Pełna transkrypcja odcinka Outsource the Work, Not the Thinking—with Nikolaj Brammer (Heimstaden)

Maybe we'll all live in the cloud uh one day, but at least uh that that seems to be a bit uh further away. 00:05:29.920 --> 00:05:32.560 And I don't think that's given in real estate uh precaution. 00:05:32.879 --> 00:05:34.079 If I can just make one comment. 00:05:34.240 --> 00:05:58.399 I think a lot of other real estate segments, if you think about office space, retail, that is being disrupted by the internet, uh, you know, uh shopping online or work from home and and and different uh consumption patterns, uh, but at least residential seems to be uh relatively uh safe from a fundamental perspective. 00:05:58.560 --> 00:06:03.439 Uh that doesn't mean that what we do can be digitized. 00:06:03.519 --> 00:06:19.839 And if you think about what we do administrating, making investment decisions, uh asset management, uh, capex, letting uh there are a lot of things operationally that we need to uh to make work. 00:06:20.079 --> 00:06:28.800 Um we sometimes say it it's not rocket science, but the complexity is in the volume and the number of things we have to do. 00:06:28.959 --> 00:06:38.720 And I think that lends itself really well again to digitizing uh what you know how we think about conducting our business. 00:06:39.680 --> 00:06:45.120 I'm I'm curious, like when you first took on this role, how did you think about where to begin? 00:06:45.279 --> 00:06:57.759 You know, we talked about that F1 pit stop analogy that you shared, um, you know, kind of uh being kind of uh it's 67 seconds in 1950 and now kind of under two seconds today. 00:06:57.920 --> 00:07:09.360 What is the corollary or the analogy around how you viewed the process and reducing the time or kind of making things more efficient in the kind of space that you're operating in? 00:07:10.879 --> 00:07:12.879 So it's a it's a great question. 00:07:13.120 --> 00:07:21.120 Uh I think one of the beauties of coming from generic consulting is that you can view any problem in the same way. 00:07:21.199 --> 00:07:29.199 So you don't need to be a specialist in technology to say, how do we think about the strategy when it comes to the digital area? 00:07:29.360 --> 00:07:37.040 Because it would be the same for uh, let's say, sales or any other uh part of the business. 00:07:37.199 --> 00:07:41.839 And I think it's about understanding what is the situation today. 00:07:42.000 --> 00:07:54.240 So analyzing where do we come from, and then it's saying where do we want to go, and then uh how do we get there, where to play, and and how to win, essentially, right? 00:07:54.399 --> 00:08:02.480 So you need to understand what systems do we have today, and where is it we want to go? 00:08:02.639 --> 00:08:25.759 So we want to be able to make the pit stop in in two uh seconds, and then just analyzing what are we doing, where are the bottlenecks, where's the opportunity to do uh things uh more efficiently, uh, and doing that from the ground up with like applying first principle thinking and a lot of uh rigorous prioritization and analysis, right? 00:08:26.319 --> 00:08:28.800 And it's about dissecting what we're doing. 00:08:28.879 --> 00:08:32.480 So, say you're conducting uh a letting journey. 00:08:32.720 --> 00:08:42.559 We publish um around 99-0 advertisements uh every day of apartments that we want people to uh move into. 00:08:42.799 --> 00:08:56.399 It's of course very important for our top line in order to both secure the right pricing, but also ensuring that we get it rented out as quickly as possible, that we get all the steps in this process correctly, right? 00:08:56.639 --> 00:09:02.080 So we need to be able to have the right pictures to make the advertisement look uh amazing. 00:09:02.320 --> 00:09:11.519 We need to have the right description, we need to uh publish the advertisement at the right time, whether that being Sunday at 5 p.m. 00:09:11.759 --> 00:09:19.840 or Monday at uh uh 8 a.m., depending on demand and depending on how long time you have before it vacates. 00:09:20.240 --> 00:09:34.639 Then when a prospect tenant sort of expresses their interest, we need to be able to capture uh you know that um interest and convert it into a viewing, and we need to be able to sign a contract as as quickly as possible. 00:09:34.799 --> 00:10:05.120 And I think, you know, to get back to you the your original question, the way you get from 67 seconds to two seconds is that you analyze the entirety of what we do, figure out where to prioritize, of course, and then you get really get to the bottom of what we do, the processes, and then you think about how you can uh make them even more efficient, often by applying uh technology and increasingly uh AI. 00:10:06.399 --> 00:10:17.120 Yeah, I I do want to definitely expand on the AI piece of it, as it kind of um uh it can obviously introduce a lot more leverage and efficiency in that analysis. 00:10:17.360 --> 00:10:30.960 But I I think I want to understand, you make it sound very simple, but traditionally, especially you know, residential real estate, I'm sure, has been really slow to adopt this way of thinking and improve and uh implementing some of these efficiencies. 00:10:31.120 --> 00:10:38.480 There's a lot of legacy systems, there's uh things that are spread across different countries uh that make things hard to communicate with. 00:10:38.799 --> 00:10:54.080 What do you feel like is different about your approach and what you're doing at Heimstad and versus kind of what you traditionally see in the industry, where at least from the outside in, it seems very slow to adopt, very like laden in legacy processes. 00:10:54.879 --> 00:10:55.440 Yeah. 00:10:55.600 --> 00:11:01.120 Um I sometimes hear people say it comes down to culture. 00:11:01.200 --> 00:11:06.639 Uh, and the first many times I heard that, I was like, yeah, that sounds like a cliche. 00:11:06.960 --> 00:11:08.399 Uh, but I really think it's true. 00:11:08.559 --> 00:11:29.759 I think it makes a difference that we've

