192: ‘Building an AI-First Practice’, with Patrick Chopson and Eric Cesal

192: ‘Building an AI-First Practice’, with Patrick Chopson and Eric Cesal

Evan Troxel: Welcome to the TRXL Podcast. I'm Evan Troxel, and in this episode I welcome Patrick Chopson and Eric Cesal. Patrick is the co-founder of Cove, AKA Cove Tool, and one of the drivers behind their newly formed AI powered architecture practice.

Eric is an architect, post-disaster expert, former executive director at Architecture for Humanity and special program instructor at Harvard. He's also a sharp observer of the evolving relationship between technology and architectural practice. His Substack called "Life as a Disaster" is worth subscribing to, and I've included a link to it in the show notes for this episode.

This episode explores what it means to build an AI first architecture firm, not by layering AI on top of outdated processes, but by starting from scratch.

Eric introduces the AI first mindset, borrowing a page from the journalism industry's shift to digital first thinking and challenges us to stop asking how AI can fit into traditional workflows. Instead, what happens when we reimagine the entire system from first principles? Patrick meanwhile has already made the leap. He shares how Cove is now delivering design services with a hybrid team of architects and software engineers building bespoke AI tools, sometimes just for a single project. What they've created goes beyond their sustainability consulting service.

It's a new kind of architecture firm, one that is operationally different, legally responsible, of course, And focused on giving experienced architects more time to design while maintaining higher coordination and quality standards. We also dig into the critical mismatch between architectural education and real world demands in an AI driven practice why large multi-office firms may be structurally at a disadvantage during this technological shift. How AI might elevate senior architects while challenging the traditional learning curve of junior staff and why building trust in the output, not just the speed of these tools is essential to firm adoption.

A key theme from this conversation, which connects to many of the other episodes, is the power of rethinking practice models from the ground up. An AI first approach forces us to confront everything we've inherited from phase structures to compensation models, and ask if we started today with today's tools and knowledge, how would we design a practice that serves both clients and architects better?

There's an extensive amount of additional information in the show notes as usual, so be sure to check those out. You can find them directly in your podcast app if you're a paid supporter of TRXL+. And if you're a free member, you can find them at the website, which is TRXL.co. Please consider becoming a paid supporter if you aren't one already. If you are, thank you so much.

It makes a tangible difference for the sustainability of this podcast. So now without further ado, I bring you my conversation with Patrick Chopson and Eric Cesal.

Welcome back to the podcast. This episode is gonna feature two brand new guests. And Eric, we just actually got to meet in person at a, at a recent event in San Francisco, which was fantastic. I. And I think I told you when I met you originally just on the phone, that I had been gifted one of your books in the past, which was Down Detour Road, which is a, a, I mean, searching for meaning I think is kind of one of the things that I, I talked about with you, uh, regarding, I like this.

Maybe, maybe architects are always kind of dealing with an existential crisis on some level, right? I mean, and we will be talking about that today actually. But, that was kind of my intro to you a few years ago that a friend gifted me that book. So, Eric, welcome to the show, and Patrick, welcome to the show.

Eric, if you want to just kind of kick us off and kind of give us an idea of, of where you've been before we get into the actual topic today.

Eric J Cesal: first of all, my, my main focus is not, and has never been on technology per se. Um, my background is in the humanitarian sector. I train as an architect, and as a construction manager, uh, I have degrees in both as well as an MBA. Um, so after five years of commercial practice, I was back in school and started volunteering on the Katrina reconstruction. Um, and that kicked off a career in post-disaster recovery that. went global, um, leading the Global Disaster Studio for Architecture of Humanity. And in 2015, I pivoted to more of a teaching and advocacy role, which included having my own podcast for a while, social Design Insights and, um, teaching at first at WashU, then at Berkeley, and now currently at Harvard.

And I think the roots of this discussion started a couple of years ago when I started writing and lecturing on AI through the lens of You know, not AI doism, but you know, this is a large impact events and you know, is architecture as a profession going to be resilient towards it? Um, and are we having the right preparations and reactions for what's coming?

Evan Troxel: I want to talk about a, a recent, fairly recent article that you wrote before we jump into that. Patrick, welcome to the podcast. Great to have you. I know we've talked before just on a, on a regular call and, and talked about kind of what you guys were up to with Cove, and since then made another a pivot speaking of Pivots that Eric just mentioned, which we'll talk about today.

But Patrick, welcome to the podcast and maybe you can kind of give us a little bit of background on your circuitous route through this industry.

Patrick Chopson: super happy to be here. Um, one of the things that is interesting is we've raised three, six and a half million dollars to date in venture funding. Tr against trying to find ways to bring carbon lower.

That's kind of like the

Evan Troxel: Mm-hmm. Mm-hmm.

Patrick Chopson: Um, how we do that. We've tried many different things.

We've tried, uh, working with manufacturers, we've tried working with architects to varying degrees of success with our analysis tool that we've had for many years. Um, but then over time we found out that most architects want to answer. Some of them will learn, uh, but many want to just have someone tell them what their carbon is in their billion or the energy use or daylight or

so on and so forth.

Evan Troxel: Hmm.

Patrick Chopson: started a consulting practice, which then, recently, uh, we've actually also expanded that to include architectural services as well. So we're one of the first ai, uh, powered architecture firms, and we've had a lot of success thus far with a lot of developers. So excited to keep that rolling.

Evan Troxel: Nice. All right. Well, Eric, let's, let's jump into kind of where the idea for this conversation started, which was with an article that you wrote and you said you don't, you didn't, don't really focus on the AI kind of doism, but you do include images of. The Terminator in your graphics.

Eric J Cesal: Yeah, do.

Evan Troxel: So maybe you can kind of kick us off about being what it means to kind of rethink practice when we talk about new tools, kind of, I don't wanna say taking over, but infiltrating, but also kind of change.

There's a new mindset that is potentially required here. And I'm, I'm curious kind of your thoughts when you're about the, the article that you wrote about being an AI first practice.

Eric J Cesal: Yeah. Uh, well, first of all, the Terminator thing is meant to be diffusing humor. Um, so I,

Evan Troxel: yeah. It's not a, it's not a menacing graphic. I, I didn't mean to paint that picture. It's like, it's like it's very cartoonish. Here's the Terminator. The Terminator is, has a a, a speech bubble and it, yeah. So anyway, we'll, we'll put a, we'll put a link to it in the show notes so people can, can read it.

Eric J Cesal: no, the, the AI first article was, was the culmination of several years of thinking and, and arose out of a metaphor. Um, the metaphor of, um, digital first, which journalism went through in the nineties. And, you know, with the emergence of the internet, there was this tension, you know, are we going to be print first or digital

Evan Troxel: Yeah.

Eric J Cesal: So if you're print first, you already know what you're doing, and then you just have to stand up a website and put the articles on there. But being digital first offers some advantages, and if we're gonna be digital first, then a lot of those that had been in place that had been sacred for the last 100 years probably didn't make a lot of sense, as well as a lot of the physical infrastructure.

So, you know, I started thinking through that a couple of years ago, you know, an AI first architecture firm. If we started off with a blank slate and said, you know, the one objective is to deliver. Designs for buildings that, that want to get made. Um, how would you do it? You know, not how would you take AI and apply it to the processes that we know, the phases that we know, the, you know, customs and habits.

Nobody knows like, why the hell we do this the way that we do. But that's the way I was taught and that's the way my mentor was taught by his mentor and, and this sort of thing. Um, knowing all the technology that we have today, um, and the technology that we could be fairly confident is coming in the next few years, how would you organize a practice?

And, you know, it's, some of it probably seems a little farfetched, but that's the point, um, is that, you know, we're, we're buried in, in a lot of practices that, you know, got their start in an era of paper. Um, and then we kind of layered digital technology and computers onto those practices and now we're gonna layer ai. top of that. And I think things emerge when you really ask that question. Like, okay, scrap everything except the starting point, end point. And let's see how we get there. And I'll offer one example and then I'll give up a mic. I mean, one of it is, is phasing, right? Um, you know, we have these very strict phases of schematic design, design development, construction documents, et cetera. and my hypothesis was that, you know, you could easily embody the interests, um, and maybe even the personality of a client in some sort of language models. There's researchers doing that at Stanford now, and I'm sure the C-E-A-C-I-A did it, you know, five years ago. Um, and then, you know, have the client as, as a copilot on the design process where you could access. That person for questions when they were actually necessary as opposed to, you know, at some pre-scheduled interval or something like that. And you could also block them out, right when you just need to work. You can say, all right, you know, you go off to the side, I'm pressing mute. Um, this is my design time.

