Why Ignoring AI Could Spell Doom for Your Architecture Practice ft. George Guida

George Guida
===

Introduction to AI in Architecture
---

[00:00:00]

Stephen Drew: Hello everyone. It's that time of year. Are we in the future? Are we in the past? Are we a computer? Who knows? Because what I'm talking about is ai. Artificial intelligence. Even my dad is playing around with chat GPT, but we're not just typing little bits and bobs in. We're not just asking basic questions 'cause we're gonna zoom into Architecture.

You might be playing around with it. In the office now is 20, 25 and now is the time. Where we're all doubling down and realizing that ai, it ain't going away. It isn't a buzzword. So what I want to do on this show is have people in the industry who are actually. Developing, using ai, doing really cool stuff that is gonna permeate Architecture practices and we're gonna get their opinions on where the way the industry is going.

Meet George: Architect and AI Enthusiast
---

Stephen Drew: So on [00:01:00] that note, I have an awesome guest that I've been get beginning to get to know who's been doing some very, very cool stuff in the AI space. And that is George with me. George, the founder of X Figu and ai. George. How are you sir? Are you okay?

George Guida: I am great. Thanks for having me.

Stephen Drew: Well, there we go, George. So now while we get to know each other, some people might not know I've met you yet. Do you wanna tell me a little bit about yourself and your background? First of all,

George Guida: Yeah, of course.

George's Journey into AI and Architecture
---

George Guida: So I am an Architect by training. I went through Oxford Brooks, I went through Architecture association. I worked many years at Fosters in London and, and then continued my education at Harvard and ended up teaching, starting my own practice and diving deep in the last few years into this world of AI and Architecture.

That since brought me to consulting large and medium Architecture firms on AI [00:02:00] strategy, tool building, all the likes of it. In the last year I have been building X figura, so I'm a sole founder. We're now a team of five people and uh, we launched the platform three, four months ago, so I'm excited to talk to you a little bit about that too.

Stephen Drew: I can't, we're gonna go into it together. And it all sounds exciting. Let me just understand a little bit. So we both studied Architecture. Right now, I've gone down one crazy little route. Now you have people that are building the buildings, and then in the practice you get some people that go in the more technical route.

Some people go in the more front end route. And then there's a few people like yourself, George, that get caught up in the technology and push that ahead, which pushes buildings ahead. So when did you first start thinking, oh. This is a space that I'm interested in. When did you start playing around with tools, technologies to further advance the projects you work at Fosters and other stuff?

Early AI Experiments and Innovations
---

George Guida: Yeah, I guess it all started first at the Architecture [00:03:00] Association where I had my first introductions to the the Grasshopper world, the world of computational design. And, but I really started getting a sense of its true value at Fosters, where performance based design is really at the crux of the design process there.

Auctioneering testing different environmental solutions early designing through numbers as well, as well as aesthetics. But that's when. Something clicked in me and that brought a passion into computational design, robotics in Architecture, which became a big thing that I was interested in. And then the natural next step became AI in its very early days.

Um, and so I was, I guess I timed things early well.

Stephen Drew: Fair play. I remember when people were talking about chat, GPT circa 2023 and it started to come in. There's been a lot of advancements since then. So, I mean, do you want to tell me quickly, what was the early stuff that you guys were playing [00:04:00] around with George compared to how far it's come now then?

George Guida: Uh, I guess my, the earlier uses was with tools like in Grasshopper using Galapagos or multi objective optimization tasks, but the, my real first uses were using. Models like style, GaN shape GaN clip Vq GaN, which are fun. All fun words. But um. I started diving deep into research in applied machine learning in my earlier days at Harvard.

So I worked with a professor called Andrew Witt, he's a professor in Architecture there, and we dived into training data sets, um, classifying different Architectural data sets, or simply using, for example, grasshopper scripts to create. Also create synthetic data sets and train those and build new 3D models.

So all it was kind of those early days, which now are more [00:05:00] consolidated into image or 3D generative models.

Stephen Drew: Fair play. Wow.

Launching X Figu: From Concept to Reality
---

Stephen Drew: We're gonna talk about your business in a second, but what was the point, George, where you were just, you were there, you were doing all this stuff, you're pushing the boundaries. There was all those acronyms there. Some of these softwares I hadn't even heard of myself, but when did you think, right, this is my call in.

I should do that. And was it a tough move to move away from somewhere like Fosters and Partners to then set up a business?

George Guida: Um, for me it became very clear that I wanted to have my, have independence and let's say had a few projects starting, uh, at the time. So I really, I felt the urge to kind of take that step. Part of that taking of that step was for me, take doing another master's, and that brought me to Harvard, which is, um, it's the Graduate School of Design, which is pretty unique compared to European or British models where you can take classes [00:06:00] across all the schools.