operated residential real estate with a founder mentality for more than 30 years, and with a founder who's still very active uh in the company, um, because it creates a culture whereby we want to analyze all the small things we're doing constantly, right? 00:11:29.919 --> 00:11:34.399 It's a culture whereby we say, okay, you know, the small things. 00:11:34.639 --> 00:11:43.679 We want to be able to provide, let's say, uh soda, Coca-Cola, whatever, for our uh colleagues here in the office in Copenhagen. 00:11:43.919 --> 00:11:51.519 Now, we don't just buy a random number of Coca-Cola's and we don't pick the first supplier. 00:11:51.759 --> 00:12:05.919 We actually sit down every time and we analyze how many colas do we want to buy uh and where do we want to source it for, and how can we reduce the amount of uh money we spend per cola uh per employee. 00:12:06.080 --> 00:12:07.840 And it's a very small example, right? 00:12:08.000 --> 00:12:26.240 But that's the zero-based budgeting, and it's really dissecting the underlying drivers on everything from how we procure coffee or sodas in the office into how we do renting out our apartments and so on. 00:12:26.399 --> 00:12:39.440 So I think that's a very, very important part, the founder's mentality and the culture, uh, which kind of leads you to also think about how you can digitize uh in the most efficient way. 00:12:39.679 --> 00:12:43.120 So that's at least uh one point I would uh point to. 00:12:44.240 --> 00:12:48.559 You know, I think they they say that how you do the small things is how you do everything. 00:12:48.720 --> 00:13:10.879 And I think that having that system zero-based budgeting approach to every part of the business, including the procurement of uh office supplies, um, sets a tone and a foundation for the how the rest of the company should operate, uh, especially analyzing the system kind of systems within residential real estate. 00:13:11.120 --> 00:13:20.480 Um you mentioned AI, obviously being uh now something that you are leveraging to introduce more efficiency when you analyze the entire process. 00:13:20.720 --> 00:13:32.399 Can you talk a little bit about your the evolution in your approach as it first got introduced, what it was being used for, and how it's being leveraged today? 00:13:33.279 --> 00:13:33.840 Yeah. 00:13:34.080 --> 00:13:43.519 Um and I would add we're we're actually uh quite proud at Heinstaden uh in terms of how we've adopted AI. 00:13:43.679 --> 00:13:47.519 Uh we do it the same way we conduct all our business. 00:13:47.679 --> 00:13:56.080 We we have a very analytical approach, and we care about at the end of the day, utilizing technology in a profitable way. 00:13:56.399 --> 00:14:16.720 Um I would characterize it like this: that we really doubled down a year ago, and it was about a year ago, uh, here in July 207, where we got the conviction that now it it had reached the technology had reached an inflection point whereby it really made sense to uh scale up. 00:14:16.960 --> 00:14:25.600 Prior to that, we mainly used AI for, you know, let's say expense management systems that could read uh invoices. 00:14:25.679 --> 00:14:30.960 Uh, we had tested a little bit uh co-pilot, uh, but not at scale. 00:14:31.039 --> 00:14:35.360 And a year ago we said, okay, let's double down on a LLM. 00:14:35.519 --> 00:14:41.519 We chose OpenAI and we said we buy uh subscriptions for all our employees. 00:14:41.600 --> 00:14:46.799 Uh we create a team, we call it AI first, not a coincidence. 00:14:46.879 --> 00:14:50.879 It's applying first principle thinking, but with an AI lens. 00:14:50.960 --> 00:14:58.320 And that team is responsible for the training, the rollout, and for the transformation uh at scale. 00:14:58.559 --> 00:15:04.639 And there we set and again coming back to what I described before, how do we approach digital? 00:15:04.720 --> 00:15:10.159 It's a bit the same saying where are we coming from, where do we want to be, where to play, how to win. 00:15:10.320 --> 00:15:27.759 Uh, when it comes to AI, I think it's about uh data, it's about democratizing, so you know, uh enabling at scale, and then it's about finding uh the 20% of the use cases that drive 80% of the value. 00:15:28.080 --> 00:15:33.519 So today we would have, and remember we have a lot of you know, blue colours and so on. 00:15:33.679 --> 00:15:40.960 We we have uh every other employee uh using uh ChatGPT on a daily basis. 00:15:41.120 --> 00:15:56.960 Uh we have 500 credits uh per user, and the first time in May, we uh we ran out of uh credits in our uh subscription because we had started to uh use codecs increasingly and and some of the new models. 00:15:57.120 --> 00:16:18.879 Uh so I think we tick off kind of having empowered uh a lot of colleagues in using it, and then we have a team, small team, uh AI first, and then a network of champions around 40 uh that use AI for more advanced um you can say use cases. 00:16:19.279 --> 00:16:26.159 Um, so I think we're we're still early, but well on track on what you can call an AI transformation. 00:16:26.320 --> 00:16:34.720 I think it goes into a new phase where it's also being really, really smart about how do we apply AI in a profitable way? 00:16:34.879 --> 00:16:39.679 How do we manage the uncertainty about the pricing mechanisms uh going forward? 00:16:39.840 --> 00:16:42.159 What will a credit cost in two years? 00:16:42.399 --> 00:16:44.320 How many credits will we consume? 00:16:44.559 --> 00:16:52.159 What does that mean for the trade-off between what you automate versus what you still want to have colleagues who do? 00:16:52.399 --> 00:17:09.279 I think you know there is so much uncertainty in this space, but if you do it in an analytical way, I think you're building the right foundation for also being agile to navigate uh, you know, whatever happens in this space uh going forward. 00:17:09.519 --> 00:17:16.160 Um yeah, it sounds amazing the way that you and the