I'm going to figure it out. So, I mean, I think we all should be engaged in those sort of speculations at the moment 'cause they're gonna be here before we know what I think.

Evan Troxel: I think there's a lot of speculating and a lot of pontificating. I mean, you're talking about kind of a thought experiment, right? And, and there's value in that, but I think a lot of people get stuck at that part too. And so I think this is where it, it gets interesting because Patrick's here and talking about what you're up to with Cove.

It's like you went beyond the pontification and you're, you're out there and you're talking about what you're doing. You're not doing it even in secret. You're kind of. Semi, I assume kind of doing it in public, not a hundred percent, but maybe this podcast gets this, this more behind the scenes about what you're actually up to, but to actually like pull the trigger and say, we are going to move forward with a different model.

And I don't know what that means. I, I, I just assume it's different. Right. So I mean, Patrick, give us an idea of, I mean, because Eric, you talked about like, we can see we're fairly certain about what's coming. I don't know if that's actually the case. I think a lot of times this is also what's holding architects back, which is like, what do you mean it's fairly certain?

Like this stuff is still changing so rapidly. What if we tie ourselves to this thing and then it, the, the rug gets pulled out from underneath us, right? And, and, and maybe the nimbleness of a startup has a little bit more ability to kind of navigate that. Anyway, I want to throw this to you, Patrick, and, and kind of get your thoughts on, based on what Eric is talking about, how you decided to actually move forward in this space.

Patrick Chopson: Yeah. Yeah. So I think like a lot of, as you mentioned, a lot of folks were saying over the last couple years, oh, this is coming someday it'll be here.

Many firms

Evan Troxel: It's here.

Patrick Chopson: But you have to think about why did large firms and many other firms banned the use of ai? It's because they couldn't validate the, um, the chain of command of thought that runs through

an idea.

So like, you couldn't actually maintain what they call responsible control. So without responsible control, you can't use AI because you can't, you're legally have to understand as an architect, like, where'd you, where did I get the answer

from?

Evan Troxel: Mm-hmm.

Patrick Chopson: like made sure that I've exercised the standard of care. So for the most part, up until maybe about six or seven months ago, it was very hard to kind of like figure that out. And so once we had solved that internally here at Cove with some of the new tech that had come out with from different people like Claude, you know, open the eye, different ones, uh, then we were able to actually start building tools around that where you can maintain and understand how you get from one answer to like another answer and map that out.

And so that's like really important. So that was the first thing that was enabling that allowed us to move forward. And the second thing we saw that really was a big game changer was understanding, uh, inter interpretation of intent by AI systems has dramatically improved over the last like six to eight months. And so once that also, those two things, interpretation of intent. Is being able to re maintain responsible control. Now you can actually build an AI powered architecture firm. And then of course we have years and years of building for the a a C industry at our back. So we actually understand how to build applications and we have a full software team of 10 people. But now those 10 software engineers collaborate with the architects. And as we're doing specific projects, we not just build an application that is general purpose. We can build applications that are for single use, like for a specific project, we can actually build an AI application that solves a very specific problem, piece of software out maybe never used again.

'cause there's not, um, as much of a, a problem with that. And we're also building like some really cool internal tools that allow us to produce drawings and things like that. And those ones are how we kind of gather a lot of information out front and be able to actually. Do the, the work of architecture a lot more smoothly with less revisions, um, and get things done on time, which turns out to be the thing that most developers right now are really for is a on-time delivery.

Evan Troxel: Does that and is that really like, because there's just less and less time available to do these kinds of things, is because it's like the, the, the tool stack that people have and their understanding of their current capabilities. Like we know pretty much how long it takes to do a thing, but that that timeframe is getting squeezed and like new tools do allow.

Efficiencies to happen, maybe less steps. Um, it's, it's not all about efficiency. I, I would assume like there's gotta be design differentiation to make projects also successful. And so I'm just curious from your point of view, like, are, have you just just accepted that as like, yeah, we're, we're gonna have less time to do it.

The quality still needs to, the bar needs to be way up here. How are we going to get from here to there? And it's by employing thinking and tools like this.

Patrick Chopson: the most important thing though is when a developer is hiring an architect, if you really break down the business problem that they're trying to solve is they need someone to assume

risk. And so

Evan Troxel: Mm-hmm.

Patrick Chopson: doing it faster, which every architect architect's afraid of when it comes to AI, is not really what they want. wanna pay for someone to review project, to come up with the best and highest use of their design. And you know, as architects, we sometimes, because of all the things we have to do, we don't have time to make sure that we explore multiple alternatives. We want the design cycle

time is just

Evan Troxel: Mm-hmm.

Patrick Chopson: You know, you saw people like big architects, they use grasshopper and things like that to improve their cycle times. And that's kind of like how they achieve some success with more resolved. know, looking designs, you know, back over the last few years, but that's like still only a two x or three x improvement and speed of iteration.

So I think like

Evan Troxel: Mm-hmm.

Patrick Chopson: as architects we need to focus on. The developer does not want us to do it too fast. What they want is they just want to get it done and they want to have it be coordinated because we're seeing many principals are retiring now, the knowledge they have around coordination not necessarily getting passed on quite as well.

And so we see more and more uncoordinated drawings delivered late where, that's the big pain point for most developers. So that's why

Evan Troxel: There's more and more drawings too. Like there,

there's a lot more drawings. I, I, I'm curious, just if you, if you guys have seen this newest graph, I've seen it circulating around a little bit, which is, you know, like the old version was, you know, here, here's design is kind of like this slow. Curve and the gently curving upward of it is like what, what you're doing over time.

And AI kind of jumps up real quick in the beginning, but then there's a bunch of iteration. 'cause it gives you something pretty close. And then it's like, okay, now tweak, tweak, tweak, tweak, tweak, tweak, tweak, tweak, tweak, tweak. And it's like, okay, we end up at the same place at the same time as we used to.

It's a different process to get there. But it sounds, Patrick, like what you're talking about is you've actually been able to kind of make the tools do what you want faster. Like you trust them sooner because of the kind of the work that you're doing Is, does that sound about right or I, you tell me.

Patrick Chopson: I mean if you build it yourself, you can definitely how you got the answer.

Evan Troxel: Interesting.

Patrick Chopson: that's the most important thing. But giving us more time for design excellence, that's

like what we're

Evan Troxel: Mm-hmm.

Patrick Chopson: for in the processes that we have in house is like, how do we get experienced really good architects with really good design sense? How do we give them more time to do the thing that.

They do best, which

Evan Troxel: Where the value is.

Patrick Chopson: the

Evan Troxel: Yeah.

Right. Eric, what do you think about what Patrick's kind of describing as as a place to kind of kick this conversation off with as far as like implementing into practice?

Eric J Cesal: I mean, I think the, the implementation is great and it's a terribly important discussion, right? You know, where is the and what are people actually, I. Seeking when they hire the services of an architect. I think Patrick is spot on. Um, you know, the, the, the highest motivation of most of the developers that I know is, is risk.

You know, they want somebody to assume the risk of errors because those errors stack up. Um, you know, I would challenge a little bit the idea that, you know, if, if, if time is second, it's a very, very close second. You know, when you, you take out a $50 million commercial loan, um, you know, you're, you're paying a of a million dollars a month in interest, um, you know, on this project.

So, um, if we have the ability to collapse the design process from 14 months to 14 weeks, I assume every developer would jump at that because it means the building gets done sooner and you know, the returns start coming in sooner, as long as it's at. The specific level of quality. Um, I think that brings with it problems in an industry that charges by the hour.

Um, and I think that's the fundamental problem that I've always been reacting to, is that architects have an idea about where their own value comes from. And I'm not convinced that that's synchronous with other people's ideas of where our value actually lies. Um, and I think it's, uh, also not synchronous with the, the mechanisms that we have.

You know, we, we claim that design is where our value lies, but the methods that we use to charge for our services are principally mighty hour and as a percentage of construction costs, both of which are designed, So, you know, I think AI and the sort of advances that, that Patrick and his team are achieving really brings into relief the, the historical problems that we all kind of knew about with architectural practice.

And I. I hope that our methods of practice can evolve as quickly as Patrick is evolving the technology crossing my fingers.

Evan Troxel: I, I have an like another wrinkle maybe to throw in here. And Patrick, I'd love to hear your thoughts on this, which is like, if it, okay, just using Eric's example of 14 months to 14 weeks. That to me sounds more valuable than, right. Like it sounds like that should actually cost more for an architect's fees.

And so I'm curious like, like how is this kind of, how are you seeing this work out in when, when it comes to practice, billing, all of those, you know, timelines, milestones, those kinds of things?

Patrick Chopson: be, I would've thought that too when we first started, so we even did this with architects when we did consulting before, where we would,

Evan Troxel: Mm-hmm.

