How much of construction's cost, delay, and waste begins with information that fails to survive the journey from design to delivery? That question runs through my conversation with Julian Geiger, Chief AI Officer at Nemetschek Group, as we look at the practical role of AI across architecture, engineering, construction, and operations.
Julian describes what he calls the industry's 90, 40, 20 problem. According to the figures he shares, 90 percent of projects are over budget or over time, the built world accounts for roughly 40 percent of global carbon emissions, and around 20 percent of material is wasted. His argument is that many poor outcomes start as information and decision problems. Each project phase may work reasonably well on its own, but handovers can strip away context. A building information model becomes a PDF, a PDF becomes an email, and a decision may never be recorded against the object it changed.
We discuss how AI, building information modeling, and digital twins can identify missing information, scope gaps, clashes, and design choices before they become expensive construction site problems. Julian shares Nemetschek's work bringing Firmus AI into Bluebeam to review two dimensional drawings, then explains the broader goal of feeding lessons from construction back into design and engineering tools. The commercial promise is easy to understand. Finding a mistake while a wall exists only in software costs far less than finding it after workers and materials are waiting on site.
The conversation also moves beyond the assumption that every task needs the largest available model. Julian sets out a four tier approach. Deterministic calculations such as structural math should remain deterministic. Stable, high volume checks may be handled by conventional rules. Smaller domain models can classify objects, retrieve data, and interpret geometry close to the source. Frontier models earn their place when the work involves ambiguity, reasoning across documents, or several dependent steps. His test is refreshingly practical: use the least expensive method that is reliably right and fast enough for the person waiting on the answer.
That discipline matters when finance teams ask for proof. Time saved on drawing reviews or tender preparation can be measured quickly, while reductions in rework or missed issues require a longer data series. Julian also notes a familiar problem for enterprise AI programs. If a firm never established a baseline, it becomes difficult to show what improved. Usage can indicate that people find a tool useful, but adoption alone does not settle the return on investment question.
Data sovereignty adds another layer. Construction files can include valuable designs, commercial information, and details tied to national infrastructure. We discuss where the data is stored, who processes it, which jurisdiction applies, and whether customer material is used for model training. Julian argues for separating genuine intellectual property from routine usage data, then matching controls to the sensitivity of each project rather than treating every data set as identical.
Finally, we consider people. In an industry facing a skills shortage, removing junior roles creates a future shortage of experienced professionals. Julian sees AI as a way to shorten the apprenticeship period and reduce repetitive documentation, while preserving a clear line of accountability: AI proposes and a qualified human decides. Could that model help construction professionals spend more of their time on judgment, design, and better buildings, and where should the industry draw the line? Listen to the episode and share your thoughts with me
[00:00:00] The leading issue of agentic AI in businesses right now is ensuring agents act with compliance guidelines. And Denodo applies guardrails across your entire data estate. By aligning your company's data infrastructure under one system, these guardrails perform consistently across your platform. So start scaling your business and start with Denodo. Simply visit denodo.com to learn more.
[00:00:27] What happens when every building is a prototype? Every handover loses information and mistakes become far too expensive once crews and materials reach the site. Well, today I'm going to be joined by Julian Geiger, Chief AI Officer at Nemetschek Group.
[00:00:52] And together we're going to examine where AI can make a practical difference across architecture, engineering, construction and building operations. So we'll discuss why some calculations should never be handed to a language model. How digital twins and building information modelling can actually expose problems earlier. And what reliable AI might look like when firms lack a clean baseline.
[00:01:21] And Julian will also explain why data sovereignty and intellectual property all need clearer treatment. And why the best operating model might be simple. AI proposes while a qualified human decides. So if you work anywhere near enterprise AI, that distinction alone is worth hearing. And on that note, let me introduce you to Julian right away. So thank you for joining me on the podcast today.