So I took business classes at Sloan, at MIT. I worked a bit at the MIT media lab on, you know, generative city building, kind of a crazy project. And then I had my studios and I had my Python. It was, you could really mix and match things together.

Stephen Drew: Yeah.

George Guida: The real step for me into understanding the value, for example, of AI was.

Partly in my thesis days. And so my thesis was repositioning the value of language within the design process. And so that was the early days of text or image to image converting from a single text, which at the time was. A bit mind boggling, and you could convert text into an IM image trained on a trained dataset.

Um, there was no discussion, there was no hype, there was no dally, there was no, uh, none of these, um, image generator models. I was pretty much, very much [00:07:00] alone in, in proposing this. But I really, once you get into the code and once you start realizing, okay, these blobs, month after month are getting more and more refined, you could actually generate.

Images of use or proposed different pedagogical models, then things started getting really interesting.

Stephen Drew: Fair play, and so I want to talk about Figu in a second, but how did you get to the business that you've got today? Then you're talking about all this stuff that's there because there's the, wow, these are cool applications. But then in a business sense, when you're doing the project also you've gotta think like what kind of practice a practice is gonna be interested in?

How can I solve a real world problem? How can I make money out of it and all this stuff. When did, how did you get to an idea where you were like, this is the one I need to run with?

George Guida: Right. So, uh, a few, there's, I took a few different approaches. One was from the consultancy I gave to several Architecture [00:08:00] firms. Even being in this, in the industry day to day, seeing this AI sprint, there's so many companies, there's so much noise. Happening. There's UI is addressed in a terrible way and for me, UI and UX user experience is really critical to developing a product.

It was just very much lacking. And so our way of addressing the noise of all the tools and all the pain points that these companies were having and trying to figure out, okay, what is best to, what is the best model? Everything's changing month on month. How do I manage all my subscriptions? How do I control that?

All centralize that or simply mapping out the creative process from a p pedagogical perspective, how do you map out, you know, my starting point, I merged a different model. I generate images. I upscale or made video tracing that across different platforms. It's huge, you know, huge challenge.

Stephen Drew: Okay, [00:09:00] cool.

George Guida: we piece, we pieced it all together into one.

Stephen Drew: Yeah, I'll bring it up then because, so probably helps if we visualize it as well.

Exploring X Figu's Features and Capabilities
---

Stephen Drew: 'cause you've got a nice bit of, um, you've done a good job here on, on, on websites I think if it helps to show what it actually does. And that's half the battle. But what am I looking at here then, George?

Is this the actual software that you've built now? So this is for Gura, which is for our audio listeners, www dot x Figu, F-I-G-U-R-A. Ai. Tell me about it then. So it says, from concept to infinite canvas. Now I apologize, George, if I'm scrolling around, it's a bit like when your student, it takes over the PowerPoint and you're like, stop.

But I'll go back up to the top. Do you wanna walk me in? Use through it? It seems really cool.

George Guida: Yeah, of course. So we, a way of thinking of us is considering us a Miro for architects. So you have [00:10:00] that infinite canvas, you have full collaboration, and it's all powered by ai. So we frame the. The platform very much in those concept design, those ideation stages, we can accelerate the process but really keep the Architect in that driving position.

And so things you can do, we've kind of broken down this project. You can trace your history. So you can start from an input image. It can be a rhino screenshot, it can be a sketch that you doodled up, and then you can get a high res. Image or you can ideate and cross. You can combine a style. Someone in your office is really great at sketching, and you wanna make a presentation with that style.

You can cross these together, make videos, make 3D models. We piece this all into one workflow.

Stephen Drew: that's very cool. 'cause sorry, I'm clicking around here. But that's interesting. So it's text to image. Then we've got image to image. And this is the [00:11:00] one that I think is really interesting as well. 'cause you've got image to 3D 'cause that's the thing, that's the one big critique that some people have.

George with chat, um, not chat, PT Midjourney per se. It's like, whoa, that's a beautiful image, but how is it built? You know, you get in one side of it almost like a film production, isn't it? I am guessing what you're doing is you are extrapolating these images into 3D and then like you said, like mirror style.

You're mapping the project. Is that right then, or have I butchered what your product does?

George Guida: Yeah, this is one of the many workflows. So actually we host 25 of the best AI models all in one place. Of those, only three actually are these 3D models at the end. So we have text to text. Text image, Google OpenAI, uh, flux, all of the best models video. And then the final one is 3D. So for 3D we have a few workflows in itself, so it's image to a 3D model.

This goes through [00:12:00] what's called a multi-view process or multiple images. So front, right, left back, and then you can condition that output. Then the third option is actually a speckle integration. So this permits you to connect your rhino, your Revit, your SketchUp, or any platform with, you know, a few clicks.

You can connect it into the platform, choose your view, and then make a view from your existing model.