Patrick Chopson: in a day and we thought that people would want to get that detailed analysis in a day, it turned out that it was too fast, and that if you did it so fast, people thought it wasn't valuable. So there is an element in the construction industry that people know how long things should

take or

Evan Troxel: Mm-hmm.

Patrick Chopson: to take. And then if you fall too far outside of that then it calls into question the validity of your process and your accuracy of what you're doing and how high the quality is. So there is some, like, you know, people just wanna make sure that proper time is being taken. So, yeah, I can see it like. For us, you know, we've been able to do things about 40% faster, 60% faster in terms of making drawings. But when I, if I go to sell a, an owner on that, they're like, Hmm, biggest risk is that my money that I've sunk into this project and to acquire the land or to like secure this loan. If I lose that, my architect's fees are nothing in comparison to that. So it's like sure that it's right, that it's been vetted, that's been figured out. But also like in the beginning when you're just doing side evaluations, um, if those things are incorrect and we can't actually build for the price that we think we are, and we're 20%, 40% off in the construction costs, then that's a bigger risk oftentimes to the developer.

So like, that they really wanna know is like, is this accurate? And so if we're like, yeah, it's more accurate what we're doing, or it's higher quality that I. probably what most people are paying for. Now. Obviously there'll always be someone who wants to do it quick, like a web services building 20,000, know, data centers or like some kind of, know, people building warehouses.

That could be a thing. But, you know, so speed is of the essence for some building types. But for many projects, like what we're finding is that it's the 50 to a hundred million dollar projects where our, our, uh, services are most receptive. Um, so that's gonna be things like multifamily garden style or like midrise or we had a couple highrise projects we're working on.

So it's like those, those types of projects we're coordination can make or break the project are like really, really ones that we seem, seem to be winning on a lot lately. Uh, which I think is not what I expected. I thought we'd do small projects first that had quick turnarounds. the bigger ones that have more coordination where we're really kind of like winning the trust of those developers.

Evan Troxel: Interesting. And, and if you can do drawings 40 to 60% faster, does that just mean then that you're spending more time on the quality level and the. Yeah, upfront. So, so you're, you're, yeah. So you're just like changing the, the, the way that the hours are kind of stacked, right? So, so you actually are getting to put more time into design, like you talked about, and ensuring that the quality's there, where a lot of people are just underwater trying to keep up and there's delivering an uncoordinated quote unquote set of drawings at the end.

Obviously architects, reputations, they're license, all kinds of things rely on this kind of thing. So I, I'm not just trying to say that architects are doing that, but I mean, it, things fall through the cracks, like they definitely slip through the cracks. You, you, a lot of times, you know, especially, you know, in the last, let's just say 10 years ago, right?

You would, in public work for schools, you could submit a set of drawings and then, because you knew it sat in the bin for six to nine months before it was gonna get reviewed. Knowing it's half baked and then go slip in sheets before you know that things have been catching up to that. Like that isn't really a thing anymore, I don't think.

But the, like these are examples of how this has worked in the past, and now you're saying like you're actually getting to spend more time on the stuff that matters in the process because of those efficiencies that you've figured out.

Patrick Chopson: know, think about too, like there's a lot of studies I'm sure that you guys have probably seen as well and other viewers, but like, I don't know if you guys realize this, but like they're saying like when study after study people with lots of experience can achieve up to 10 x improvement in using AI of some type that's well fitted to their industry, but someone with no experience achieves a negative increase in productivity. Uh, so I can see that on as I train people on our team that we have to invest a lot in training people who are junior

on how to

Evan Troxel: Hmm.

Patrick Chopson: I. Someone who's older, uh, and more experienced. Like for example, ed Aikens, a IAS on our team. He, um, you know, he picked up driven workflows in a few days as soon as we started using it, producing gorgeous drawings, beautiful images, but everything was like, looked a lot like what I would expect to see from him if he'd spent several weeks. But he was like doing it in just a couple days. And so like, then he spent his time designing he understands what he is looking for. And so like, that is like to get the beauty. And also the, the, the hard part is like you have to invest much more in people. If you want AI to be successful in your

business.

Evan Troxel: I, I knew there was gonna be a people aspect here because this is the, the thing, the, the, the fear. There's obviously a lot of fear in, in these tools, but also just in the unknown, right? So I'm curious, Eric, what you think about this people side of, of the conversation now.

Eric J Cesal: Um, uh, yeah. Where to start? Um, no, I mean, it's got me concerned. Um, you know, Patrick, I hear a lot of the same things and, you know, I experience them in my own work with ai. You know, when I ask. Clawed a question, and you know, it's a complicated question. The answer comes back in less than a second. I'm like, nah, that can't be right.

You know, like, do something else. Um, I

Evan Troxel: Fake it. Just fake it for me. That's what you just, just take.

Eric J Cesal: just,

Evan Troxel: If it did take a little longer, would I trust it more? That's a, that's an interesting aspect to this.

Eric J Cesal: Well, yeah, I mean, when they released the microwave, there was a lot of suspicion around it because, you

Evan Troxel: Hmm.

Eric J Cesal: people could not get their heads around the idea that food was being cooked in so short amount of time. And for all of human history, you know, eating uncooked food was, you know, occasionally fatal.

Um, so people have a lot of concerns about it. And, you know, gradually people acclimated to the idea that like, yeah, I can, I can cook my dinner in, in two minutes or three minutes in the microwave or, or something like that.

Evan Troxel: Now we rely on it. We,

Eric J Cesal: Yeah. Now it's like absolutely necessary. But that's a good point too, because, you know, the sort of hesitancy Patrick that you were describing, I wonder how long that sticks around and how long it takes for the entire a EC value chain to acclimate. Um, deeply respect the fact that, you know, you, you soak up that extra time with more design to, you know, bring the work to a higher level of quality. Um, I also know that their architects weren't gonna do that once they get, you know, hands on the tools and they're gonna,

Evan Troxel: hmm.

Eric J Cesal: outbid everybody in the room and say, look, I can get it done in 47 minutes, just like, give me the job and like,

Evan Troxel: Race to the bottom. Yeah.

Eric J Cesal: I mean, I think that's where, you know, I get concerned about the people aspect is that arctics have this notorious history of devaluing their own services and their own worth. And I think it's a profession wide project to figure out. How as these technological changes are taking place, we manage to preserve or enhance the value as as perceived on the client side.

Patrick Chopson: And that's, and that's where like in our opinion, you know, obviously 'cause this is a protected profession 'cause we've got the stamp, it really lends itself, unlike many other industries, the AI innovation will have to be done by architects themselves because of like the legal liability of the stamp itself.

So like for myself as the licensed that, uh, stamps a lot of the drawings here, I have to be, and also the person producing the software, like I've coded my own a DA bathroom generator. That's a part of the whole process, right? Like, I wanna know exactly what assumptions we made, how do we make them? Go through the, with the team that's making it, check it line by line, make sure that makes things that are plausible and that meet code. But like, you're right, like there'll be a time when that won't be the case, that there'll be less careful people building things. But I think it's really incumbent upon those who are building the tools like us and others to like how a broader conversation with the ac, especially the, a part of the community, start educating people, start collaborating, um, sharing things like what we're doing here, you know, the ideas back and forth such that we can make sure that the systems we are designed are biased towards Like, for example, we're analyzing everything from a carbon perspective in the backend, or not it tells that to the owner, or we actually highlight that. Maybe not a good idea these days, but, like we're actually checking it, you know, so that we can ensure that our billings are

by default the

Evan Troxel: Hmm.

Patrick Chopson: we can achieve. But that's like kind of my, my dream has always been how do I create that maximum impact. come to see that it's maybe affecting the architecture as maybe a faster way to do that. So I don't know. There's like, but there's a lot of like, we have to be, uh, engaging with the a i a, um, different folks. I know has kind of been on a rocky period recently in terms of, you know, how to get the profession to work better together. I think like if we do all work better together, we can come through this whole AI transition a philosophy and rules and things that make sense. I think I would advocate for an ai, an AI of conduct from the ai.

So everyone

Evan Troxel: Okay. Are.

Eric J Cesal: absolutely. Yeah.

Evan Troxel: I wanna, I wanna throw this out now because, and it hasn't come up on this show yet, but it's, did you guys see the survey that a i a published regarding, uh, adoption of AI tools just amongst existing firms? And it, and it was put out there as kind of like a report, but it was, it was just based on survey results.

