and complex, and it has to bring together necessarily so many stakeholders. And I think that in and of itself is both the challenge and the promise of digitization in the industry, right? It's just why we don't have an E-Trade for skyscrapers, because you would have to go through and look at this enormous amount of documentation across thousands of different individuals and try and figure out how it's doing and what's actually happening.

10:43But I also do think there is this change management piece. So yes, a lot of crop tech software hasn't succeeded in the past because we haven't had the right kind of backend technology to drive people's behavior forward in the sense that before LLMs were around, it was just really hard. But I also think all industries are about people. Real estate is about people, perhaps even more so than most industries. And if you are used to getting a phone call from your broker, this is how you, you know, not even an offering memorandum or a property brochure in your inbox, but a phone call. Hey, I've got a hot lead, right?

11:22Here's something you might be really interested in. And then you go back and forth a bit on, you know, the details of the property. Maybe the facade needs refurbishing before you could call it like a true class A asset. Maybe it's not quite close enough to downtown Los Angeles to be the type of asset you're looking for, but you sort of start to dig through the data and then you would open your laptop and then you would open your Excel and then you would open your PDF. The way that people do things today is the biggest barrier to any business making money in real estate. However, I also believe that what people like about that process and what they don't want to changed, even as they become AI native businesses, is really the kind of interplay between the relationship and strategy.

12:07So you like that broker's phone call because it's deal flow, right? We're in VC and we are running VC backed businesses here, right? What that phone call says to you is, oh, I'm about to buy something no one else knows about yet. How do you reproduce that on technology? You can't. So someone probably still has to make a phone call. But what I could surface to you in my platform is like, hey, no one's called you about this yet, but actually, and this is a real life use case, right? We can do off market prospecting for you where you would have formerly had to watch via Google SQL queries and so on.

12:40Hey, a large property manager of assets in the Southwest has died. And now they're up for distribution in a way that we didn't expect. I think this is the first chance the market is going to have, right? To like get in on this portfolio because they're going to break it apart. We can identify that that might happen, but you still have to have the conversation. So that's why we focus a lot too on like, we're not here for you to build a smaller team. And we talk about this in the sales process. The point of working with Fifth Dimension is so that your analyst, your chief investment officer, your CFO, your broker can then have that data and intelligence service you, but then you're still going to go look at the property physically.

13:23You're still going to go call the broker and have a conversation. And in fact, now, because all this work is happening in the background with AI agents, and the right intelligence is being serviced to you at the right time, you can do 10 times more of that. Amazing. And so you obviously have a very interesting product that gives them a lot of insights. How have you gone about selling? At what stage did you move from founder led sales to bringing in more team members to assist with that? Tell us a little bit about once you kind of had product market fit, let's call it. How did you start to get into more customers?

13:57I think if you're a Series D business, you probably have like the slickest, well-oiled, you know, go-to-market machine. You're selling everywhere in a way that makes sense. It's localized, but it's globally coherent. But like we're a Series A, right? We're a Series A stage business. Encompasses the transition out of seed through Series A to Series B readiness. It's like an enormously big amount of variation, right? And how much you're actually building out your team and how much is repeatable about your sales playbook. And I happen to run a business now where we have three offices already across the world because we have to be global because our customers are global investors, which also adds a whole bunch of noise to the mix.

14:40So for me, first of all, obviously I'm a product person. I've been a career product and technology officer before I went into the CEO role. And so the only thing I essentially care about is creating such an enormous amount of value for my customers that I also create shareholder value. That's the equation for me. How do I go out and create so much more revenue for the businesses that I work with by transforming them from, some of them don't even have cloud-based software today, if we're talking about some of our Asian customers and some of our customers in the US that just haven't yet bought an ERP even.

15:16and we're then taking them into the era of AI. That to me is a project that requires the engagement of a series of product-led people who also do sales. So about two years ago, 18 months ago, is when we really started selling a lot of contracts in the United States, which is now 80, 90 % of our revenue, having started the business originally in London. And so I had AEs, I had account executives in London, But then it's like, cool, now I'm opening up business in the US and I'm doing a lot of founder-led sales in that seed stage. And I know I need to set up a machine here to make this Series A ready.