organization have approached it. 00:17:16.319 --> 00:17:17.920 There's a lot of questions I have. 00:17:18.079 --> 00:17:24.000 Um maybe we start with just kind of this AI first team that you you've put together. 00:17:24.240 --> 00:17:31.599 Tell me a little bit about the makeup and how they've been able to roll out this transformation at scale. 00:17:31.680 --> 00:17:51.039 Because in in some other previous conversations I've had, sometimes there's some resistance, sometimes there's a knowledge gap, there's an agency gap, there's an art of the possible gap, especially across larger organizations in terms of what one team is doing and how they another can really uh learn from those examples. 00:17:51.200 --> 00:17:53.039 There's a lot of different moving parts. 00:17:53.200 --> 00:18:01.039 So tell tell me a little bit about your team, their mandate, and how they've been able to, I guess, cross the chasm on those areas. 00:18:02.000 --> 00:18:03.200 That's a great question. 00:18:03.519 --> 00:18:13.519 Um so the team, we have uh three people in the uh in the team, two full-time, one student. 00:18:13.680 --> 00:18:28.799 Uh the characteristics of the team is they're just, you know, call it generalists uh who would be project managers or business development or consultants, but with an interest in technology. 00:18:28.880 --> 00:18:31.039 Um, and I think that's very deliberate. 00:18:31.200 --> 00:18:35.119 Uh I don't think you need someone to be an expert on AI. 00:18:35.200 --> 00:18:36.480 I think it moves very quickly. 00:18:36.559 --> 00:18:48.480 It's more about change management and transformation, but you have to have a knack for technology and be genuinely uh interested and you know, read up uh and test things yourself. 00:18:48.720 --> 00:18:57.279 Um that core team uh has focused a lot on uh empowering the organization. 00:18:57.440 --> 00:19:00.960 So we we created training modules. 00:19:01.119 --> 00:19:09.839 I'm usually not a big fan of corporate learning, but I thought for AI and just having a basic introduction to AI. 00:19:10.000 --> 00:19:14.960 Here is ChatGPT, here is how you sign up, here is do's and those, don'ts. 00:19:15.119 --> 00:19:18.880 Also, from a compliance perspective, I think worked really well. 00:19:19.039 --> 00:19:21.200 Then we used that as a certificate. 00:19:21.359 --> 00:19:27.119 So once you've taken it, you get a enterprise license for uh ChatGPT. 00:19:27.359 --> 00:19:40.240 We encouraged people to also use ChatGPT and whatever OpenAI offer for private usage, meaning that you know we told people if you have to plan a vacation, try with ChatGPT. 00:19:40.319 --> 00:19:41.440 It's super helpful. 00:19:41.519 --> 00:19:51.519 Uh whatever private uh things, because we think the more people play around with it, the more they understand the frontier, what it's good at, what it's not good at. 00:19:51.759 --> 00:19:56.720 So I think uh then in addition, we uh we've made competitions. 00:19:56.960 --> 00:20:02.240 So, you know, the colleague who made the best custom GPT and shared it. 00:20:02.319 --> 00:20:08.960 Uh you could nominate a colleague, you were allocated an uh amount uh if you won. 00:20:09.119 --> 00:20:12.960 We had four winners that you could then spend with your team. 00:20:13.200 --> 00:20:17.599 Um we have some internal communication. 00:20:17.839 --> 00:20:22.000 We have been very sort of focused on that from the beginning. 00:20:22.240 --> 00:20:24.480 I think people respond well to stories. 00:20:24.640 --> 00:20:27.680 So start storytelling is super important here. 00:20:27.920 --> 00:20:40.480 Get the use cases out, get pictures of real people working with real tasks, smiling because they had AI helping them do something which was just a hassle to do manually. 00:20:40.720 --> 00:20:46.400 And having people who see that recognize themselves in this and saying, okay, well, maybe this is also for me. 00:20:46.640 --> 00:21:01.200 Because if you look around in an office and you see all the people sitting in front of a computer typing something in, a high percentage of the time what they're doing is a test they don't really want to do because it's not stimulating. 00:21:01.359 --> 00:21:08.240 Someone just have to do it, and quite often AI is a tool to do that more uh efficiently. 00:21:08.559 --> 00:21:17.119 Um and then the final point, and now I'm I I recognize here I have long unstructured uh answers. 00:21:17.519 --> 00:21:18.720 No, this is extremely helpful. 00:21:18.960 --> 00:21:19.839 That works. 00:21:19.920 --> 00:21:25.920 Uh the last point I want to make is we quite early created a community of what we call champions. 00:21:26.079 --> 00:21:31.839 Uh now we have 40, and I highly encourage you know organizations working with AI to have that. 00:21:32.000 --> 00:21:33.279 And how do we select them? 00:21:33.440 --> 00:21:40.559 Well, we didn't select them, you just know who's interested and who's going to build some more uh cool stuff. 00:21:40.720 --> 00:21:48.000 Uh we the only thing we said is we wanna we're in nine countries and we have everything from you know blue collar to investment people. 00:21:48.160 --> 00:21:59.759 So we want to have different profiles in this community, and then we have meetings uh at least once a month with extra training, rolling out codecs, uh, and and having more advanced trainings. 00:22:01.119 --> 00:22:03.119 So that created a good community. 00:22:03.359 --> 00:22:06.960 The very last point I want to make, sorry, and I think this is important. 00:22:07.519 --> 00:22:13.839 A year ago, when we said, let's create an AI first team, we also said, we're a real estate company. 00:22:14.000 --> 00:22:21.519 Are we really going to be at the forefront of AI and knowing what are you know the best tools available and so on? 00:22:21.759 --> 00:22:22.079 No. 00:22:22.640 --> 00:22:25.920 There are consultants, there are specialists with this. 00:22:26.400 --> 00:22:39.759 So what we