And that's the survey, I mean. I just maybe, maybe Patrick, from, from your point of view, what did you think about that survey? Because, because it was really a mix, right? It was like, uh, okay, of course this is gonna be what an a i a survey results look like. It's like some architects are just saying, we're just watching from the sidelines, and other ones are like, yeah, we're, we're, we're going, okay, you just painted a picture of nothingness to me, I mean, it, it is just like that, that is such an obvious outcome of a quote unquote survey.

But the way that they presented it was like, here, here are the facts of the industry. And I felt like it was actually, like I asked, who's the audience of this survey? Is it architects? Because, I mean, it's a public facing. Article that they put out there. And I'm thinking, well, if, if future clients read this, it's kind of damning of the industry.

Like our architects aren't adopting the latest, greatest tools. Everybody's adopting these tools, why wouldn't architects? And, I mean, there's a lot of nuance in there, right? When it comes to risk and standard of care and control of the documents and all those things that a lot the public doesn't know about.

But I, I kind of, they don't, they don't care about that stuff either. Right? So I'm just curious from, from your point of view, what, what you thought of that survey and, because I think it is kind of a state of the industry, but at the same time there's gonna be a lot of AI first firms like you who would never, not, not necessarily you, but they wouldn't even talk about their use of AI because it's, they, they want to keep it a secret.

Like a lot of firms have always wanted to keep their technologies and their special sauce and, and special podcast air quotes a secret.

Patrick Chopson: Yeah, I don't know. I, I don't, I, I just have seen have, having tried to sell technology to architects for years and years that architects don't adopt new technology easily.

Evan Troxel: Mm-hmm.

Patrick Chopson: know this. Uh,

Eric J Cesal: Yeah.

Evan Troxel: Mm-hmm.

Patrick Chopson: so,

I

Evan Troxel: Live, lived it.

Patrick Chopson: I don't think that it'll be very quick that most firms adopt, uh, ai, to be honest. Like it'll, they'll need to see some other people do it.

It's kinda like

the whole

Evan Troxel: Yeah.

Patrick Chopson: When you see the penguins, they get close to the edge and the south pole, and then there's like the sea lions and the orcas down there swimming around. They, they get closer and closer until one penguin falls off and then they all look down. They say like, well, did he die?

And

if he doesn't, then, then they

Evan Troxel: I think if you look closely, there was, there was one right behind it that just kind of gave a quick push.

Patrick Chopson: So that's, that's I think how we'll see it. We see, we've seen that like from cad, uh, to bim, there was like a 20 year adoption cycle and I think like Bim, if I'm not mistaken, it was 10 or 15 years and then like probably we'll see a having again of that to maybe

Evan Troxel: Hmm.

Patrick Chopson: seven years. You'll see, I think most firms using ai, but like initially the first movers will have like, us and others will have like time to actually. Understand what it's we're working with and really perfect that process, uh, potentially ahead of other things. And then of course, once we perfect it, I'm sure we'll end up sharing that with others. But our thesis is that most people don't wanna try something new

when

Evan Troxel: Yeah. Right,

right. Eric, did you see that survey that I'm talking about? I'll put a link to it in the show notes, uh, for the listeners, but, but I thought it, you know, again, like it, it, it's just to me it wasn't, it was a nothing burger of a report. Oh,

Eric J Cesal: Um, I didn't actually see it, but you told me about it at the conference, uh, when we saw each other. Um, and I assume it's like every other. AI survey and architecture that I've seen from Reva or architecture or whatever.

Evan Troxel: Mm-hmm.

Eric J Cesal: and yeah, you have those striations of, you know, the people who don't wanna deal with it at all, which I agree with you, Evan, is, is hazard in terms of like how we face the public. And then at the opposite end, there's the ones who are quote air podcast, air quotes, is that what we call it?

Evan Troxel: Yeah.

Eric J Cesal: using ai. But what they mean by that is, you know, chat GBT to make an occasional query and midjourney to make thousands of, you know, renderings that nobody leads or, or wants. Um, so yeah, I mean I, I think that, um, we, we need that statement of AI governance that Patrick was talking about, and we need to be conscientious about how architects are perceived as using AI in front of the client facing public.

Evan Troxel: I wanna shift gears a little bit and talk about students in kind of the next generation. Eric, I'm curious, does this come up in your, what you're teaching at Harvard?

Eric J Cesal: Um, no. Um, but

Evan Troxel: This is a separate thing,

Eric J Cesal: yeah, no, I mean, I teach in the global development program and I

Evan Troxel: okay?

Eric J Cesal: post-professional students, so I'm not, um, not presently teaching design studio or, or anything like that, although I have, um, I see it coming up in my classroom regardless, and

Evan Troxel: Mm-hmm.

Eric J Cesal: tell you the way that it comes up in my classroom, and I think that, know, the analog's perfectly there. Um, I always tell my students that they should use ai, um, and that, you know, AI catches are not sophisticated enough to catch 'em. Um, and I can't always tell either, but they probably shouldn't use AI because it's normative. So in my class, you know, I teach on community-based approaches to disaster and resilience.

And we're seeking innovation. We're acknowledging that our existing methods for keeping our community safe aren't adequate. So we're looking for better, newer, improved methods for doing that. And AI doesn't capture that very well, right? If you ask, you know, Chad, GPT or Claude out of the box, you know, how do I keep my community safe for disaster?

It's gonna give you a pretty generic set of things to do. And I tell my students, that's not adequate here. I think that applies to all learning and all students across the board. Um, you know, AI makes it easy to give the easy response. And what we want for the future is, you know, critical thinkers, innovators, entrepreneurs, people who are really gonna push things forward.

So I think in that sense, there's, there should be serious alarm around how this is going to be used in the classroom and how we keep. Teaching innovation and design,

Evan Troxel: Kind of setting a baseline and then going beyond using it as that, and then going beyond that with your critical thinking skills, your design thinking skills.

Eric J Cesal: if you understand it as a

Evan Troxel: Hmm.

Eric J Cesal: And I think, you know, I've heard about models, um, Randy Deutsch at, uh, Champaign Urbana, Phil Bernstein at Yale, are really like teaching to it in that way and saying, look, here's ai, and then here's all the thinking that we're gonna do on top of the ai, and I think that's the right way to teach it.

Evan Troxel: Yeah. Patrick, from your point of view, when it comes to tools, I mean there, working in architectural practice, I would argue that there is a, an analogy here, which is like, uh, a lot of times the buildings end up looking like the, basically how the tool works is how you get to Yep. I can tell that's a Revit building or that's a SketchUp building.

Right. I'm again, with, with the air quotes, but it's, it is kind of, that's a thing, right? For sure. In practice and. Because we are often limited by the tools and, and what you can make in the tools. So it obviously has, has to do with the skills of the operators and what they're capable of doing, what the tool's capable of doing, understanding how much time it actually takes to do work arounds in these tools to get them to do what you want.

Right. Um, often to, to terrible results. Right. And wasting a lot of time from your standpoint, like has this really unlocked some additional freedoms? Because, and again, like I have to, I have to put it out there, like, the caveat here is your software developers coming into this and so not every architecture firm has that, I'm gonna call it a luxury, right?

May, but, but maybe you don't see it that way, but like you're coming at this from a different perspective, I think than, than traditional practices where they're buying something off the shelf like you're building versus buying in a lot of cases.

Patrick Chopson: Yeah, no, that's, that's interesting. I, I think like, man, there's a lot to unpack there.

Uh, I guess

Evan Troxel: Yeah. Sorry.

Patrick Chopson: when I think about like, uh, just kind of like what I'm, what's on my mind is when I think about like the junior folks that work in a firm and how they're, they're the ones who are going to get most impacted by the bias of the tool that you created because they won't know how to overcome it,

or

Evan Troxel: Hmm.

Patrick Chopson: make something that seems technically. Uh, complete, but it does, but it's not beautiful.

Or

Evan Troxel: Hmm. Mm-hmm.

Patrick Chopson: create a piece of writing where they won't be able to, they won't be as good at writing or as good at like saying, here's what I'm trying to get. And then the AI as it interprets the intent like create something very interesting or compelling. Whereas like if I have a tool that generates images and it's trained on my hand drawing style and then I have a junior person use that, then it makes things that are like what I would do. we've actually have like created different profiles for of trained image regeneration models for different principles so that the work like their, their work if they're having a junior person work with it.

But I think like that is an interesting point though. It's like the bias of the tool. I've always been a tool bias person in studying that. And, uh, I think I. What I look at when I'm thinking about building an AI system is not it necessarily like an image generator where it's trained on things from the past, but train it on how to think about a design problem. And so I tend to look back to like somebody like Nicholas, uh, Nicholas Negroponte or uh, maybe Christopher Alexander with pattern language. And I'm like, what is the design thinking methodology? And then how do we organize the information that we can use that design, think thinking, and keep everything in mind as we're like making And then hopefully we create a, a model and a way of working that's biased towards an actual of making things beautiful and interesting and compelling, rather than trying to teach against a certain previous existing dataset so we can generate new things based on our human So it's like how do we achieve a, a tighter fit between a human intent and the machine's resolution of that intent?

what I'm.