[00:01:51] Can you tell everyone listening a little about who you are and what you do? Yeah, great to be on today and thanks for having me. I'm the Chief AI Officer at Nemetschek Group. I'll talk about the combat in a moment. And my mandate is twofold. One, make our software AI first for our customers. And secondly, make the company itself AI first internally. And I joined the company about two and a half years ago. I came from Google originally. And we're led to manufacturing as a product vertical globally.
[00:02:18] And now really translate that learning into the AEC, the architecture, engineering and construction industry here at Nemetschek Group. Well, it's a pleasure to have you join me on the show today. There's a lot I want to talk with you about, especially around the world of construction. Because on this podcast every day, I try and get people thinking differently about the ways that technology impact their lives, their work and their world. And construction is typically one of those areas that we don't often think about.
[00:02:45] But construction is responsible for substantial waste, pollution, delays and cost overruns, of course. So which of these kind of problems can AI or technology realistically address today? And which are primarily organisational still? What are you seeing here? That's a really great question. And it's actually the question that led me also to join the Nemetschek Group ultimately. When I look at the industry, it's really perplexing.
[00:03:13] We have an industry that has projects that are 90% over budget and over time. 90%. I think few other kind of industries would accept that kind of KPI. We have 40% roughly of global CO2 emissions coming out of the build world from the buildings we construct and how we operate them. And 20% of material is wasted in the industry.
[00:03:36] That is a massive KPI set that pretty much would be unacceptable in any other industry, certainly in manufacturing. So why is this the case and what can AI fix? So I think the core issue of why this 90-40-20 problem exists is an information problem.
[00:03:55] These are decision problems that happen because information does not flow between the different participants and because incentives are often not aligned properly between the incentives that you have in the value chain of the construction industry. AI can help in many ways to change all of that and lead into a new equilibrium. We have a paradigm shift underneath the working models of the industry of the past.
[00:04:23] I think AI is something very different from any other technology that we had before. Cloud, for example, just made the same processes happen a little bit more efficiently or made tools suddenly live in the cloud so you didn't have to have a desktop computer that is as powerful as one before. These are all incremental changes. But now with AI, you can really address this in a very, very different way.
[00:04:46] So, for example, looking at rework, direct rework is about 5% of the total construction cost in the industry and 10% roughly if you account for all the indirect costs. That's a big thing that usually comes from information breakdown and identifying problems as early as possible in the value chain, i.e. already in the design stage that what the architect, for example, is designing is not constructible. This is something that AI can do if we build it smartly.
[00:05:16] And if you catch it early, there's very low cost or zero cost to fix the issue. While if you catch it late while it's already on the construction side, the material is waiting for construction and the working crew is already there ready to execute. That's when stuff gets really expensive. That's one thing. The other thing is rule checking, for example. There are so many clashes that are only found late in the construction cycle.
[00:05:41] And detecting those clashes early is something that we've been doing at Nimitz Check with a tool called Solibri for a very long time. And now we can augment that with AI and make it even larger in terms of the capabilities that it can check for. And then, of course, you have the operation space when it comes to operating the buildings. Here is solutions like Spacewell and the Nimitz Check Group. We use AI to detect anomalies in the building operations, in the HVAC systems and so forth.
[00:06:07] And that can really reduce cost and operational overhead as the building sits through decades of operation. So a lot of things to address here. But on the other hand, of course, you have still some industry inertia to overcome. Right. So the industry is structured around the old way of working. And this structure will probably not change as quickly as we all would love it to change.
[00:06:29] And here you have contract structures, for example, and regulatory requirements that are built around paying, for example, for change orders or splitting liability across different firms. And those kind of things that sometimes lead to incentives that are not aligned with the ultimate interest of building the building in the best possible and fastest and cheapest possible way. And before you join me today, I was doing a little research.
[00:06:55] And the big stat, 90% of projects are described as late or over budget. I mean, I've got to ask, where does information break down between design, engineering, construction and operations? What's going wrong here? That's actually a really interesting one, right? So it does not really break down inside the phases. Every phase itself is pretty self-contained and it's usually owned by one stakeholder, right? So the architect does the design, the engineer does the engineering and so forth.