did is we signed a retainer 12 month with a company called Hype Mechanics, Boutique AI Consultancy, 10, 15 people advising us in the beginning. 00:22:39.920 --> 00:22:47.279 And they're introducing awesome tools like N8N and other things, saying, you know, you could try this for this use case. 00:22:47.359 --> 00:22:58.480 And we had some bigger, more complex uh projects we collaborated on, the AI first team, Hive Mechanics, and you know, potentially people in the champion community. 00:22:58.720 --> 00:23:05.680 So I think there are many things you need to think about if you want to do an AI transformation. 00:23:05.839 --> 00:23:11.279 Uh but it's not different compared to many other transformations, right? 00:23:11.359 --> 00:23:19.599 And it comes back to understanding where you're coming from, where do you want to go, and what does that mean for what you prioritize and how you win there. 00:23:20.559 --> 00:23:21.599 I think this is great. 00:23:21.839 --> 00:23:28.720 Really good uh framework and thinking for others that are may either be considering it or going through it themselves. 00:23:28.960 --> 00:23:35.119 Um, another piece that I wanted to uh talk about a little bit was the data side of things. 00:23:35.279 --> 00:23:46.400 Because underlyingly, in all the conversations that I've had, in order to get the highest leverage, you need to make sure that you have quality, clean data across the different parts that AI is working in. 00:23:46.640 --> 00:23:55.519 Um, what has that part of the transformation been like in terms of working with the different teams, making sure there's good data hygiene definitions, et cetera? 00:23:58.720 --> 00:24:00.799 You can say it's a journey, right? 00:24:00.880 --> 00:24:05.599 Uh and I think all companies would uh attest to that. 00:24:05.839 --> 00:24:11.119 Uh I personally think data is hard and it's super interesting. 00:24:11.200 --> 00:24:19.759 And I think I have the privilege of approaching it not as a data scientist, but also sometimes asking, you know, the stupid uh questions. 00:24:20.160 --> 00:24:22.559 Um, we talked about it prior to this call. 00:24:22.720 --> 00:24:27.680 You asked me what is the data landscape like uh in real estate? 00:24:27.839 --> 00:24:29.759 And I thought it's a great question, right? 00:24:30.000 --> 00:24:35.599 Because how do you even think about the dimensions in a data uh landscape? 00:24:35.759 --> 00:24:56.960 And asking that question to data scientists uh at Heimstadt who are really good and and and we do a lot of great stuff, but sometimes just getting a simple question from the side means that you you you have to take a step back and really think about the architectural structure of your uh you know data. 00:24:57.279 --> 00:24:59.279 Uh where does data come from? 00:24:59.440 --> 00:25:01.440 Where do you have your semantic models? 00:25:01.599 --> 00:25:03.920 How do you distribute it in the business? 00:25:04.160 --> 00:25:10.240 Um and and I think you know this is a complex, uh this is a complex area. 00:25:10.400 --> 00:25:19.599 Um I think at the core of it, right, we have buildings, they have data, uh, we have contracts with people who live there who are the tenants. 00:25:19.759 --> 00:25:22.640 And then within these pillars, a lot of stuff happened. 00:25:22.880 --> 00:25:32.400 Like a tenant has a force, as you said, uh, that is broken, that creates a ticket, and then you know, a lot of data derived from there. 00:25:32.799 --> 00:25:38.079 Uh, we have had a, and now I come back to history and culture. 00:25:38.240 --> 00:25:55.759 We've had a culture of obsessing with knowing facts, so actually spending a lot of time training people that you need to insert this data point, even if you can't see an immediate benefit of it, you still need to insert it because we will need it in the future. 00:25:56.000 --> 00:25:57.519 So that's a cultural part. 00:25:57.599 --> 00:26:09.279 Uh, I think one step further is saying we devise a system that runs a process whereby there is no way of avoiding having to record this data point that we uh know we need. 00:26:09.599 --> 00:26:11.759 Um, and that's an evolution. 00:26:12.000 --> 00:26:27.599 Then when you have the data, uh you build the models, semantic models, because you have you will always have different systems of records where you need to combine them to conduct an analysis that the business uh needs. 00:26:27.839 --> 00:26:31.359 So that work, uh we're on Microsoft Fabrics. 00:26:31.519 --> 00:26:47.440 We have invested heavily in in the past you know five years or even before that, um, which I believe puts us in a really good place because it means now that we have a lot of data, we can serve uh AI. 00:26:47.599 --> 00:26:52.480 And I think that's really where AI becomes interesting. 00:26:52.640 --> 00:27:12.799 And I don't think we're there yet, but I don't think we're far away from having that junior analyst sitting on the side who can look up you know data, say you're curious about what is the vacancy in this specific region in northern uh Sweden and how has it developed the last five years. 00:27:13.119 --> 00:27:26.559 Today, you would have to ask that question to someone who would go into a system, download the data, make an analysis in Excel, maybe put it on a PowerPoint page. 00:27:26.799 --> 00:27:30.000 And that would take at least a couple of hours, right? 00:27:30.640 --> 00:27:40.400 I think we're not far away from you can just post that question in a chat interface or whatever you prefer, and then you get a high-quality answer. 00:27:40.559 --> 00:27:49.680 And I think once you get there, you you really come to a different place as a business, especially as a real estate business. 00:27:49.839 --> 00:27:54.720 Um, and and that's you know part of the future we're seeing in too. 00:27:54.799 --> 00:27:56.400 And I don't think it's too far away. 00:27:56.559 --> 00:28:02.160 And it's it's not binary, it's a you know, gradual