Evan Troxel: Yeah.

Patrick Chopson: Always thinking about as we're building stuff.

dunno.

Evan Troxel: I'm curious how, how do you actually extract that from people? Because I, I had a guest on recently and they talked about expertise and how experts don't even know what they know. Right. They just know it, and it comes out in different ways throughout. The different parts of a project or in conversations or, and, and, and so I'm curious because you, you talked about that, that idea of like, keeping it all in mind, right?

And, and that's hard for humans to do, to keep it all in mind. Like, like you may hear about a particular design constraint and then it's like a, you immediately forget about it and you design something and it's like, oh, it doesn't work because I forgot about that one rule. But the computer doesn't forget and you can just continually add additional constraints so that they're always in mind, you know?

So, I, I'm curious just how, first of all, how do you extract from, from individuals? And then do you put that into your models so that it is always in mind?

Patrick Chopson: Yeah. So the closest analogy I guess, of where this has already been com figured out is in the software industry where now you can code very large applications, uh, from scratch

with,

Evan Troxel: Mm-hmm.

Patrick Chopson: to have like the standards that you're referencing. You have like the structure of the project that you're looking at.

You have to have all these documents that help you vibe, code is what people call it nowadays

instead of,

Evan Troxel: You're the first, first person to bring up Vibe coding on the podcast. I've been waiting for this to come up.

Patrick Chopson: So like we actually went through this first with our software engineers. So of the things that was super valuable about this whole process is that. actually have an industry where they have been totally, in the game has changed. You can code now using ai, a hundred percent the whole project. when we took our software team, we were like, guys, you're trying to use copilot. You're still writing things line by line. Let's all work on learning how to code using a different way of

coding.

Evan Troxel: Mm-hmm.

Patrick Chopson: of the most valuable things that we have in our business is that our, um, head of software engineering is a older guy who's been through four different ways of coding through his career.

So starting from like, you know, I think the eighties, two 20, 2000, 2010s, you know, he's been able to like see these different ways of doing what he's doing. then when we showed, when he and I sat down, like how are we gonna design an AI coding best practice, we actually had templates to pull on from the past of like how to this.

And so like. Changing how people thought about coding was very, very challenging.

And so

Evan Troxel: Yeah.

Patrick Chopson: once they got over that hump, now they're all, know, building stuff at lightning speed and they're building more things than they could ever build as a team of 10, before. So that's kind of like, I don't know, it's like we've already been through the process now with the software engineers, so I kind of have a template for how to like, architects get over the same hump.

That's a very long answer. So

I apologize, but.

Evan Troxel: it was great. That was, and there's so much stuff in there that, that I want to, I want to explore and I don't, I don't know if I can even recall it all talking about keeping things in mind, but that my, my first one that I want to talk about here is just this idea of you driving this process because I, and that sounds very top down.

Versus a grassroots kind of a thing, which I think is what most people are experiencing in firms today. It's like the technologists are the grassroots, they're the ones who are the doers. They're writing the scripts, they're checking out the new tools, they're experimenting, they're doing these things, but there's somebody at the top saying, no, we have to do it the way we've always done it.

Right. And so, I mean, I know this is kind of a common trope, but at the same time it, it's not untrue. I mean, a lot of people experience this and so from you driving this process, change this, right? Like, like that to me is a, is a big difference.

Patrick Chopson: I mean, that is the difference. Uh, right

Evan Troxel: Yeah.

Patrick Chopson: of have the, the foxes are in the henhouse to a certain extent when it comes to like, innovation the a, EC. Like, you have folks that don't, that aren't on the edge, but you do have a few firms where people are. do think that a lot of your Gen X, um, kind of principles who are now assuming control of a lot

Evan Troxel: Mm-hmm.

Patrick Chopson: uh, or millennials in some. You know, limited cases, I guess, uh, those folks do have, uh, sometimes more of a technology bent to them. so it's possible, it's only possible to create an AI powered architecture firm if the leadership is aligned around like, we're doing this different

and we're gonna

Evan Troxel: Yeah.

Patrick Chopson: And then everyone's gonna feel uncomfortable for a while.

But if you kind of like live in that state of discomfort, it's, that's where you're gonna find the innovation to actually do it. And then some things you just can't do with new tech. You know, like hand drawing is still the fastest way to communicate with ai, I think. Uh, your design tint. So we

have

Evan Troxel: Interesting.

Patrick Chopson: on every desk.

Evan Troxel: Nice. What can I ask you? What generation you are actually.

Patrick Chopson: I am a geriatric millennial, I guess is what they call me.

Evan Troxel: I, I was the Gen Xers out there. We, I, I heard this recently and it really put, threw me into like a, a perspective, oh, okay. We are the last analog generation as Gen Xers, and obviously we are not digital native. And so it was just kind of a new way to say that, right? It was like, oh, we're the last ones who grew up with analog as like the, the way it all started.

And then many of us transitioned, but we weren't, we're not digital natives. Eric, what do you think about this? I think it applies to education too, the thing that we're talking about here, but, but when it applies to education and the architecture profession, what do you think about this? Like, top down and versus grassroots and like the, the position that many firms are stuck in, like literally stuck in, um, because of the, the leadership.

Eric J Cesal: Well, a few thoughts. First of all, shout out to Gen X, Evan. We gotta stick together. So, um, uh, I appreciate that. Um, I always tell people, like I, I was probably like the last class at school that was taught how to do hand lettering. Um, and I

Evan Troxel: Mm-hmm.

Eric J Cesal: do it. I still write like that. But, um, a few thoughts, uh, you know, about the, tool thing.

You know, a metaphor that I used in, in that article that you referenced in the beginning. is that, you know, when, when 19th century traders crossed the mountains crossed a mountain range, they followed the path that was most hospitable to the horse, right? And so, lowest elevation, fewest rock slides, things like that, 20th century engineers went through the mountain to take a tunnel, they followed the path that was most favorable to geology. nowadays, when we fly over the mountains, we ignore both, and we just follow the path that is appropriate to jet stream. So I think we need to be open to the idea that technology is speaking to us about the way in which it wants to be used. And

Evan Troxel: Hmm.

Eric J Cesal: AI evolves, it won't always be the case of, you know, us trying to make AI do this, but us listening and saying, you know, look, by the nature of this technology, it suggests that we do something different.

And maybe by coding is an example. Love that. Um, you know, just kind of changing the paradigm about coding. Um, the other thing is predictably a concern, right? Because, you know, I'm the disaster guy. Um, you know, I wonder how these tools progress, um, and at what point they diffuse into the wider population.

You know, it was one of my first concerns about Midjourney, you know, two years ago, two and a half years ago, uh, you know, Arctics were having a field day. And, and my thought was, Hey, you know what? Everybody else can use Midjourney too. and you know, I wonder as these tools evolve and they make the process of design more intuitive, more accessible, you don't have to spend years learning, you know, Adobe Creative Suite or Revit, you know, it's more of a conversational platform that you're working with. Does that, does that open the door to additional predation and additional devaluation of, of architectural. Services. I agree with the prior comments that you really need to know what you're doing, you know, at least at this point, um, to know when the AI is wrong and to apply that human judgment to the equation. Um, but yeah, uh, experts don't know what they don't know and, you know, neither do regular people. So I think as a regular person, you know, meaning like a client, it comes very, very hard to delineate between something that is professional level quality and something that is 95% of professional level quality.

So, you know, I think the tool conversation is, is interesting, but I have, I, I feel like we have to keep it in frame.

Patrick Chopson: Yeah.

Evan Troxel: I'm curious, Patrick, from your point of view when it comes to that, like what he's talking about. It used to be that people would come with their own SketchUp model or their own, you know, they would be drawing, I. Floor plans in Excel, for example. Right. Um, but, but it's happened, right? Uh, and so is this the, the new version of that?

I think it is, right? But the other, the other thing that we have to keep in mind is that the ai, and maybe maybe this isn't the case, but it has been the case, is AI doesn't understand space. And that's where architects, the value comes in when it comes to these things and actually figuring that stuff out.

And, and I'll layer on top of that, that I think a lot of times being a quote unquote architect, you know, a graduate of an architectural program, you still have many, many years until you understand those kinds of things in actually producing architecture and experiencing it and understanding. Quality of those different things and how they affect the people that inhabit these spaces, right?