[00:07:24] Where it breaks is actually at the handovers. And it never really flows back, right? So you have two things, the information as it travels forward in the value chain is breaking, but it also doesn't have a loop back into the previous part of the value chain. So every handover is a translation loss. So the model, just for context, how the industry operates, the designer, the architect does create something that's called a BIM file, right? A building information model.
[00:07:51] That's a PD representation, an abstract representation of the building. But then that gets translated into a PDF later for the construction site. And a PDF is a lossy conversion compared to that model, right? But then the PDF becomes an email. The email becomes a decision that nobody really records against the object that it changed. So you have information loss at every single boundary.
[00:08:14] The second thing you get is there's also, due to the accountabilities of how the industry is set up, a lot of, let's say, not really distrust, but basically people tend to recreate a lot of data sets. They don't just take the data set as coming from the previous phase, right? The engineer would redesign a lot of things that is actually coming from the architect.
[00:08:37] And that leads also to a lot of additional work that goes into that value chain, because ultimately the engineer is responsible to sign off for whatever comes out of his or her phase. And to really own that, you want to have trusted data in and you better verify everything. And those kind of steps really lead to a lot of duplicate work on the one hand, but also a lot of clashes and inconsistencies that then get resolved, unfortunately, in the construction site versus earlier.
[00:09:07] There's no loopback, no real circular, iterative, the building design process in place. It's very waterfall-like. And just to bring to life some of what we're talking about here, are you able to share an example of AI building information modeling or digital twin that can identify a problem early enough to change the outcome of a project? Do you have any stories that you'd have to name any names? Yeah, actually, we have a really good example right in our group.
[00:09:35] So we just recently acquired a company called Firmus AI, and we brought that into our Bluebeam software. Bluebeam is a software that is used by pre-construction and construction personnel that run Bluebeam to prepare and execute on construction sites. And what this solution does, now that we have it integrated in Bluebeam, it reads the 2D PDFs that construction really runs on and then does that drawing review, right?
[00:10:02] Just before the construction even starts, it flags missing information, scope gaps, and inconsistencies. And that saves those people on the construction site and the planners for the construction many, many hours, not only on the one hand, but also really on the other hand, detects issues that might have gone undetected otherwise, right? And that allows us to trigger that correction loop much earlier in the construction process than it was previously the case.
[00:10:28] Now, obviously, that's just the first step, right? And it's still too far down the value chain. What we want to do is actually now with EI, go through our portfolio that covers the entire value chain and really work right in Archicad, for example, that is the tool used by architects or in Alplan for engineering to surface issues to the architect, to the engineer while they're still designing the building. And that's the kind of advantage of an image group. We see the entire value chain.
[00:10:58] We see it on the Bluebeam side in construction, but also in the design phase in Archicad. And we can bring that together and connect the dots between what kind of decision on the design stage leads to what kind of problems in the construction site later, right? So my ultimate vision is I want to tell the architect that if you draw the wall this way, you're likely to see a five-week or five million construction issue down the road. Please fix it right now, right?
[00:11:24] If we get that kind of iteration going at that point, then the industry will really change. And I think the 90-40-20 problem will be massively different going forward, right? So I think in summary, I think we can take with AI a lot of the ignorance out of every single individual decision that is made due to lack of awareness or lack of data, lack of insight across the whole value chain.
[00:11:50] We can also solve a lot of the translation problems that today make it impossible to really even trace a single wall through the entire value chain. A wall element in a design phase has a different basically identity than a wall in a construction phase. And connecting the two semantically, that will also make a lot of those insights possible.
[00:12:15] So, Nemechek, you also develop AI across a wide portfolio of software brands. So I'm curious, how do you share capabilities and learning without almost forcing every product and customer into the same model? Tell me more about that and what you're doing here. That's a great question as well. Historically, we've been operating a lot as a collection of very independent brands. So more like a financial building company almost.