Stephen Drew: Right.

George Guida: a few processes out there. 3D is definitely coming up this year. It's very much been the year of video. It's been very much the year of AI agents, but I definitely think in the next few months it will be much more, we'll see a lot more development in 3D.

Generative AI applied to

Stephen Drew: interesting.

George Guida: Yeah.

Stephen Drew: Right, because yeah, there's lots of these different things out.

Challenges and Future of AI in Architecture
---

Stephen Drew: I mean, what, what's really interesting, so you've designed this for the Architect in mind basically, [00:13:00] isn't it? Because in theory, I mean, I have access, I can get access to the API, I can get access to Gemini, I can get access to all these things.

But you've got all these technologies here, there everywhere. So. What you're doing is trying to bring everything together under one roof, George, for the Architect in mind, right, because you can, I mean, my experience already is, like you said, you can go off on a tangent on just videos

George Guida: Right.

Stephen Drew: generating that.

So with that, where's the challenge then, George? Do you look at different language models and some are better for certain things, some are better for others. Is that what the constant battle is right now and getting all the practicalities as well?

George Guida: Yeah, it's, uh, it's a challenge, but also the beauty because we're in a moment where all the models, it's not converging to the one model. There. We'll definitely see in the next few months, probably the convergence into more conversational models, which I kind of, I critique because it's because of its [00:14:00] mono modality in its in one way.

Um, but there's definitely. What we offer is the combination of all of these. So each model is good or bad for certain use cases. If you wanna relight an image, you might use what's called flux context max. If you want to zoom in or zoom out, change the view. You might choose another model, and it's always important to realize.

Every model, every company has trained their data sets with different inputs. There's no such, such thing as an objective data set. And so that underlying subjectivity is A, something you can influence, but B is something they bring forward and you can leverage or you know, be challenged by in your model selection.

Stephen Drew: Mm. Got it. And do So is half the battle when you've been building this software then been learning all this stuff, and then do you find generally a, are all these different. Are [00:15:00] all these different? So you got what? I'm trying to think what's out there. We've got chat, GPT, you've got Crock, you've got all this stuff as well.

You got Claude. So I mean, just in my business, I, I use Opus to, you'll laugh at this, right? Um, to output h tml coding, whack it on my website. I made this like, uh. Notice period. Calculator George, and then a, the amount of traffic I got on Google was loads. So it was really good for that. But then the same task I brought into chat GPTA few months ago, and it was rubbish.

You couldn't do it. It couldn't make it. So, uh, two part question. Are there certain ones that you, as an Architect, you think right. Google 'cause it can have a million, whatever. That one's really good. This one's really good. That one really good. How much of your time is working out? All this stuff. And then the second part is, oh, do you have to be fluid because they're constantly evolving and getting better and better all the time.

George Guida: Yeah. Um, as we're seeing every few weeks, if not, you know, few [00:16:00] months, there's like hundreds or there's so many new models coming out all the time. And so I would say actually for text only, we only offer the AI models. But as you've seen, you know, uh, in the coding industry. People tend to flock to Claude and, and so you know, chat GBT performs better or worse in other categories for applied to Architecture.

I think a really great model. It's also open source. There's a dev developer version. It's called the Flux Context, uh, dev model, and that's by a company called Black Forest Labs. They're kind of the next one, uh, sec. Another company to stability, AI who open source, let's say the famous stable diffusion model.

And so by open sourcing it, you really give, you know everyone the access to these tools. [00:17:00] Not only. Forcing people to, okay, let's just use the mid journey. Let's try, you know, let's get under the hood. Let's get into the code, let's train our own models and really give everyone those chances.

Stephen Drew: Yeah. Because in one way the beauty of what you're doing and I, I like the idea of open source 'cause there's so much that could happen. 'cause if one thing changes it or, and you built all these different procedures on it and then they change their rules, then that's annoying. 'cause then you've gotta move some to something else George, haven't you?

Which could be not as potentially as good. Right. So where do you think, okay, I, where do you think it's all going with these right now? So at the time. Recording this. So Mr. Musk is releasing GR four, right? And then, oh yeah. And actually let's talk about that 'cause he's not having a fun time. 'cause on X is, it's calling itself some unfortunate last names, but where do you see it going?

Because I think that Gemini is quite, quite a good product there. But you mentioned that OpenAI [00:18:00] use a lot of it and it, it was the first to market and all my staff use. Chat, CPT as well. Do you think sometimes the names of things certain gravitate towards some people using it or what? What's your thoughts?

George Guida: Well, what's evident is that being first to market is almost everything, especially in the AI startup race. Gemini was slow being, being, entering the market for some reason. Even though they launched, let's say they had the foundational papers were, you know, authored by them. Uh, the same race applies to us.