progression towards it. 00:28:02.640 --> 00:28:04.079 Yeah, I don't think so either. 00:28:04.160 --> 00:28:09.440 And uh, but it does, I think to kind of the point you're making, it has to be very intentional. 00:28:09.599 --> 00:28:11.599 And it's a cultural thing almost. 00:28:11.680 --> 00:28:18.400 Like, you know, we're going to be looking and needing this data in the future for some of the future questions that we might uh be asking of it. 00:28:18.640 --> 00:28:21.119 So um, yeah, I think the approach is fascinating. 00:28:21.200 --> 00:28:27.279 And like you said, everyone has their own very uh windy journey with data within their organization. 00:28:27.599 --> 00:28:44.000 Um, one final part I wanted to, or thread I wanted to pull on based on what you said, is you were saying that it's obviously there's so much leverage you can get with AI, but the challenge is finding those 20% of use cases that provide 80% of the value. 00:28:44.240 --> 00:28:49.759 And I'm curious how in practice um that gets exercised at Heimstaden. 00:28:50.400 --> 00:28:53.519 Um, and also how that gets measured uh as well. 00:28:53.680 --> 00:28:55.920 Like, how do you know the value is being created? 00:28:56.079 --> 00:28:57.279 Do you have a committee? 00:28:57.519 --> 00:29:01.599 Is there uh a council that like gets together to review this stuff? 00:29:01.839 --> 00:29:06.640 I'd love to hear any um tactical feedback that you have around doing that. 00:29:08.880 --> 00:29:23.279 Um I really like when people can answer uh questions briefly, but I think you asked the brief questions, but which are very good and therefore does require a longer answer. 00:29:23.440 --> 00:29:33.119 Uh how do you find uh the right cases to prioritize with AI, or how do you find the places in the business where there is potential? 00:29:33.279 --> 00:29:36.880 I I think it is a great question, and it's it's not an easy one. 00:29:37.039 --> 00:29:45.440 Um the way we think about it is there are different levels when it comes to uh AI value creation. 00:29:45.839 --> 00:29:49.039 There is the, you know, let's say there are three levels. 00:29:49.200 --> 00:29:53.039 Level one is just the personal productivity gain. 00:29:53.279 --> 00:30:00.640 Let's say it's 10% more productive from you know making better research, uh, drafting communication. 00:30:00.880 --> 00:30:06.079 Everyone can do it by just downloading an LLM and start using. 00:30:06.559 --> 00:30:14.160 Then there is the level two where you create, let's say, simple tools that bring significant value to your team. 00:30:14.319 --> 00:30:18.240 Like I said, we create 90 advertisements per day. 00:30:18.480 --> 00:30:23.759 Uh if every single one has 10 pictures, that's 900 pictures. 00:30:23.920 --> 00:30:33.119 The difference between having you know nice pictures where the lighting is correct, and nowadays also having virtual styling. 00:30:33.200 --> 00:30:42.640 So it's not just an empty room, but you virtually place, you know, uh whatever a bed or a table or something like that, is a super easy task for AI. 00:30:42.799 --> 00:30:48.559 You can create a simple tool that takes the raw picture and converts it into what good looks like. 00:30:48.799 --> 00:30:57.200 And and you know, that's a simple single prompt, but then you can share it across all hundred uh leading officers at Hangston. 00:30:57.279 --> 00:30:59.680 That's what I would say is uh is a level two. 00:31:00.000 --> 00:31:01.519 How do we find these cases? 00:31:01.759 --> 00:31:12.960 Well, I think you stimulate uh the 40 people uh in the champion community that I mentioned before, and then people will naturally find these uh things. 00:31:13.200 --> 00:31:20.160 Then you have the level three, which I think is where you start really redefining how you do things. 00:31:20.240 --> 00:31:24.000 Uh and I'll just take the easy one, right, which is customer care. 00:31:24.240 --> 00:31:26.720 I think it's a lot harder than many others. 00:31:26.799 --> 00:31:36.480 Uh, but everyone can see a future where you know AI can not only just answer questions but also initiate uh processes, right? 00:31:36.559 --> 00:31:52.640 So you have a faucet uh that is damaged, and you can just call somewhere or take a picture, and then you can you know create that uh ticket, and and the problem will be recorded in our systems and and and solve itself. 00:31:52.799 --> 00:32:01.119 That's where you take, let's say, an end-to-end workflow uh and and you completely redesign how it's done. 00:32:01.279 --> 00:32:05.039 Uh, and AI is basically going to uh automate it. 00:32:05.359 --> 00:32:13.519 And that's where you know we will have a lot of tasks where we will no longer need humans in the future. 00:32:13.680 --> 00:32:21.680 Hopefully, humans find other things to do, but there will just be a lot of the things we do today where I'm sure you won't need uh humans. 00:32:21.839 --> 00:32:35.039 And I and I think, you know, for the level three, I think you need to stimulate some brainstorming where you need to, you know, look at your business in in a different way. 00:32:35.119 --> 00:32:40.799 And I I think that's workshops and creative people and and interviews. 00:32:40.960 --> 00:32:48.000 Uh the level two is where you can ask a lot of subject matter experts how are you doing? 00:32:48.079 --> 00:32:49.440 Could you do it differently? 00:32:49.599 --> 00:32:55.759 Um, but I think you know how you find them depend on what level you're looking at. 00:32:56.000 --> 00:33:00.240 That was kind of uh the long answer to one of your questions. 00:33:00.400 --> 00:33:12.319 Then you had a second second one, which I think is equally important, and which I mean I would be happy to hear if anyone has uh found a silver bullet for how to do it, which is to say how do you measure impact? 00:33:13.119 --> 00:33:25.359 Um and we're firm believers that if if you can't measure things, you can't improve them, like we really care about