[00:12:42] And in that case, we did really well over our history. But now with AI, we've realized one thing. And the one thing is that AI does benefit from shared capabilities much more so than any other previous technology generation.
[00:12:59] When it comes to connecting insights from all the phases of the AC value chain and making those critical high value budget insight decisions that we want to make with AI and create high value outcomes for our customers, we need to connect the dots. And that means for us.
[00:13:49] And that means especially that we want to make AI and multiply across our group and help not just accelerate one product, but accelerate AI across all of our products by any investment that we make. And that is not just for efficiency on our end. I think that's frankly the least of the considerations.
[00:14:08] It's more for the capabilities that we can actually develop for our customers and the experience we can create and the speed with which we can launch AI capabilities into the market that is driving that kind of approach. One example of that is our metric AI assistant that we've been rolling out since 2025 across our different brands. So, for example, in Archicad, we launched the earliest version as the first beta.
[00:14:34] But now this is increasingly sitting across all of our products and it's becoming a little bit of a connecting fabric across all of our products. And this also helping standardize some of the workflows that we offer to our customers when it comes to AI. And of course, there is a lot of excitement around all things AI and agentic AI right now.
[00:14:56] But there is also a little, let's say, nervousness about AI usage costs that can rise rapidly when products operate at scale. And I would imagine this would be even more of a concern at construction scale, too. So how are you deciding what should use a large language model? What should use a smaller model and conventional automation or, in fact, no AI at all on occasion? Yeah, that's a very clear kind of approach here.
[00:15:23] Not everything needs to go to the latest and greatest frontier models. I think we have a four-tier kind of approach. The first tier is no AI, right? So anything that must be deterministic and needs auditable results, that needs to be deterministic. And that means no large frontier model AI. So, for example, code compliance checks, quantities, calculations, structural math.
[00:15:47] That is basically a rule that is right every time instead of 99% of the times, right? You don't want your structural math to be off every 100th decision. That doesn't work. So no AI definitely plays a huge role going forward. Still, then you have in the second tier regular automation and rules, right? So high volume, stable patterns, low ambiguity. This is where rule checking, for example, is a classic case.
[00:16:14] You don't need AI for this and you don't want AI for this if it is a stable pattern, right? And low ambiguity. This can be automated with rules and should be automated with rules going forward as well. The third one is then small or domain-tuned models, right? So here it is all about classifying objects, extracting data, retrieving data, parsing document information, understanding geometry and so forth.
[00:16:40] Those small models, often machine learning-based models, they are cheap, fast, and they can really run close to the data. This is a lot of the volume, actually. And then last but not least, you have the frontier models, the Gen AI, the large language models, and what everybody is talking about now.
[00:16:58] These are really amazing for reasoning, open-ended reasoning, especially if you have a more ambiguous situation and you cannot really find stable patterns and high-volume use cases. This frontier model approach scales really well for complex situations overall.
[00:17:17] You get multi-step agents, you can synthesize across many different documents, and you can basically process a very complex set of data into a meaningful outcome. And this is, I think, the new capability that is now changing a lot of the things in construction as well. And if I may draw an analogy here to some of the other industries I mentioned earlier, I was working on manufacturing before.
[00:17:42] And manufacturing has a very different kind of approach versus the AEC industry because the parameters of the industry are very different, right? So manufacturing produces hundreds of thousands of widgets or more in one factory that is pretty standardized. That is an environment that really works well for classic machine learning.
[00:18:05] You can train an algorithm for exactly that kind of scenario for that single construction line in that single factory, and it will do defect detection, for example, really well. AEC, we have a different problem. In AEC, every single building is a prototype of one. There is no construction site that is like the other.
[00:18:26] Even if you have standardized housing, the different context of the construction site is so different from any other construction site that classic machine learning was really hard to adopt in the past for the AEC industry. What is now very different with frontier models is that, let's say, general reasoning and general intelligence capabilities.