You know, we try to like be as fast as possible to deliver the latest models and kind of try to be that first to market applied to Architecture. It is unforgiving. Um, Gemini for at least for image generation, is really great. There's the, um, a flash and a flash model it's called, and it's very much conversational and you can have a [00:19:00] long conversation and say, add some people in the foreground.

Change this to a snowy scene, and it keeps the coherence of that image. The, let's say, one of the pitfalls of their model, for example, is that they prioritize inference speed. So these, the images can load really fast, but the quality sometimes is not as high, for example, as the open ai, the chat GBT equivalent model.

So it's, it's always, it's always that trade off. But, um, the point is that. You know, you have to keep that flexibility. You have to kind of stay on your feet and, and be aware of what's next.

Stephen Drew: Yeah, I, because I can see the point of your platform as well, because this is such an in-depth thing. You have to know all these different things, and I think bringing it in seamless makes complete sense, especially on a project thing. Maybe we can talk about you setting up a business as at the moment, because I think personally as a business owner, it's quite inspirational.

Entrepreneurial Journey and Lessons Learned
---

Stephen Drew: So you were [00:20:00] getting going. Was it initially you, George, at the start on your own? Burning the midnight oil, you've got an idea, you wanna get it going. You've got a concept. How are those early days? 'cause now you've got five staff. You're working remotely with all these cloud people around the world. Give me the jour, the entrepreneurial journey with the AI part of it.

George Guida: So I started, uh, it was myself and a, a co-founder at the time. And, uh, you know, we were together burning that midnight oil and developing the vision. It like creating that vision is, is quite a challenge today. I. especially finding your market in all of this AI noise. Uh, so we, we spent a few months drawing, drawing it out, and then gradually started onboarding and launched our MVP.

In December we launched the product. We about three, almost three and a half or so months ago, and it's, it's been kind of steady from there.[00:21:00]

Stephen Drew: Steady. Full on. Yeah. And I imagine there's quick wins as in the business sense. What lessons have you learned? So for me, I mean, I've learned a lot of lessons of being, so you need to be frugal on some things. You need to jump on some other things. I've made a lot of mistakes in my business, George, on hiring, which is quite ironic.

And recruitment or, or maybe not jumping on opportunities, not getting people on board, maybe missing certain things. And also, when I was at the start of the Architecture, Social, I was trying to do everything, George, you know, we were doing this, we were doing that. And then it wasn't quite streamlined and people were like, what the heck are you doing?

So there's lots of mistakes that I've made and I've learned from Were there some mistakes that you made at the start, which you thought, ah, you know, and then suddenly you go into it and you realize like, that ain't working.

George Guida: I think, yeah, it's, I mean, it's a, it's a rollercoaster of a journey. Uh, super exciting, but also the highs comes with lows. Uh, I would say. It's a common challenge that many founders, especially [00:22:00] developing software has is when to when to get it out. Like how precious should you be about delivering something a little less unfinished?

It's never gonna be that, like that final, what you have in your mind. And so we ended up launching it. It wasn't perfect. There's the usual bugs in the software, but it was, let's say it took a big step. To just get it out there in it, in it, in its imperfect state.

Stephen Drew: Yeah, well it's never finished, especially with stuff like ai. So I'm kind of like, I'm probably the worst person to ask or, or talk to about that, George. 'cause I do on websites, for example, you, the concept is you have a live side and you have a staging side. But I use the ST and it like just to debug stuff, which is just totally the rogue way.

But I'm, my view in business is you're never fully ready. You just get it out there. So I agree with you. However, also. In the game industry. 'cause I mean, I don't have much time now, but when I did do I like to play video [00:23:00] games, the amount of hate that a game will get if it's released and not quite right.

So you've got, yeah, you have gotta get that balance. I think it needs to be working. We have a few kinks and you say to people, listen, you are on the beta, you know, let me know and I'll fix it. Maybe that's part of part of it. But yeah, no, it seems really, really cool. I applaud anyone doing a business as well.

So you do have staff around the world?

Challenges of Remote vs. In-Person Work
---

Stephen Drew: I have staff in person. What's the challenges on having people remote versus in person? 'cause I'm guessing one of the massive pluses is that you can hire the best people around the world, but then how do you feel about it? You think remote's the way they go when you do a digital platform?

Or would you rather get people in the office Architecture style?

George Guida: I think a hybrid scenario would, would probably be my ideal, is kind of having that flexibility, having people around on certain days. There's always a challenge, especially growing an online office culture.

Stephen Drew: Yeah.

George Guida: how do you cultivate that? Could you [00:24:00] have your like Friday drinks or your Friday pizza hour when your time zones are completely off?

So it's, you know, morning coffee and uh, uh, maybe a lunch and then a dinner for the next two staff.