measuring results and outcomes. 00:33:25.440 --> 00:33:30.880 Uh the question is, what is what was the root cause of something you did better? 00:33:31.039 --> 00:33:38.000 And and when can you say, point at the AI and say that's exactly what led to uh to this outcome? 00:33:38.160 --> 00:33:40.559 I think it's very hard in in practice. 00:33:40.720 --> 00:33:54.319 You can easily do estimations like saying, okay, if I roll out this tool to a thousand people and they say they become 10% uh more productive, then the value is you know, whatever 10% of the combined uh salary. 00:33:54.480 --> 00:33:55.680 That's one way of doing it. 00:33:55.839 --> 00:34:13.599 You could also say uh back to the example with the pictures, you can say, okay, if we have better pictures that are virtually styled and just more visually appealing, I think on average I will get 0.5% higher rents or whatever lower vacancy. 00:34:13.840 --> 00:34:20.320 Or, you know, before a letting officer spend 10 minutes per picture refining it, whatever. 00:34:20.480 --> 00:34:27.360 Like you can make all of these uh estimations, and I think you should, because it informs what you prioritize. 00:34:27.599 --> 00:34:30.639 Uh, but at the end of the day, it's really, really hard. 00:34:30.719 --> 00:34:33.440 And I think you don't get away from business judgment. 00:34:33.599 --> 00:34:41.280 And then I think as a leader, you at the end of the day have to say, I have a hard time you know proving this 100%. 00:34:41.920 --> 00:34:44.960 I just need to second guess constantly. 00:34:45.119 --> 00:34:54.880 And then I, as a leader, have to have enough conviction to say, I just know this uh creates value because I can see the cost, right? 00:34:55.039 --> 00:35:06.960 Uh and and and then the final point, and I think a lot of people, especially in tech, underestimate this storytelling and marketing of what you do is super, super important. 00:35:07.440 --> 00:35:10.559 So you also have to find those small wins. 00:35:10.880 --> 00:35:21.199 You know, just spend half an hour saying, I make a nice picture and I describe what I did and the outcome, and then I send it around and I say, look at what we do here. 00:35:21.360 --> 00:35:36.880 Uh, because the alternative is a lot of people are sitting, you know, in the tech area and they're doing a lot of stuff, and then someone from the business comes a quarter later and says, What have we actually done that uh created a real business impact? 00:35:37.039 --> 00:35:41.760 And then they're they're sitting there and have a hard time articulating it, uh, right? 00:35:42.079 --> 00:35:49.760 Um, so I guess what I'm also saying is uh, how do you know if you created value? 00:35:49.840 --> 00:35:51.280 Uh it's hard. 00:35:51.679 --> 00:36:05.119 But people working with tech and AI have a responsibility for at least making the case constantly that it does provide uh uh value to the business. 00:36:06.079 --> 00:36:10.639 Yeah, you know, even though I'm asking short questions, you're giving very good, comprehensive answers. 00:36:10.719 --> 00:36:12.480 And I think there's a couple things in there. 00:36:12.639 --> 00:36:20.480 You know, you kind of talked about your own maturity model around how you think about AI adoption and how to measure kind of within each one of the different levels. 00:36:20.719 --> 00:36:38.559 I also think another piece that you mentioned that answers a former question was just about always being um, you have your, of course, 40 champions in the organization, but constantly sharing the work, not only just to show the art of the possible, but because you can get so much more done with AI, um, you actually forget. 00:36:38.719 --> 00:36:46.239 I I find it like I there's so much that I'm getting done that I'm like, I have trouble recalling all of the things that it assisted with. 00:36:46.400 --> 00:36:55.920 So constantly being able to share will not only help inspire other people, but it will validate okay, we are getting a lot of value here, and there's a lot of leverage to be had. 00:36:56.400 --> 00:37:00.079 Um talking about the go no, please. 00:37:00.800 --> 00:37:03.199 Just a quick one, just a practical example. 00:37:03.280 --> 00:37:11.920 Uh Andres, who's heading our AI first team, he has, and it's just screenshots from emails or you know, one page of PowerPoint. 00:37:12.159 --> 00:37:29.360 He has a library with more than a hundred of these small examples just coming after each other from real people where he can say, you know, this is an example of uh something AI enabled, where people uh actually wrote me and said, This works really well. 00:37:29.519 --> 00:37:31.440 Here is what I did and what I got out of it. 00:37:31.599 --> 00:37:37.119 So having that library, I I really encourage organizations to build and keep. 00:37:37.599 --> 00:37:38.239 Yeah. 00:37:38.559 --> 00:37:39.840 Very good feedback. 00:37:40.079 --> 00:37:45.360 Um, as we start to close, um, we've talked about the measurement of impact. 00:37:45.599 --> 00:37:56.719 And to look at it holistically, you kind of talked about a destination state, um, you know, and uh taking a real inventory of where you are and really breaking it apart that way. 00:37:56.960 --> 00:38:03.360 You know, F1, we we keep coming back to this 67 seconds in 1950, two seconds today. 00:38:03.679 --> 00:38:12.719 On that journey, in your mind as the chief uh or as the CDO, where do you feel like you are today from that 67 to two seconds? 00:38:12.800 --> 00:38:16.000 And how how much further do you feel like you have to go? 00:38:17.199 --> 00:38:19.679 Um it's a great question. 00:38:19.840 --> 00:38:25.840 But but you know, you also know the two seconds is not the end state uh in the Formula One, right? 00:38:25.920 --> 00:38:30.880 Like we we have no idea how we'll get there, but it will be you know less than one second. 00:38:31.039 --> 00:38:36.239 And and by the way, I just want to add, it's not Formula One uh