[00:18:48] This really helps us to scale AI at that level much more between different contexts, different building projects, different construction sites. And that is now the game changer for our industry because we work on prototypes of one with generalized reasoning on top of that, that can actually work. So, in summary, the question for me is never really which model is the smartest.
[00:19:12] It is what is the cheapest thing that is reliably right and that can answer it before the person on the site gives up waiting, right? So, that's the direction of following. When a CFO that might be listening is not worried about the rising token costs around AI, they're also famously concerned about that big ROI question that now surrounds every tech and, in particular, AI project.
[00:19:39] So, what kind of evidence convince an architecture or construction firm that an AI feature is actually delivering a return rather than simply making the software appear more modern? I suspect this is a question of getting more and more now. Yeah, absolutely. And I think there's different kind of use cases that have different, let's say, different kind of validation level needed, right?
[00:20:04] So, some use cases, and I'll give you some examples in a moment, some use cases are straight on obvious, right? So, if you have AI use cases that save your team hours per week, hours to produce a tender, reduce the RFI cycle time, reduce time to do a drawing review by five hours, that's one value dimension that I think is very easy to quantify because you get to reuse that capacity for higher value tasks.
[00:20:33] And that is a massive benefit. So, this is pretty straightforward when we come in with solutions such as Bluebeam Max that introduces those Roplane comparison features that are AI-based. That really is pretty straightforward to argue. Then you have a different category, which is maybe something where you reduce reward percentages or increase the number of issues you caught before something is executed, right?
[00:21:02] These are, over time, better to detect, right? But you probably need a little bit of a longer data series to prove that this is actually happening and there is no confounding factors that drive this KPI up or down. So, that's something that customers will probably take a little bit of time depending on the situation and the use cases to really get their heads around and evaluate the impact.
[00:21:29] And most importantly, also, there is sometimes you don't really have a baseline and a good data set to compare against. And that's another challenge when you actually quantify the RFI. If you don't have a baseline, it's really hard to make a claim on the improvement, right? So, data quality is important for ROI calculation. And then, of course, we look at sustained usage and we even ask customers and internally our customers are using surveys to check with their users.
[00:21:56] And there's a lot of studies by now, independent of the AEC industry, that ask, for example, in R&D what software engineers say about potentially losing access to agentic coding tools. And I think the survey results are very unambiguous there in terms of my productivity would plummet compared to where I am today. And that kind of survey result also gives you at least a directional input to your ROI calculation.
[00:22:24] You mentioned data a few moments ago and construction data can contain commercially sensitive designs, intellectual property and indeed information that is tied to national infrastructure. So, how are sovereignty and model training concerns affecting adoption now? It's something that I hear more and more about, especially across Europe. Yeah, that's an important question. So, sovereignty, especially in Europe, is a big topic.
[00:22:54] And it's a very complex topic that has a lot of, let's say, confusion attached to it, right? Let me try to break it apart a little bit. When it comes to sovereignty, it has components around where data is going, where it is stored, who and where is it, who is processing it and where is it processed? And what is it used for besides that actual use case?
[00:23:20] If I'm designing a nuclear power reactor in Europe, that would be maybe one example where sovereignty does play a big role. The companies and the governments involved with that would most likely have some requirements along those categories. Maybe not all of those categories, but many of those categories.
[00:23:42] So, the data set on the design of the nuclear power plant might be classified as sensitive, right? So, the data needs to either stay on premise or be shared with a system that is a, first of all, trusted partner that has security certifications, but also hosts the data within the jurisdiction that is under control of that government.
[00:24:08] So, their deployment options of our backend infrastructure are very important to meet those requirements. The other thing is when it comes to then processing that data, having it processed in Europe, for example, is one thing, but also is the design itself used to train models that make it easier to produce another nuclear power plant by a competitor going forward, right?