Stephen Drew: Yeah, that, that was my challenge. At first. We were fully. Remote, but around London. And then there was a point where you're just like, why are we doing a meeting? Let's just go in person and have a beer and talk about it. So yeah. Fair enough. I mean, I'm excited where the business is going and I'll bring it up in a, in a bit, but while you're here, so this makes sense then for businesses who want, who are doing projects, like you said, the mirror of, of these staff, however, you've got a lot of.

Stuff in your background, George, and, and you also do teaching, and that's how we met. Now that's a mutual contact. So for people, maybe not in a position to use your product, but maybe like a student, right.

Advice for Students on Academic Projects
---

Stephen Drew: So do you have any advice on things that people should look at for their own academic projects, like good [00:25:00] tools that people should perhaps consider bringing into their arsenal of tools that you would use if you were a student now?

George Guida: Yeah, it's a great question. There's, I mean, I feel like all students now should have some exposure to ai, even as critical as some might be, and it's completely natural. Uh, even at a level of AI ethics, understanding bias, understanding data sets or agency as designers, what is, how is this, what's happening to the creative process from a tool perspective?

Um, I'm obviously biased, but are the point of xFi Good is to package all of the best in one place. And just as an example, it's like the image generation you like flux produces. Ideogram produces really high quality images. For upscales, we have magnifi, and that's, let's say, the industry standard for video.

We have gem like Google's Gemini, VO two and three model, [00:26:00] which is image to video with audio, which is kind of the latest and greatest. And then for 3D we have trip rodent, uh, for example, which are, let's say the best, and this is just framed within the generative space.

The Role of AI in Architecture Education
---

George Guida: For the more advanced users, you know, diving a little bit into the code, which is something I'll be teaching this semester.

I, I try to allocate no more than one class a term. I'll be at UPenn. Teaching a class on AI and Architecture. I teach no code background Architecture, students, how to use these tools to build tools, tangible or you know, web-based platforms, which are essentially design tools for architects. And so that requires getting a bit into the code, learning how to debug with Claude or chat GBT and then build, actually build things.

Stephen Drew: Yeah. Fair play. I was just looking on Figu. You've got a free [00:27:00] start pack so people can play around with it. Now I could actually see how it could work with some students as well. Fair play. As long as they don't go off on a tangent and burn all those credits. As so you gotta get the, get the balance, but equally experiment.

George Guida: models. It's, it's the, the Google VO two and three models, which are their price pricing model is crazy expenses. So as long as you stand away for those, you can use the academic, it's 30%, 30 to 40% discount for all all students, all educators.

Stephen Drew: There you go. Do you need to get the uni to pay the go on, George, we, we should get that. That would be really cool. I really like it on the subject though, of academia, right. So I think personally, my bias is like, I, I would've loved AI when I'm there. Not to rip everything off, but it's just, I can see it being an amazing tool and just speeding up, doing the mundane, getting things going.

Having said that though, do you know the Architecture that me and you [00:28:00] studied? You mentioned you were at the a i, my part went on part two. Right. I remember you gotta do all that coursework. You go write that big bloody essay and stuff. I can see AI working really well with that.

Future of AI in Architecture
---

Stephen Drew: Do you think the old way of learning is gonna be changed by ai as in.

Can we teach the same way as before, or do you, does, does courses have to evolve with the tech as well?

George Guida: So the World Economic Forum just released the future of Work report for 2025, and some of the key. Skills, which they're projecting that are needed in, let's say the next generation of workers are very much flexibility, adaptability, and resilience. And so we need to teach students how to have these soft skills or how to have these skills, but also to maintain that critical thinking and not just rely, or let's say, depend on the chat GBTs or depend on these tools now.

I just, [00:29:00] um, I taught a course with the Department of Education in the US and I taught, um, a few thousand high school and middle school teachers in New York. And part of what I taught was, you know, future classroom models. And one interesting one, which I brought forward was this idea of bringing back this flip classroom approach.

And what a flipped classroom essentially means is that how could we prioritize in-person time to really give the time for critical thinking assignments, teamwork, collaboration, soft skills training, you know, working together, working critically, working with our critical hats on, and then everything at home.

You know, the lectures or you could even have AI help you, like you give an AI chat bot or an AI agent to help you augment lectures or you know, the passive stuff, but flipping it in some way.

Stephen Drew: Yeah. I think the, when people talk about [00:30:00] ai, they jump to the idea of AI doing the design, right? Whereas. It's in my head. It's just like a very, very good apprentice. If you, if you coordinate in the right way, you can get the mundane stuff done. Do you think. That's echoes what you say, but on that subject, like this, this be devil's advocate now, right?

I don't see AI replacing an architect's job. I just don't see it personally. However, Microsoft just did a massive layoff of people, right? Because oh, was at Google, one of them did. Either way, anecdotally, they've let those people go. And the other thing is, I remember 10 years ago, it was like the best skill, Steve, if you can, is to learn how to program, right?