alone. 00:38:36.400 --> 00:38:45.039 If you look at Turifang's average uh speed cycling, a hundred years ago it was 36 uh kilometers per hour. 00:38:45.119 --> 00:38:46.559 Now it's like 44. 00:38:46.719 --> 00:38:50.880 Uh, we just beat in the marathon running, going below two hours, right? 00:38:50.960 --> 00:38:53.840 It's a human being with a human body running. 00:38:54.079 --> 00:38:57.360 It was around three hours uh a hundred years ago, right? 00:38:57.440 --> 00:39:01.119 And and you just know it's going to be faster in the future. 00:39:01.360 --> 00:39:08.719 Um I so I think I I have no clue, right? 00:39:08.800 --> 00:39:20.639 Uh I I think it it is there a physical constraint that means we wouldn't be able to run our business, you know, uh with systems, technology, AI without any humans? 00:39:20.800 --> 00:39:22.159 Uh no, I don't think so. 00:39:22.320 --> 00:39:23.440 I I think that's possible. 00:39:23.599 --> 00:39:25.519 So that's one way of looking at it, right? 00:39:25.599 --> 00:39:27.679 It's that is that the goal in itself? 00:39:27.920 --> 00:39:32.000 No, because I'm sure there will be a lot of uh other things to do. 00:39:32.239 --> 00:39:40.400 A hundred years ago, uh, or maybe a bit more, 38% of the US population worked with agriculture, right? 00:39:40.480 --> 00:39:41.599 Today it's less than 1%. 00:39:42.000 --> 00:39:44.400 People found different things to do. 00:39:44.559 --> 00:39:47.360 So I'm not terribly uh concerned about that. 00:39:47.599 --> 00:39:56.559 I think in real estate, it if you made it from the beginning uh to the end, I think we're quite early in that journey, maybe halfway. 00:39:56.719 --> 00:39:59.679 Uh I think we're the best one out there, Heimstadt. 00:40:00.000 --> 00:40:00.960 I I really do. 00:40:01.119 --> 00:40:06.960 But but I see a lot of uh potential uh to become a lot faster. 00:40:07.119 --> 00:40:24.559 And you know, when I worked at Bain and Company, they had this graph with digitization across industries, uh, having TMT typically as the most digitized maybe financial institutions, and you would have physical assets like real estate at the very bottom of this. 00:40:24.719 --> 00:40:26.320 And I think that is right. 00:40:26.480 --> 00:40:30.000 There is a reason why I can be a CDO at Heimstadt. 00:40:30.239 --> 00:40:36.000 It is because to begin with, we're not in the most uh complicated uh and digitized industry. 00:40:36.239 --> 00:40:44.639 Because if we were, uh I would probably have to uh up my technical uh capabilities to qualify, right? 00:40:44.719 --> 00:40:52.159 Uh I'm sometimes joking that you know, amongst the blind, even the one-eyed uh has a great uh side. 00:40:52.320 --> 00:41:01.039 Um and the point being that real estate is is really quite analog, uh as a stereotype, at least. 00:41:01.199 --> 00:41:02.800 Um yeah. 00:41:03.119 --> 00:41:12.480 So as a as a leader uh trying to navigate this uh modernization on this continuum that we're talking about. 00:41:12.639 --> 00:41:20.719 And like you said, you know, the ingenuity, the human spirit, what we're able to push not only in F1 and beyond is is incredible. 00:41:20.880 --> 00:41:28.239 I was actually watching a video on um marathon times and how it's just like continued to reduce throughout the past couple days. 00:41:28.400 --> 00:41:29.360 I don't I don't get it. 00:41:29.599 --> 00:41:36.159 I know there's efficiency, I know there's clothing, I know there's different types of exercises, but I still find it amazing. 00:41:36.559 --> 00:41:38.639 But but and I just love it, right? 00:41:38.719 --> 00:41:47.599 Because what I love the most is when there was something people thought were just true, like you have to train in this way or you have to eat in this way. 00:41:47.679 --> 00:41:52.800 And then someone raises her hand and says, but have we do we actually know? 00:41:52.960 --> 00:41:56.239 Have we measured it and have we tried something else? 00:41:56.480 --> 00:42:04.480 Um and and it's often that you know, the story that people just buy into, it feels intuitively correct. 00:42:04.719 --> 00:42:08.159 Like, you know, the more you train, the better you become, for instance. 00:42:08.320 --> 00:42:12.559 But then someone says, But what if I train less, but I train in a different way? 00:42:13.039 --> 00:42:14.239 Okay, let's measure it. 00:42:14.320 --> 00:42:15.119 Does it work? 00:42:15.360 --> 00:42:25.519 And and and I think that's you know what is so inspiring about sports, but also I think very much an analog for business, certainly ours. 00:42:25.920 --> 00:42:32.559 Um, and you know that that's a mindset, and that's why we come keep coming back to culture. 00:42:32.960 --> 00:42:45.039 Um and technology is just an enabler of being able to do it uh the right way, but it really starts with the mindset uh, right, that you have to ask the right questions. 00:42:45.199 --> 00:42:53.199 Um 100% and not be afraid to challenge the status quo, be that first principles thinker and say, why why is it done this way? 00:42:53.360 --> 00:42:55.360 Why does it have to cost so much money? 00:42:55.599 --> 00:43:00.800 Um, yeah, very uh we could have a whole other podcast on that topic. 00:43:01.039 --> 00:43:10.960 Um as we close, I'm curious about you personally, what your use of uh AI looks like. 00:43:11.280 --> 00:43:15.920 And as you continue, let me frame it a little bit. 00:43:16.320 --> 00:43:23.519 I have found that with myself, it used to be about delegating as much as I possibly could to AI. 00:43:23.840 --> 00:43:31.760 And then I found that for certain things, I was maybe losing the development of a muscle around critical thinking. 00:43:32.000 --> 00:43:43.679 And I knew I was like, I actually want to be involved in this because I'm losing the nuanced details that make either an output, an article, or a conversation that I'm about to have uniquely mine. 00:43:43.920 --> 00:43:47.280 So I've had to adjust my use with AI.