15:56So of course I need to hire a head of sales and so on. But my reaction is almost like, por qué no los dos? Because we are still changing so much and we're still in a transition phase, I'm still running my own pipeline. I also still have a US head of sales. I am still doing founder-led sales at Fifth Dimension. And that's integral to our understanding of how we have to keep developing product to make people happy. But you definitely need more than just you. Also, I would say just to get enough reps to get enough signal about whether what you're doing or not is working. So to your point about product market fit, in enterprise sales, you can have really high value customers, right?

16:34I can have 50 customers and be making 50 million of ARR. so you have to sit there and say like how do I get enough exposure to what the market is doing and what the market needs in order to build a truly scalable go-to-market engine which is also the motivation for me hiring people that aren't just me to do sales if that makes sense yeah absolutely and how do you balance like token usage with pricing and making sure that you kind of are building a robust business model as you scale. And while tokens might be cheap now, the pricing might change, you know, staying nimble to those kind of pricing challenges.

17:13I speak to my investors about this all the time, obviously. What are the two main things that you worry about as an AI business? You worry about your unit economics as they relate to compute and token costs. And then you essentially worry about your organizational design. Like, do I have not not only the right sales team, but if I need to do digital transformation, which we do, the product works out of the box. Let's be clear. I actually still need a pre-sales and post-sales team who can sit with those folks and teach them how to do prompt engineering, teach them how to click buttons to get the outputs that they want for an investment committee memo.

17:50And so that stuff is really complicated as the market talks about, oh, we're looking at token maxing as a way of understanding how productive our employees are. And then within three to six months, the backlash of Google saying, sorry, we can't pay this much to automate our workflows with foundation models. We're ripping all of this out. Cloud code is too expensive. We're not doing that anymore. And it is enormously expensive. Anthropic talks about having 500 % NRR. And we have this joke internally at Fifth Dimension that the 500 % NRR at Anthropic means you paid them five more times than you thought you would this year, because you had no idea what the compute costs would be of getting your work done.

18:33And so we approach it totally differently. And you've gone through now a few funding rounds. How did you find that? What did you find the sort of biggest challenge? And what advice would you give to anyone else that's sort of in the fundraising process at the moment? Look, I'm a woman CEO of a deep tech native. And for anyone who's listening to this, I've said this before on my platforms, but things have gotten more sexist since I started in technology 15 years ago, not less. And they've gotten more all of the bias things. And I actually think this is why I also wanted to start an AI business, because when you start 10 or 100xing the outputs of every individual in the universe, there are consequences of that.

19:22And the consequences of that are you've just created a whole lot of noise through which you have to assist and find signal. And every time I've gone and I fundraise, honestly, it must be, I've just raised a$26 million Series A, and I have three offices around the world. I'm always going to be hearing when I'm talking to a potential investor also about why my numbers aren't like lovable's numbers. Why am I not an overnight 10-year success, just like the business that everyone is talking about today? And I'm creating an AI decision-making engine business because I also think there are moral and immoral ways of approaching problems like this.

20:03And investing is really hard. And trying to raise money as a founder is really hard. But I think the process, in my view from when I started in technology, you know, as I was saying in the early 2010s to now, it's noisier, it's harder, and it's more difficult. And like numbers support this. The growth multiples that you have to put up today versus classic SaaS multiples to be taken seriously for a series A are insane. And so what do I think about that? You just, again, have to find a way of keeping yourself sane in that environment. And you also have to be have enough deep convictions and be right about at least 51 % of them in order to keep going.