And now I'm like, I don't need to do that. I got my Claude Opus. I'm using, I'm vibing. I'm making this clunky piece of software. Nowhere near as good as the Gura, but I'm playing around. So what's your sentiment and all that I just said.

George Guida: Yeah, it's um. A lot is [00:31:00] certainly changing and there's, there will definitely be a, a big reliance of for these tools and it's important to. To teach, let's say the traditional, traditional task, but also teach the next generation to, to actually work with these replacement is, um, is definitely a big topic.

It will happen across all industries in a Architecture. I don't think also the architects will be replaced, but I like to say that architects who do the early adopters, the architects who do engage with AI, will be much more competitive or much safer in the, in the short and medium. Let's say long term.

Stephen Drew: And also on that point, I agree using it, it's like a super power up. But then what, uh, my view is as well, if someone has a personal connection with a client or development, then they're gonna have that human contact. So I see this, well, the role of an Architect in client facing being their communicating with the client.

That's more and more and more powerful, [00:32:00] whereas. Maybe if someone's back of house doing all this stuff, that's the role that could be a little bit more in jeopardy. That's how I feel about recruitment.

The Changing Landscape of the Architecture Profession
---

Stephen Drew: That's how I feel as a business owner, that the lower hanging fruit, I would put my business hat on. Any data entry job that someone's been doing in the Architecture, Social, you are redundant, literally, and you're gonna beis.

But if someone's got that connection with a client, why would I ever replace them? You know? I mean, what'd you think about that sentiment?

George Guida: Yeah, it's, it's important to position AI very carefully, and so there's clearly like client facing, public facing, uh, these like interpersonal requirements, soft skills that we need. You can't replace those. But what what's even more important for Architecture is simply positioning, you know, our role to have, uh, within this idea of our standard of care.

Standard of care. Like we have a duty for the public, for the public safety, welfare, um, and safe and [00:33:00] pub safety and welfare. And so we, we need to like really be careful of, um, what it means if an AI is replacing some of that process. What is. What does it mean for an AI to do code compliance on your drawing set?

What does it mean for an, like an AI system to automate, go from as like a concept design, Revit model to a construction design Revit model? Like where does that position us also as, as curators or as people signing off and keeping that liability?

Stephen Drew: Yeah, because on that note, I find it, especially after grandfather, like look at the amount of policies that architects in the UK have to do, even like fire regulation not to be sealed. And now you've got the PRI Principal design act and all this stuff. It's quite tempting though, George need to get bit of AI in to know all the rules.

'cause I do think that the burden on an Architect is so high at the moment. You've gotta build the building, you've gotta get qualified, you've gotta learn all this stuff. You've got all the critical [00:34:00] thinking. So maybe it would be nice if AI could do a little bit of that

George Guida: Uh, which is, it would be, it would be nice, and maybe there's a future for that in which that could happen, but not as a way to replace us. It should be always, you know, have it as a maybe, potentially as a collaborator. So it, it's, you know, you have to also remember that AI is black boxed. At least for, for example, generative ai and to a certain extent, machine learning.

So finding compared to the traditional like software process where you can trace the branching of decisions, like you know exactly how something is derived. With ai, the risk of hallucinations or the risk of not knowing how that decision is reached, it removes its value of like this idea of giving it accountability.

So imagine the accountability of a scenario like Grenfell, which is let's say, absolute worst than let's hopefully never get to that stage. [00:35:00] Um, but would you then want like an AI system to evaluate, you know, the compliance or the standards of your drawing set? With within its current Architecture. You know, if, if you can't trace those results results, then you still, you'll definitely still need an Architect.

I'll tell you that Anthropic is one of the few companies that is trying to address this idea of explainability, of essentially finding how these decisions are being made. But today we're, you know, we're still not there.

Stephen Drew: Yeah. And for, have you followed? Um, I can never say the name of it. Forgive me. My Welsh accent. Philanthropic, right. You know the, I loved it. Do you see they did an exercise with their AI and they said, right, you're gonna run a vending machine business in the. In the office. Apparently it started off really well, and then someone tricked it or asked it to order some black silver cubes, and then it did that, lost [00:36:00] all the profit.

And there was one bit, it made me laugh because George, it was hallucinating and someone was like, yeah, where are you? And he was like, yeah, I'm just in front of the vending machines wearing the suit and tie. So. It's not there, but again, it, it, it is really, really interesting. I was gonna ask you on that, so where we are now at the time of recording, if anyone's watching this, so this is like mid 2025, right?

So in terms of thinking about two, there's two parts of it. The start, first of all with, 'cause we talk about AI now, and then I want talk about the future of Architecture. But your sentiments on it. Are you feeling.

Optimism and Concerns for AI in Architecture
---

Stephen Drew: Optimistic towards where AI and architecture's going at the moment and the future. Then George?