Because it's not a 2D image on a screen, right? It's a spatial experience and it a progression of spaces. It's function and adjacency and response to the environment and lighting and all of these layers of, of nuance that a lot of people just, you know, here's, here's, here's what I want it to look like.

It's like, well, there's a lot more to it than that, but give us kind of your perspective on that.

Patrick Chopson: I, I guess like as the AI becomes more aligned around your design intent like, what is you're trying to do, I think that if you don't have good taste, it's gonna align more closely to your poor taste. Um, then not, you know, so it's like, if anything, as the tools become more sophisticated, they'll interpret what you're trying to do and, and like. still, there's been a whole long thread on Twitter, uh, or XI guess now, um, where all these property developers, that's where they have their conversations. They're not on LinkedIn. They have a rip roaring time on X

and they always are chatting

Evan Troxel: Mm-hmm.

Patrick Chopson: and this and that. But one of the things you see a lot of is people like, Hey, I threw this image of this strip mall into, into my thing and I got back a beautiful rendering and, and it doesn't look beautiful, uh, what they made, but they,

But but from their perspective,

Eric J Cesal: it is. Yeah.

Patrick Chopson: think it looks pretty good, you know, like, but what they were saying, which I think is the part that a lot of architects who might see that and freak out is they were like, I just need to show the person who's gonna give me money that this could look better. And I think like our architect as architects, we should think about, like, that's kind of like why they were hiring us to begin

with or doing

Evan Troxel: Mm-hmm.

Patrick Chopson: study is they just needed to have to secure that money. And then they spend the time to like getting the construction documents done and know. No reasonable developer wants to do that work themselves.

They're just, it's just not what they're looking to spend their time on. They wanna find the next deal and they wanna analyze deals and for them, everything's a spreadsheet anyway. The whole creative side is not something that they really are wanting to do. So I think like you'll probably just see, like, maybe you'll see people use it to like up with a concept similar to like the Excel analogy of someone drawing a plan in Excel and you'll, it'll still be not that great compared to what an architect will have access to in the future the AI will have progressed on our side much far in advance of what a consumer will have access to because we'll either be billing ourselves or know, maybe someday a big company will figure something out, but I doubt.

Evan Troxel: An unnamed large company. Yeah.

Patrick Chopson: That makes a billion dollars a year. They'll, they'll buy some, some something, you know,

that will do something. But

Evan Troxel: Mm-hmm.

Patrick Chopson: have access to tools and AI agents that will be far superior than what a regular person will have. us with all of our training will be able to operate them more effectively, even if they are more sophisticated. I don't know. I, I remain optimistic that the profession will be, anything reinforced by good applications, but I fully acknowledge there'll be bad applications too.

Evan Troxel: Yeah. The, the idea that, um, there's, I mean, equal access for all right, it's the same, it, it, it's just like a cell phone, right? Like the mo the richest people in the world have the same cell phone that you do, right? Uh, they don't have a better phone. Um, and, and I think the tools in this, for the most part, are like that.

But to your point, Patrick, obviously there are specialty tools and, and if you can build tools yourself, I'm, I'm just curious, and I think this applies again, you guys are both in kind of different aspects of. A EC, probably a better way to say that, but I'm just, I'm curious, like what do you see as kind of from, from a standpoint of experimentation, how is and, and making tools?

I don't know, Eric, if you're making any tools or if you're just really using the tools, but, but is this a, a thing of like, you have to get in there and experiment and learn what these tools are capable of or, or are, and, and, and how much time can you actually spend doing that? Or should you be doing that so that you are at least understanding what the landscape is out there?

Eric, why don't you take that?

Eric J Cesal: was that

Evan Troxel: Let's let, it was at both. It'll be eventually at both of you. So yeah, I mean, from, from your standpoint and especially when it comes to teaching and writing, like how, how are you, what, what's your stance on that and how, how are you approaching this?

Eric J Cesal: Um, how am I approaching?

Evan Troxel: like tool usage, experimentation. How are, how are you taking that and, and, and even tool building, like, are you doing anything on your end to enhance your capabilities so that like, again, like as these more specialized tools do emerge, that you're ready to actually put them to use.

Eric J Cesal: Yeah, no, I mean, I build tools all the time. I mean, I don't design buildings anymore, and that's not a part of my professional life. So,

Evan Troxel: Mm-hmm.

Eric J Cesal: not in that space. Um, but, you know, I develop specific single use software for all sorts of things that, that I need to be doing. And, um, I think, you know, I always begin with the question of, okay, you know, what is the start and what is the finish?

You know, what am I actually trying to accomplish? And then what is the best approach to doing this? You know, is it some off the shelf application? Is it, you know, some app done by some startup that I sign up for free and then, you know, two weeks later it's charging my credit card every month and I forgot about it and, and that sort of thing? Or is it actually a bespoke solution? I mean, is it. Straightforward enough that I can, you know, just jump in the Chad GBT or Claude and, you know, code it outright. So I, I mean, I think that, yeah, I just circle back to this idea that this needs to be a period of time where we're very open-minded about what it is that we're trying to do and what is the ultimate start and the ultimate end. Um, you know, Evan, the last time we, we talked, we talked about Uber, uh, as, as a case study. And I think that, taxi cabs were very, very protected market. They were very well entrenched, um, legally, uh, illegally, uh, culturally, you know, I mean, this is like a very kind of fixed part of civic life, um, especially in some cities. And, you know, it was disrupted because they failed to realize that nobody gave a shit about taxis. People just wanted to. Get from one place to another without using public transportation or their own vehicle. And technologists were able to step in there and say, you know, if we take a point A and point B, like how does technology solve that problem?

De novo. So, you know, if I was at, at an architecture firm, like I would be asking that about every process. Um, you know, the smaller ones and the larger ones.

Evan Troxel: Mm. Yeah.

Patrick Chopson: No, for certain, and like, I think when I look at, look at it like this, if, if you're a leader of a firm you're not willing to like. Try out these tools and find the boundaries yourself. You're never gonna know what you can and can't ask someone to do. think that's

like the way I

Evan Troxel: Mm-hmm.

Patrick Chopson: with ai, you just have to try because the boundary between what the human can do, what the AI can do is not a hard line. It's a very squiggly line. If we kinda look at like some things you'll be like, oh my God, I can't believe I can just press a button and get x and other things. You'll be like, I should be able to do that. But it's far outside the capabilities. Um, and so you have to constantly test that boundary, but then the boundary is always changing.

So as you can add more new models, you to understand what you're doing, you have to kind keep picking away at like, here's a problem I've been trying to solve for the last iterations of the AI model. Can it do it now? And you try and you're like, oh, oh my God can do it

now. Then that's like,

Evan Troxel: Right.

Patrick Chopson: you gotta keep experimenting.

'cause then I can't go to my software team and say like, I. Do it this way. 'cause I read this on a, a blog somewhere like that, that's not the right way to lead. Uh, as a leader, you have to, if you wanna lead on technology, you have to it yourself and you have to learn how to do it yourself. And then if you're the per, it's like, there's this guy that, he was my best boss I've ever had. I was unloading trucks one summer in college and he didn't come back there and say, you guys are going too slow. He would say like, you guys are going too slow. Let me show you the speed that I want you all to work at. And he would go in there and he'd start throwing stuff off that truck. He was a reasonably. older guy and he was out doing us young kids like just moving these boxes, you know? And I remember like thinking like that he'd work that way for like 15, 20 minutes. He'd be like, no, keep doing that for the next four hours I'm gonna come back here. I wanna see this truck unload. And I think like that's the kind of like person that you really wanna work for them.

'cause you're like, I don't wanna let somebody like that down

'cause they're doing the

Evan Troxel: Mm.

Patrick Chopson: well. And I think

Evan Troxel: Mm-hmm.

Patrick Chopson: leaders that are more your servant leader type, I

guess is what they say

Evan Troxel: Mm-hmm.

Patrick Chopson: like that. You gotta like show, show it first and then other people will believe you.

Especially when it comes to AI

Eric J Cesal: Were you

Evan Troxel: So when,

Eric J Cesal: or by the

Evan Troxel: yeah.

Patrick Chopson: by the hour, you know?

So.

Evan Troxel: More boxes in less time. Yeah. I mean, this is, this is architecture too, right? It's like, do do it in less time. It's like, uh, well better design in less time. It's a tall order. And so I, I'm curious from like a tool stack standpoint, Patrick, can you, can you give us some ideas about

Patrick Chopson: Hmm.

Evan Troxel: what, what you guys have taken on versus what build versus buy?

Is there any buying? Are you doing it all? Like, like give us just kind of a, a glimpse into what, what's going on at Cove?