00:43:47.360 --> 00:43:50.880 And I'm curious as to what your personal journey has been. 00:43:51.760 --> 00:43:52.320 Yeah. 00:43:52.480 --> 00:43:56.960 Um, it's it's a really, really good question. 00:43:57.039 --> 00:43:59.760 And I think it's something so important for people to. 00:44:00.159 --> 00:44:01.199 reflect upon. 00:44:02.559 --> 00:44:22.880 I see a great risk not only for me or Heimstaden but also for society that AI creates a lot of noise that it uh you know the the brain starts working differently uh because you outsource the the critical thinking. 00:44:23.440 --> 00:44:46.320 Um I I fully subscribe to there are a lot of things you can do a lot you know uh quicker like writing code productivity you know it's it's perfectly clear that's that's become you know you can write a lot more code right and I think what we're seeing is it's going from software uh engineering and writing code into business. 00:44:46.559 --> 00:44:54.000 So as a business person I just see it's going to be the same with Excel and PowerPoint and all the things that the business people know about. 00:44:54.079 --> 00:44:56.400 It starts with coding and then it moves in there. 00:44:56.480 --> 00:45:00.639 So we have this amazing productivity game. 00:45:01.039 --> 00:45:10.159 But I think it's a big risk if people just focus on that and get overwhelmed with the amount of things you can now produce. 00:45:10.320 --> 00:45:20.079 And therefore if I'm being honest I'm very reluctant with outsourcing anything that I think sharpens my brain. 00:45:20.320 --> 00:45:23.280 So I'm using AI a lot for learning. 00:45:23.440 --> 00:45:41.519 I think it's incredible at that but it's based on your own sort of curiosity and asking the right uh questions and then think then I think there are very sort of mundane tasks where I know exactly how this should be done but it just takes time to do. 00:45:42.159 --> 00:45:45.519 And I come from a real estate background. 00:45:45.679 --> 00:45:49.519 So let's say I want to do an investment model in Excel. 00:45:49.599 --> 00:45:57.199 I I know perfectly well how to do it but it would take me five hours because you you need to put it into Excel and so on. 00:45:57.360 --> 00:46:14.880 That you can prompt your way into I'm not afraid of outsourcing that but if I'm thinking about how to transform our business with AI and I write a prompt how should I as a CDO strategize about AI? 00:46:15.039 --> 00:46:17.599 I'm sure I would get something quite convincing. 00:46:17.760 --> 00:46:36.559 I'm sure I could share it with others and they would look at it and think that makes a lot of sense but then I I wouldn't really own it and I'm just afraid of really losing out on the yeah my my core ability to think myself and and problem solve and understand things. 00:46:37.199 --> 00:46:38.719 So I'm very concerned about that. 00:46:38.800 --> 00:46:43.760 And that also means I I just I guess this is a longer way of echoing what you're saying. 00:46:43.920 --> 00:46:50.480 I'm afraid of outsourcing things that requires critical thinking. 00:46:50.880 --> 00:46:51.360 Yeah. 00:46:51.679 --> 00:47:09.119 Yeah that that quote that's been going around you can outsource your thinking but you can't outsource your understanding and I think the thing that makes us uniquely human and uh brings that human judgment layer to everything is our own personal experiences, our own personal views. 00:47:09.280 --> 00:47:21.119 Because ultimately the AI, even if you ask it hey the uh strategic direction for the company, it's going to pull on all of the previous data that it has, which is examples from its past. 00:47:22.159 --> 00:47:31.920 And why then while they might be good to reference, uh I think coming up with unique novel first principled ideas are uniquely your own. 00:47:32.159 --> 00:47:34.559 And um so I totally agree. 00:47:34.719 --> 00:47:37.519 And I think that's uh that's a perfect place to end. 00:47:37.679 --> 00:47:42.480 So uh Nikolai this has been a fun conversation uh really enjoyed it. 00:47:42.559 --> 00:47:48.000 Uh if people want to learn more about Heimstaden and what you all are up to where can they go? 00:47:49.840 --> 00:48:11.760 I guess our website maybe LinkedIn or give us a call send an email um we we have a lot of great apartments you can rent but it does require you to uh be European okay sounds great well Nicol I thank you so much for the conversation today I I thoroughly enjoyed it thank you have a good one bye bye