George Guida: I am optimistic, but I definitely think that, um, let's say the, the RIBA a RB, NCARB, A I A in America, all need to play a, a, a stronger stance in essentially helping [00:37:00] define where the profession. Might head or what the values that we bring as a profession, especially in light of ai. So currently I am in the Reba's expert advisory group, and these are discussions we're having every meeting of how, you know, how forward, how much of a stance should RIBA have.

For example, in, in saying, you know, what, whether the next five years or whatnot, versus keeping a very broad, you know, approach. And we just launched about. Two weeks ago, the 2025 RIBA survey on ai, and you can see the adoption rates from last year are almost at 20% increase in its use in practice. And so what does that mean for the next five years?

What does that mean for the next 10 years? If we frame what the value of Architecture is today, if you look at what, what's called the futures cone, we need to ensure that those probable futures put us as architects in a good position for clients to value us. Give us, [00:38:00] let's say the appropriate compensation or the appropriate values that we need to thrive as a profession.

Stephen Drew: Yeah, fair enough. Maybe one more broader, 'cause I really like your perspective, so if less, uh. Pinpointed on ai, but Architecture as the industry. 'cause, so I studied Architecture in 2006 has been such a change there and a lot of frustrations with the rising costs in, in living life and then Architecture, salaries kind of going up, but with the constraints of businesses and all this things.

So there's a lot of frustrations I think in Architecture, but then there's a lot of exciting stuff as well. How do you feel in the wider, um, picture George, about Architecture, the state of the industry at the moment?

George Guida: Um, there's, uh. I'm generally i'll, I'll stick with my optimistic stance, but I will recognize that there is a lot of fragmentation, especially from a software perspective. Um, what AI and what the profession I, I like to think needs to embrace [00:39:00] is, um, this potential expansion of our scope. Now it's kind of, it's already been expanding.

There's its edges are very, are quite murky. You could argue. With the rise of ai, I like to, I like to think that we can offer more and we can diversify our scopes, especially since projects inevitably will have to move at a faster rate.

Stephen Drew: Yeah.

George Guida: so what that means, for example, is, you know. As funny as the term is what?

How could we apply vibe coding to Architecture? Could we rethink practice as a new startup? You know, how could we encourage a startup culture within any size practice? And so that's something I teach my students. How can you vibe code the next tool or the next software? How could you rethink Architecture as a profession into a more entrepreneurial mindset and actually put things out?

If you look at today's state of where, you know, Architecture software [00:40:00] is at least the ai, a EC software, we need to make sure us as architects are in those driving seats rather than external, uh, forces, um, you know, jump in.

Stephen Drew: Yeah. Well I think that makes sense 'cause a lot of frustration in Architecture is that we've lost certain things from being the, what was it called? The, the big arc, basically, you know, the way now on the design team you've got many things, whereas the before you'd have the master Architect, so a. One of the conversations I was having with someone else was the, the, the principal designer, actor in the uk, the role of principal, designer being part of the Architect.

But I think there's a massive opportunity, like you say, for the ai, the tech, the vision, and doing all that stuff. So I personally find it quite exciting as well at this point. George, I normally say to the guests, if you've got one or two questions for me now, we've just met. Whichever, once or twice before I do the recruitment piece, I've got the online platform.

I've got all this stuff. I've mentioned that I use AI in [00:41:00] my business. I got five staff in the uk. What questions would you like to ask me live? Because we haven't planned this and I want you to, now you know me a bit like what would you like to know my thoughts on?

Recruitment and AI: A Business Perspective
---

George Guida: How do you think in the next five to 10 years, um, that your practice, like what, what will your practice look like in five to 10 years in light of ai?

Stephen Drew: So with me, because in terms of recruitment, so a lot of it is CVS processes, candidate screening. There's a massive opportunity on candidate screening, but exactly like you said in Architecture, I can't just trust him. Let's pretend I'm recruiting for you when I go up. Josh might be looking for this and that you just know.

Without that information, that training, it's gonna go with some wild suggestions on, maybe not pick up on certain things. So I think there's a, I don't exactly like Architecture. I don't see it replacing the core role, but where I think it's really good is just doing the absolute data entry, the bulk of stuff.

Processing information, really [00:42:00] trying to, a good example, so they use passing technology in cvs so people get freaked out about this 'cause they think, oh my gosh, I gotta get my CV. I've gotta make it right so that it's passed. Okay. Generally, most. CVS can be passed with anything. So I've got my own custom rules on what to look for, and then I use like, not an agent, but something like that, AI already to filter out keywords, put them in certain things.

If you use the word Revit, then put it in the short list. If it does that, flag it up to a consultant. So that's only gonna get, um, more and more evolves. Um, another one, George, not directly to my business, but in my business I get now bloody, um, excuse my language. It's all right. It's my podcast so I can swear a bit.