Patrick Chopson: I mean we, we've taken a philosophy that you probably can't find something that does what you need. And really, nowadays with ai, you can copy any software. There's no barrier now to making software for real. You can literally just take a screenshot of something and you can make it in a few days.

Uh, so

Evan Troxel: Hmm.

Patrick Chopson: in that world then, it's kind of more incumbent upon the company, I think, to look at what is it that we should make? Start from there, against that. So we use, uh, we use a lot of React tailwind. Uh, that's kinda like one of our, kinda like front end. And then on the back end we use a lot of Python. Uh, we use a lot of things like that. And we use AWS obviously for like servers or you know, some people use Azure and, you know, whatever your tech stack happens to be.

But I think like a lot of, uh, architects are obviously not gonna be like, okay, well how does web development work? But we feel like web tools are like the best because they allow us to connect more things together

than building like

Evan Troxel: Mm-hmm.

Patrick Chopson: app or something like

that.

Evan Troxel: Sure. Yeah. I, I'm, I'm curious from your point of view then, when it comes to the words that you just said, react, tailwind, a WSI mean, those are foreign. Words and acronyms to a lot of architects. Right. Especially, you know, traditionally trained architects. So like, what do a architects need to be doing about that?

Are they hiring a staff? Are they outsourcing this to a consultant or outsourcing it to, you know, anywhere? I mean, I, my, that's a question I think a lot of people would have. Well, where do I start?

Patrick Chopson: Yeah, I mean I think like if you look at some of the tools that are out there, if you kind of start following the AI hashtag on like honestly on x, I know

Evan Troxel: Mm-hmm.

Patrick Chopson: platform that for political reasons some people like to be on, but like it is where the discussions happening around ai, if you wanna see what's going on real time as a new model comes out, people start testing that thing.

They use different tools. They'll say, I tried this tool and it's totally awesome, or This one sucks, and you can see that and then you can go try it yourself. that's where you want to be on LinkedIn. You're gonna see stuff that's maybe three, four weeks behind and in some cases even months old.

And so it can be a little stale.

Evan Troxel: eternity.

Patrick Chopson: And like, and you can also just see that like people who don't know what they're talking about. a lot of those people on LinkedIn. Uh, whereas on X, like you get exposed pretty quick if you don't know what you're talking about. So you kind of start to, you see a real learning culture of sharing

stuff on, on X.

Evan Troxel: Interesting.

Patrick Chopson: like, if you wanna see what's, what's happening and what tools are the right ones for you're watching this podcast, like make sure you know it's the latest, uh, at that time. 'cause then there's something, there's some things I've tried last year that I was like, this is game changing now.

I look at it, I'm like, Hmm,

not really a thing now.

Evan Troxel: Interesting. Yeah. It's moving so quickly. So yeah, the, the, the way that I would explain it is like x it's, it's unfolding in real time. It, it, it kind of, it's always been that real time

Eric J Cesal: Yeah.

Evan Troxel: timeline, the feed of what's happening now. And LinkedIn is like the polished version. Like, I would like to now present to you the final results of the thing.

Right. Um, and, and it's way more like, I mean, obviously they're bigger posts. They can, they, they're, it's a different audience. I mean, it's like, show, I'm gonna show you my accomplishments and now you will applaud for those accomplishments, right. Versus this kind of just exposing the work in real time version.

So I think that's a really insightful thing to say, even if it isn't the place that people want to be hanging out, there is some incredible information. Shared and, and like, you get to actually look under the hood and see what's going on behind the scenes from Yeah. Eric, Eric, any, any ideas or like thoughts that that brings up for you?

Eric J Cesal: uh, I would agree a hundred percent. Um, I mean, I think Twitter is where those conversations happen. It's where I get a lot of my news.

Evan Troxel: Mm-hmm.

Eric J Cesal: and I think you both characterize it correctly. You know, like some model comes out and three hours later, you know, I. Some guy is like, Hey, here's what I did with the model.

Like, here's like the use case and you can explore through that. I mean, personally, in terms of my explorations and keeping up the news and stuff, I would, I would also throw YouTube in there. Um, I think you need to filter it a little bit, but, you know, there's, there's a lot of the same effect. Only slower than Twitter, faster than LinkedIn.

Evan Troxel: Because it takes more, more, more, more to actually create of a YouTube video versus just Yeah. Throw something on x. Right.

Eric J Cesal: phenomena. You know, you get somebody that, you know, takes something new and make some incredible experiment with it. And then, you know, 12 hours later is giving away the, the step-by-step how to, for free. 'cause they wanna click views or, or whatever. just because of, you know, my situation, I, I favor a lot of academic papers as well. Um, and I know a lot of people don't have access to those kinds of journals if you're not affiliated with an institution, but I really like to be aware of what's going on in the lab, um, and what researchers are doing because they know that no one is ever gonna read their paper. You know, so they just kind of put it out there and they're like, eh, you know, here's what I did.

Give me

Evan Troxel: Here's a thread. Yep. Here's a thread.

Eric J Cesal: Um, but no, you find a lot of interesting stuff that slips under the radar there because

Evan Troxel: Mm-hmm.

Eric J Cesal: of times, you know, academics are so, so down in the weeds, um, you know, with their work that they don't particularly understand the professional applications of it.

They're not, they're not looking towards that necessarily. So if you know a little bit of something about practice and you know how buildings get made and you can suffer through an academic paper, you can often draw a lot of connections about what's coming.

Evan Troxel: Patrick, um, I have a question about kind of hosting your own models versus just using publicly available models. There's a lot of models out there. Are you guys doing any privately host, like you're taking their model, putting it in onto your hardware and running it locally? I'm, I'm curious because a lot of firms have constraints around that with the types of projects that they're doing.

Um, not only with. What information can go outside of the network. Right. But also just like, I think this is also part of that trust thing too, and and control thing. So where do you guys stand on that? Or is it a, a little bit of everything.

Patrick Chopson: So like we, we do have to abide by FedRAMP style stuff, you know, in terms of like what, what hosting it yourself is probably the one of the least secure things you could do. Uh,

to be

Evan Troxel: Hmm.

Patrick Chopson: uh, AWS if, if you're in their, one of their super locked down data centers, your data is a lot more in terms of if you're making, making tools.

So, like, in some cases, like, uh, web-based tools can be more secure than a local application. But, um, when it comes to like, sharing your model something with a model, like it's, I don't know. You there's Open AI is probably the one that's taken the, or Claude, to be honest, or the with from Anthropic. Those two are focused on winning government contracts and so they

Evan Troxel: Mm-hmm.

Patrick Chopson: created versions that are more secure than others. Uh, I think the biggest risk you could do is like, try to use like deep seek or one of those open source models that you don't know who created it and try to run it locally. That would be like, to me, the most risky thing you could do.

Uh, 'cause you

Evan Troxel: Mm-hmm.

Patrick Chopson: know what biases went into it. I think you, I'll oftentimes think it too, it's like this is a continuum and if you design your code where you can swap out the engine, which is how we design everything, we'll swap, we'll swap out our model every time there's a new model we could switch from, know, one model to another without any. Problem. You know? So I think like you just have to design your system such that the brain can be popped out. Anyone can be put in, uh, fairly easily.

And because, you know, things are,

Evan Troxel: Yeah.

Patrick Chopson: you start building something to the time you finish building something, even if it's a few weeks, it'll already be out of date.

So as well just kind of like, focus so much on like the, I don't know, I, I just focus on making sure that it's something that's backed by a entity that's trying to win contracts with the federal government.

That'd be

Evan Troxel: Yeah.

Patrick Chopson: that would be my, uh, advice, uh, if I, from a security standpoint.

Evan Troxel: Okay, well, final question, I'm gonna turn the tables to for you both and you're putting you on the spot, and so do you guys have any questions for each other? I, I know that that to me, like may, maybe this, maybe this podcast sparks something there, but I, you know, Eric, do you want to, do you want to jump, I mean, I would love to pass it over to you and just say like, what, what have we, what do we need to get into the end of this episode before we sign off?

Eric J Cesal: Um, I, I really do have a million questions, Patrick. I hope we, we talk again soon. But, um, I guess if I had to pick one, um, you mentioned something earlier about, you know, larger firms disseminating technology or, or maybe even your yourselves passing that along. How do you see that landscape unfolding?

Because, you know, there is, really kind of an unfair. Situation. Uh, I don't know if it's unfair in some sort of moral philosophical sense, but you know what I mean. You have large firms with staff that can dive into this AI stuff, and then you have a whole bunch of very, very small or one person firms, um, that maybe don't have the resources or the expertise.