20:50So I think you should only, of course, raise money from venture if you think that's the right thing for your business. And that's the trajectory you want to be on. Did you find any difference between raising in the US and in Europe? Yeah. And I mean, this is probably a bit particular to AI nerds, but AI is all anyone's investing in really these days anyway, much to many funds detriment, but that's a topic for another time. What I would say is that the US is still so much on its own trip on basically like foundation layer technology. So even though we've published white papers and have open source some of our proprietary rag models, et cetera, like that isn't taken seriously because we are an applied AI business.

21:37And so I would have conversations with investors who are really interested in what I was doing and the fact that I was changing an industry that hasn't really been changed since the beginning of time and saw the opportunity, the TAM, you know, money's in their eyes for all of that stuff. But everything that's crossing their desk in the the United States right now, and I'm especially thinking about San Francisco, is still cross LLM coordination and comparison play, control token expenditure play, cybersecurity, AI native cybersecurity play. It's all completely the industry drinking its own Kool-Aid type of businesses that for the most part, I think 90%, 95 % of businesses that are getting funded at Series A in the United States right now.

22:20Whereas, and this is the one nice thing I have to say about Europe because Europe is still, I said 80, 90 % of my revenue today is in the United States because Europe doesn't eat its own dog food, even though it's producing really beautiful verticalized AI companies right now. They are actually investing in vertical AI. I think the most beautiful thing about the European VC market right now and how it's been able to grow and develop, one of our partners is Speed Invest. They let our seed ground and now I'm watching some of their businesses get acquired by really large model labs, right? Because all of them cared from day dot, how does AI transform actual businesses?

22:58The question wasn't, how do we create a thousand cottage industries surrounding AI that will make more money off of AI? Because we all live in an effing San Francisco, like Silicon Valley echo chamber. I feel like the European VC question was always like, how does this actually work when you put it into real businesses? And like that was definitely reflected in the feedback that I got in the marketplace. Now we have great American investors as well. But I would say highest level generalization, I think Europe is in this great position to win on vertical AI, while the US is still going to take several years to get interested and say, Oh, I don't even know what financial services AI looks like.

23:37What is this weird thing you're doing? Yeah, super interesting. You missed the boat and you don't have the capital to kind of play that game so you accept the fact that well hang on what do we have well we have lots of really amazing companies here and great talent here and real understanding of how the businesses actually operate across different geographies and everything so let's really focus on the bit we can understand and control and fund which is that sort of more the application there and almost have to accept the fact that the money and the innovation and the timing of getting into like a foundation model business is probably only really applicable to like the valley okay we've covered some really great ground we're going to move on to our final two questions so the first is a future unicorn prediction it would be great to hear of a company that you've spotted that you could go on to do really great things?

24:33Yeah, sure. So I totally back Slow Engineering, Pari Singh's company, who I also know personally, it's a Seedcamp portfolio company. Seedcamp are also investors of ours. But basically what I just said, I think to sum up some of the conversations we've been having here today, AI is going to transform everything, but it's not going to happen overnight. People always change a lot slower than technology does. And some of the most obvious like money saving optimization plays right now for AI are leaning into robotics, supply chain issues, et cetera. And we saw this right with past wages technology.

25:13And so I think flow with what they're doing, where they're creating knowledge graphs and using knowledge graphs to help physical hardware engineers optimize their scenario planning, right if I change this material here that affects a whole bunch of other things about at this temperature this will happen to this part of a car which is a great application of knowledge graphs that business is going to do amazing amazing things yeah awesome great pick an incredible company backed by like Seedcamp as you said EQT Ventures Sequoia have recently come in in the last year or so super interesting and one that people should go and have a look at because it's just fascinating company.

25:54And then Kate, our final question is our dinner party guest game. So if you could have dinner with any three people, who would they be? So the first I think would have to be Michel de Montagne. So longtime fan. I don't know if anyone listening to this cares as much about literary theory as I do. But basically, he's known for creating the essay as kind of a literary genre. But much more important than that, he kind of founded the scientific method before science was really a thing at a time when society was controlled by religious fundamentalism. On a similar theme, I think Alex Edmonds, who's a professor of finance at London Business School, he wrote a book that came out pretty recently called May Contain Lies.