But like you get this, I'm getting automated sales calls. Like, hi, this is a robot. So that's, um. Completely annoying because no one's gonna want to do that. So I think they've put the AI in the wrong bit. That's the front facing bit. That's the bit where people are always gonna have the jobs in the Architecture, Social, [00:43:00] whereas really down the line, I think as much of the data entry, making folders, logging things, keeping up to complying with GDPR would be a good one.

You know, put rules in place doing all this stuff. So. I see it as a massive opportunity and I see it will structure the what I, who I hire in the business as people that have all relationships. And so doing that aspects, that's what I predict will happen. Yeah.

George Guida: is funny. I I just went through the same process we had. I had 10 cvs and put them all into chat, GVT and to Claude and compared the two, made a table based on X, Y, Z criteria. I did my own assessment, my ranking, let's say myself. It looked very different from the chat beauty list. So it's, it's interesting.

It's always this like, you know, balance. Like some cvs were like pinpointing the exact right words and [00:44:00] they were like spot on. But then when you dive deeper into like the portfolio or the written statements, then you'd get a slightly different picture.

Stephen Drew: That's right. It, um, you've hit the nail on the head. Uh, it's just understanding an Architecture portfolio and the new nuances is just so, so important. So you mentioned you were at Fosters years ago, so you'd have like large international projects. There are certain companies where someone from Fosters is more likely gonna go to or the other way, like certain companies are more likely gonna appreciate.

The skills of someone from Fosters, whereas I don't think the AI would get that yet. It would be like, oh yeah, there's someone's from Fosters. They're good. They can go anywhere. I'm not saying that's not wrong. That is true. I've seen someone from Fosters go and build a real estate development company and at first I was like, whoa, that's so, so different.

So everything can happen. But that initial nuances of like. Okay. What companies like hiring from each other? I don't think it's there yet, but George, I do think you can train it to do some [00:45:00] stuff. And then maybe what you do over time is you'd have it like, okay, these are my five suggestions. What do you think?

And then they would go, yeah. And my logic for that is I know it's gonna have to do it, but the amount I always find it. Have you ever heard behind the scenes on like the, uh, the dating apps like Tinder?

George Guida: I don't think so.

Stephen Drew: They're right. It's fascinating. And this is pre, they got algorithms. So then based on the amount of attention that people get, they go up and down different brackets.

There's all this wild stuff that goes on there. So this app will go, we think you're eight outta 10 handsome based upon what other people are doing. So you imagine how that's gonna speed up. And the where it gets really sketchy though, listen to what I said, right? The, the apps behind the scenes say how.

Handsome, you are right. Where it gets sketchy area for recruitment is you've got, um, like you got if certain things in recruitment, like ageist, this and [00:46:00] that, right? So where wherein lies the balance are the right choices. Are they right? Are they ethical? Are they inclusive? That's gonna be a very hard minefield to do.

And recruiting for the private sector is very different than the public. In the public. You gotta get certain demographic, different backgrounds, you gotta really go through it, which is great, but it's so different than private. So I don't know. We'll have to see bit of a tangent, but

George Guida: like a startup in itself,

Stephen Drew: yeah, well it,

George Guida: a recruitment service, AI powered.

Stephen Drew: I dunno if it could be me, but, um, George, if you fancy Figu number two, then there's an idea for you.

Conclusion and Contact Information
---

Stephen Drew: Maybe On that note, before we go, I would like to remind everyone where they can find you, George. So tell me how people can get in contact with you while I bring up all your links and stuff again.

George Guida: You can contact us at info at xFi AI for any questions. We've now partnered with several large and medium sized [00:47:00] Architecture firms. We have these educational pacs and we partnered in the us, the uk, and in Istanbul at the moment. So happy to extend that. And, uh, you can find us at www dot x ai or as seen on the screen.

Stephen Drew: Very good. Well, I'm excited to see where it goes, and I think that there needs to be more people like yourself, trailblazing in the space. So thank you so much for joining me, George. I really appreciate it. Stay on the stage. One second. So before we go on, I say thank you to the audience. Did you find this useful?

What do you think about George's, uh, product and direction? Do you think it's gonna help you out with. Projects, do you think we shouldn't be doing that? And we should be going back to pen and paper? That's a bit overkill. But either way, drop George a comment and connect with him. And if you are an Architecture business, then do have a look at these projects as well, because the amount of hours that it takes to learn all the [00:48:00] ai, all the language models, all this stuff, George is doing those hundreds and hundreds of hours that you can just plunk it into a project.

So I do definitely see the value in it. Do check out X figu. And on that note, I'm gonna end the show now. Thank you so much, everyone. Have a good day wherever you are and see you soon. Take care. Bye-bye everyone. Bye-bye.

George Guida: [00:49:00] everyone.