[00:24:36] These are the questions that come up in that space. And there, we are pretty clear and we're working on making our position even more clear going forward on how we deal with customers' IP. And I want to be very clear on this. Not everything, not every kind of data set is IP, right? There's a lot of, probably 95% of data is really not really IP.
[00:25:00] It's usage patterns and other topics that really help improve the software, but don't really reveal anything about the project that you're working on or your individual customer situation or what your interests are. I think there's an important need to clearly separate IP from non-IP data. And then for the IP data set, really handle this very carefully and in line with what customers really need.
[00:25:27] And that need depends on a lot of the, let's say, industry that you're working on, the nuclear power plant versus, let's say, residential buildings. And one other area we must check on here is the fact that the industry is also facing a severe skill shortage right now. So where can AI support experienced professionals and newer workers that are entering the scene without removing the human knowledge needed to catch some of those dangerous mistakes?
[00:25:54] Because we also hear about entry levels being, roles being removed as well. It's a real almost paradox, isn't it, that we find ourselves at the moment? Yeah, there's one general issue across all industries, right? So if you're not hiring junior people, you end up not having senior people down the line. And that is a general issue that all industries need to contend with. From our perspective, AI really compresses the apprenticeship period for newer workers, right?
[00:26:23] So it buys back judgment time for the experienced people, but also makes newer people or university graduates come to terms with the industry standards and really understand how to operate productively in the industry much more quickly than ever before, right? So it speeds up that onboarding time dramatically, but also makes senior people much more productive.
[00:26:45] The approach that we're taking with AI and that I see as very valuable, especially in our industry is AI proposes and a qualified human decides. So that's a clear kind of accountability split, especially in a regulation heavy industry like ours, the human sign off will be very important. And that is therefore instrumental that our AI that we're developing can make great proposals and make those proposals also explainable so humans can verify.
[00:27:15] So the system really has to show the sources and a human can actually chat. And that change will help quite a bit with the skills shortage because that increases the leverage that every single human in that value chain has, right?
[00:27:28] So an experienced architect, an experienced engineer can spend way less time on documentation, grunt work and many other topics and really focus on deploying that high quality judgment that was honed over years and decades of work into the most high leverage kind of decisions that human can make.
[00:27:51] And whenever I talk to architects, I always ask who actually went to university to spend hours and the majority of the week to do documentation, code lookups and all those things, right? And nobody raises their hands, right? Everybody went to university with a vision in their mind on what it means to be an architect.
[00:28:11] And that vision usually was I'd be in the sketchboard translating the vision I have in my head about the building and the site it is on into something beautiful, functional, that delivers high quality of life for the people occupying it, right? And that is unfortunately a small part of the work of an architect. The larger part of the architect work today, at least still, is a lot of the tedious stuff surrounding it, right?
[00:28:37] And I think this is the first thing we really want to address comprehensively is to make that burden lighter and help architects, engineers and others to focus on the best value they can bring to the job. And also zooming out as a society, if you look into our cities today, we have a lot of cookie cutter buildings in many ways, right? So a lot of standardized buildings and the true kind of icons of architecture, they are not everywhere, right?
[00:29:05] Not every single building you see in every city is an icon of architecture. And that is not because the architect that worked on it was not talented, not at all, right? I think it's just a matter of is there enough time and resources to focus that precious resource on designing a unique building? And in nine times or 99 out of 100 times, probably not.
[00:29:28] And that is, I think, where AI can also have a massive impact on our cities if we get architects and engineers to focus on unique buildings and make unique buildings constructible at a very reasonable cost. I think this is sort of the society impact of AI and AEC. Well, I absolutely love chatting with you today. We've covered so much in a short amount of time.
[00:29:52] And I think for a lot of people listening, when they're doom scrolling down their news feed, they hear a lot around how the sector is facing a wide range of challenges from pollution, material waste, project delays, budget overruns, rising costs and growing skill shortages.