And I've just been struggling to understand how that plays out. You know, do we end up with like 10 architecture firms in the future because they kind of captured everything during this AI wave, or is there some technology transmission whereby, you know, large firm pioneers, something and to the way the chat, uh, open AI and, and some of these other companies have been working like, we'll give you the model that's like three months old for free and like you pay for the new one.

How do you, how do you see all that playing out?

Patrick Chopson: so I guess I look at it this way, like in private equity where you roll up firms and combine them together, they kind of are shooting for a 10% of the market to create a major player. So like when you think about I think will happen. Like there'll be, someone will come along and start rolling up small firms to create a new large mega firm. think that'll happen. I don't, when I look at, I've in large firms, there's so many. The multi-office firms that are out there will struggle because they'll be hard to enforce a top down. Like AI adoption strategy individually within those firms will be brilliant people. There'll be certain offices that may be more or less using ai, but like, I don't think that, I think being a large firm right now is a disadvantage.

Uh, it's kind of the dinosaurs versus the mammals situation here. Um, I think if you're a smaller firm, you can enforce a, a more aggressive technology approach. And if you're a principal, you can, like, if you're the one who learns it first and then enforces it, you can kind of like say like, look, you either can work here and do this, or you can go work somewhere else, but I know I can, you know, build a firm that's gonna do this thing. I think that's what you'll see first is that the smaller firms may actually it on faster and. Maybe not like five person firms, but like a 30 person firm is a good example. that size is probably more nimble and more than likely the principles can get aligned around a particular spend or investment approach. So I do see that probably happening and then I think you'll start to see that as the smaller firms that have adopted AI grow a little bit larger. We'll see a, right now there's a hollowing out. We have the small to big, it's too big on the top, too small on the bottom. So I think we'll see more of a graduated.

I. Range. There'll be the boutique firms who are never gonna use ai. And you come to me 'cause you want your building hand drawn.

I think that will still

Evan Troxel: Yeah.

Patrick Chopson: But then there'll be the

Evan Troxel: Mm-hmm.

Patrick Chopson: are like ai. You don't wanna use AI for your project, where the old, old school people come to us because we're, the expensive Louis Vuitton handbag kind of thing.

Like, that'll still happen too. So, but I think there'll be a continuum of people in the middle though, who are all using technology to deliver a unique level of design service that far exceeds either the large boutique firm the outside or the tiny, tiny, you know, firm that works for the homeowner. You know, there'll be like those guys in the middle that, that group will become probably 50% of the market is my guess. Um, in the next five to six years, you'll probably maybe 10, you know, if we wanna be conservative, but know, at least by 2030 you'll start to see some. Some firms that are like in that middle ground, probably occupying more and

Eric J Cesal: Hmm.

Patrick Chopson: of the, buildings that affect the maximum number of people.

Eric J Cesal: Interesting.

Evan Troxel: Patrick, your turn. What do you got? You got a question?

Patrick Chopson: I guess. Like man, I'm, I'm just thinking like, I need to hire a lot of people. So as I, so as I'm thinking about like, you know, new folks coming out, I've been engaging a lot with, uh, Casey Castor, you know, over there at S Arc and some of his students, I've hired them they're really kind of well fitted to this problem.

You know, they're kind of, they

Evan Troxel: Mm

Patrick Chopson: use ai, they have experience like doing some random project where they may or may not have had success in creating a building. But certainly something really interesting is the idea that's being explored. just wonder, like, as you look across the educational landscape of like what we need to do to kind of help folks get ready. What are you seeing that's been the most successful at teaching? 'cause I'm actually teaching a lot of people who are, are hiring in-house, uh, to do these

things. And it's like,

Evan Troxel: sure.

Patrick Chopson: what are you finding is a more successful way to help kids overcome, like, know, the whole like, oh, AI can just make anything.

But does that mean, like the meaning of it, the, like the, the true soul of what you're making? Like, I don't know how, is there anything that you found that's

successful with them?

Eric J Cesal: No, you gotta give me the hardest question. Um, no, I mean, it's, it's super challenging, right? Because I mean, I, I am old fashioned enough to believe that you really can't be a good architect unless you build, you know, like unless you've like, been on the job site and swung hammers and like that, that touch and feel of materials, that that execution of the process gives you the intelligence to, um, intelligence broadly speaking, um, to actually like design things in, in the right way. Um, and I think not to cast dispersions on, on architects who don't um, you know, they, they have a role as well, but I think that, um. You know, we need to stay connected to that in some way. And I worry a lot for education and we've already seen this effect show up in other professions like medicine where, you know, the AI is not good enough to do what the senior person does, but they're good enough to do what the junior person does so that junior person can't get into the job market and start making those, you know, early mistakes and, and gaining that experience.

Um, so I think that's a broad systemic problem. And I would address it by zooming even farther out and perhaps being more useless. Um, I think AI is going to fundamentally change what we hire for. you know, working in disaster zones, you know, is funny. You know, I would hire a lot of architects and I was always prioritizing certain things. Adaptability, resilience. Uh, humility, um, you know, people who could be humble in front of a problem that they didn't necessarily understand. And, you know, when I survey the architecture, job boards we're still hiring in the same old way. You know, we're like, I want somebody who knows this software and has worked

Evan Troxel: Yeah.

Eric J Cesal: kind of project and, and this sort of

Evan Troxel: Yeah.

Eric J Cesal: think those considerations are just getting less and less meaningful by the day. in terms of training people, I always found that if you hire that person based on, I hate to say it, but the non architectural qualities, you know, the things you can't teach, um, you know, courage, resilience, uh, adaptability, um, personability, that sort of thing. Um, then they can teach themselves at some level, right? And they can adapt to whatever's coming and they can learn from the twists and turns that, you know, the world is bringing. Towards them. So, I mean, I think that as educators we have to be cognizant of that shift as well and, you know, talk to ourselves about, you know, the skills in the future.

Like they're not gonna be like this software or how to do like this calculation or, or something like that. I think there are certain things that are going to be demand as we fully realize an AI future, and we have to teach to those and you have to hire to those. So,

Patrick Chopson: yeah.

Eric J Cesal: Yeah,

Patrick Chopson: I, I just seem like when I'm looking at folks that are successful, it's the problem solvers who are really abil the ability to think your way out of a paper bag sometimes is like what I've seen as the quality that I'm always looking for. Uh, but I've really seen that like, an uneven. Ability of people to reason their way out of a problem, uh, when, you know,

uh,

Eric J Cesal: No, and I think it's challenging in architecture because I think design training has a way of marrying you to process. You know, we, we kind of go through school in those early parts of our career and we develop methods that work. You know, I, I design with a micron pen on yellow bum wad, and that's just like how I things.

And architecture rewards that kind of monogamy because like, you can produce great results consistently. Um, but, you know, bag or no, like being able to say, you know, this is what the problem is, so this is what my reaction should be, and if that doesn't fit with the method or skills that I have, gotta get new skills, develop new methods.

I think that's, that's what we're looking for.

Patrick Chopson: yeah.

Evan Troxel: Change management. I mean, it's so difficult. I mean, you talk about just kind of like this. Re you, you're, you get used to the way that, that these tools work. They, that you get rewarded for that. And that just re this reinforcement learning. I mean, it's, it's hard to escape those loops and especially, I mean, it takes 30 years to get into a position of, you know, stability in, in this industry

Eric J Cesal: Yeah.

Evan Troxel: and then what, right?

And then, and then it's like cruise control turns on and like, don't mess with, with this system. Right. Um, and, and I we're, there's a ton of that going on too. And I, I think that the, one of the biggest things that I keep seeing is. There's kind of this cheerleading by certain individuals in this industry that are like, use ai, and it's just so generic, right?

It's like architecture has to transition to, it's like super generic. It's like, well, what is that? And, and I think it's just that on some level, it's like the message has to be that simple. And at the same time it's so overwhelming because tool fatigue and headlines constantly changing. And even we're saying it today, it's like, well, it didn't do it three weeks ago, but now it does.

Right? And it's like, how do you keep up with that? And so a lot of that is just, all of these things are putting pressure on people. And, and at the same time it's like, well, like the project deadlines are, are, are looming, right? We gotta, and, and so we're gonna do it the way that we've done it because we know what it works.

So, uh, ugh. Tough situation. It's a tough, it's a tough situation to be in. For sure. Well, thank you guys for taking the time to have this conversation. Thank you both for asking the best questions at the end. That was, that was amazing. Uh, and we'll have links to both of, uh, Eric and Patrick in the show notes for this episode, so that you can connect with them and read Eric's writing and see the latest on the AI first firm that Patrick is championing at Cove.

And this is, this has been fantastic. Thank you both so much for, for your time today.

Eric J Cesal: Thanks you both.

Patrick Chopson: Yeah. Thank you.