[00:30:08] And just to hear you talking today about your AI strategy that is being developed across Nemechek's portfolio and AI costs, return on investment, data sensitivity and intellectual property, sovereignty, all massive talking points at the moment. So it feels like a real exciting space. I'm curious, what makes you want to jump out of bed in the morning?
[00:30:28] What excites you about the work you're doing at the moment that you're going to be doing for the rest of the year and indeed into 2027, which is scarily just a few months away now? But what excites you? Look, I think AI is, from my perspective at least, the biggest change in all of human history. It's the technology that allows us to develop technology faster than ever before.
[00:30:54] It is allowing us to scale reasoning and intelligence across any industry. So for me, being alive right now, why this is happening and seeing AI as sort of a new intelligent species starting to coexist with humanity and making sure that AI is helping us and figuring out how to integrate AI with humanity.
[00:31:20] And along the way, hopefully also figuring out what it truly means to be a human in the age of AI. I think that is something that gets me up every morning and applying that to the AEC sector that really addresses all of us today. Every one of us lives in a building. Everyone goes to cities, goes to work, drives on streets.
[00:31:42] It's the impact that AI and AEC has on our lives, on our day-to-day experience of life is, I think, one of the biggest across all of the industries. And just making it 5% or 1% even better along those KPIs that we discussed earlier will have an outsized impact on humanity and the experience we all have. So that, for me, is a mission worth getting up for in the morning.
[00:32:07] And at Nimicek, we're uniquely positioned as a leader in the AEC software industry to do exactly that. So I'm excited for this year and I'm even more excited for next year as AI develops further and also as we integrate AI even deeper into our portfolio.
[00:32:23] Well, one of the things that stands out in our conversation today is your passion in working to modernize architecture, engineering, construction, operations, all through AI, digital twins, building information, modeling, all incredibly cool. I love that line you said a moment ago. Are you just so excited about being a human in the age of AI? And I think that's something that so many of us feel right now. It's almost being like a kid in a candy store, so to speak.
[00:32:49] But for anyone listening that would just like to carry on this conversation, find out more information about Nimicek Group and everything we talked about, where should they go? Visit Nimicek.com, of course. That's our website. There you can find all of our brands. I know anyone in the industry that is or close to the industry will recognize many of our brands. Nimicek itself is not very much in the foreground as a brand yet.
[00:33:16] We've been advertising our brands very heavily over the last decades, and we led with the brands. Now you'll hear increasingly more about Nimicek itself over the next several years. And we also have an Instagram channel for Nimicek AI specifically. If you want to look that up and follow what we're doing on AI in AEC, please do so. I'd be excited to have you as a follower there. Awesome. Well, I will add links to everything you mentioned, including the LinkedIn pages for both you and Nimicek as well.
[00:33:45] So I encourage people listening to go check that out. As for yourself, I'd love to stay in touch with you and get you back on early next year and see how things have continued to evolve. But more than anything, thank you for sitting down with me today and sharing your story and bringing all of this to life. Really appreciate you, Tom. Very nice. Thank you so much, Neil. Thanks for having me today. I think my biggest takeaway from Julian is that choosing the smartest model is often the wrong starting point.
[00:34:12] The better question is which method can produce a reliable answer at the lowest cost and in time for someone to act. Sometimes that means a frontier model. Others, it means a smaller domain model, conventional automation, or dare I say it, no AI at all. And it's that useful test that is incredibly useful for any business, not only construction.
[00:34:37] So big thank you to Julian for showing how Nimicek Group is applying that thinking across the built world while keeping qualified professionals responsible for the final decision. Where in your own organization would a simpler rule outperform an impressive model? And how would you prove it? Well, you can find everything we've talked about today over on the blog post associated with this episode at techtalksnetwork.com.
[00:35:06] You can even leave me an audio message while you're there. But I have eaten into far too much of your time already, so I'm going to go now, but I'll be back again tomorrow with another guest. Bye for now. Bye for now. Bye for now. Bye for now.