We’ve all heard it before – “AI is going to change everything.” If you’re an architect, chances are you’ve been skeptical. The hype has been loud, but the practical impact? Less obvious. At least, that used to be the case. Recently, something has shifted. The tools are sharper, the results more useful, and the people who once rolled their eyes are now quietly making room for AI in their workflows. The conversation isn’t about whether architects will use AI – it’s about how they already are. The question now is less about resistance and more about readiness. Welcome to episode 182: How AI is Changing Architecture.
[Note: If you are reading this via email, click here to access the on-site audio player] Podcast: Embed Subscribe: Apple Podcasts | Spotify | Android | iHeartRadio | TuneIn
The Changing Role of AI jump to 11:19
Twelve months ago, AI in architecture was still largely considered an experiment – something to poke at in your downtime, more curiosity than commitment. Today, the conversation feels entirely different. AI has moved from the margins into the center of architectural discourse, with a clear shift in how we think about productivity, communication, and creativity in the profession. Tools once seen as novelties are now embedded into daily workflows. While we’re still far from AI replacing architects (or anything close to it), we are already in a moment where those not engaging with these tools are beginning to fall behind. This is not to suggest the industry has fully embraced AI – but the momentum is undeniable. The conversation has moved past whether it should be used. Now it’s about how, when, and where.
The biggest evolution hasn’t only been in awareness – it has been in the tools themselves. Platforms like ChatGPT have progressed from GPT-4 to GPT-5o, delivering faster processing, stronger contextual understanding, and multimodal capabilities that allow users to blend text, code, and visual inputs. The image-generation world has taken a similar leap. Midjourney’s latest versions provide far greater control over materiality, lighting, and visual refinement than ever before. Iteration speed continues to improve. Prompt fluency is on the rise. The average user – someone without a background in machine learning – can now produce content that previously required a team of visualizers or writers. This isn’t science fiction anymore. These are tools architects are actually using.
One of the more interesting shifts is how these tools are being used not only to draw and design, but also to manage, research, and write. In my own work, ChatGPT is part of my daily process – sometimes for researching zoning, sometimes for refining narratives, and sometimes to test the tone of a message walking the line between marketing and technical accuracy. Midjourney has become a reliable resource for conceptual renderings and mood boards, translating loose ideas into compelling visual studies. Even with daily use, I know I’m only scratching the surface. Dozens of tools remain outside my orbit – either because they don’t apply to my current responsibilities or I haven’t had the time (or budget) to experiment. That’s the reality for most architects: we’re curious, but we’re also managing deadlines.
This next phase of AI feels more serious. The maturity of the platforms is bringing the value proposition into focus. Find the right tool for the task, and you can expect to work faster, more efficiently, and with sharper outcomes. No one needs to master everything. Instead, success is about knowing which tools are worth exploring. Writers might rely on ChatGPT. Designers might turn to Midjourney or Veras. Development teams may benefit from TestFit or Spacemaker. Expertise in every platform isn’t required – but an understanding of their capabilities is becoming essential. Like it or not, AI is no longer a novelty. It’s a skill set that will define how architects think, communicate, and compete moving forward.
Here is a list of AI platforms that I have been exploring – none of these links are affiliates and they are only placed here for your convenience.
ChatGPT – Natural language processing and content generation.
Used for writing narratives, zoning research, email drafts, project descriptions, and general admin support. Essential for BD, PMs, and marketing roles.
Claude (Anthropic) – Long-context summarization and research clarity.
Ideal for summarizing long codes, meeting transcripts, or planning documents. Strong complement to ChatGPT for detailed reference materials.
Midjourney – High-quality image generation via prompt-based inputs.
Used in conceptual design and early mood studies. Helps translate vague ideas into visual inspiration for client presentations.
Veras (EvolveLAB) – Revit-based rendering plugin using AI for stylization.
Allows real-time, stylized visual outputs directly from BIM models. Supports iterative ideation and presentation development.
TestFit – Generative feasibility studies for site planning.
Optimizes massing layouts, unit yield, and parking logic for multifamily and mixed-use. Valuable for developers and feasibility-phase architects.
Hypar – Rule-based, parametric building generation.
Automates layout logic, particularly useful for housing and office typologies. Supports rapid prototyping within defined constraints.
Spacemaker (Autodesk) – Early-stage site design using AI for performance feedback.
Used in urban and site planning to optimize daylight, wind flow, and view corridors. Facilitates data-driven design decisions.
Notion AI – Document summarization and admin task automation.
Streamlines note-taking, meeting prep, and task lists. Useful for internal communication and team organization.
Perplexity AI – Research-focused, citation-based answer engine.
Ideal for academic or code-heavy queries. Great for architects who need fast, trustworthy research without deep-diving into PDFs.
These tools are not replacing architects – but they are quietly replacing tasks. From zoning research to first-round renderings, AI is expanding how much one person can do, and how fast they can do it. Choosing the right tool depends on your role: designers lean toward visual generation, managers favor communication tools, and marketing teams benefit from narrative support. The more you understand each platform’s niche, the more strategic your adoption will be.
AI in Schools jump to 28:15
Andrew: From my experience, academia is still undetermined on its approach to AI tools. For example, in my department, some professors in our department are really diving into AI tools, and some still don’t touch them at all. I’m somewhere in the middle, although I would prefer to be exploring it more, but the courses I teach don’t always allow for that exploration. In a many cases, I think I’m using it more than my students are. I try to get them to see the applications beyond just making flashy MidJourney images, but right now, most of them don’t see it as anything more than a way to make something “pretty.” And even then, they often struggle to prompt well enough to get what they actually need.
I’ve tried different approaches to bring it into the work. One semester I had them use ChatGPT to make building user profiles by asking questions like, “What’s the job of a museum administrator? What does a curator do? What does it take to run a café?” What is the daily routine of an “X” User? I hope this helps them to get some baseline program information and user consideration. Since we don’t always have real “clients”, I attempt to have the AI act as one. I’ve also encouraged them to use it for code-related questions. Sometimes they’ll come to me and ask, “What does the building code say about this?” and I’ll turn it back on them and ask, “Did you try ChatGPT yet? Let’s see what it says.” But if you don’t prompt with enough specificity, you don’t get the right section, number, or exact code language back. That’s part of the bigger problem: they’re not great at prompting for technical information. This is twofold; they lack prompting skills and they lack technical knowledge. So it is a big ask for them, but I hope it could push them to become better in both areas.
There are some studios are pushing the generative side harder. There’s one professor who had a class studying roadside gas stations, like hundreds of them, and feeding those into MidJourney or ChatGPT to reverse-engineer the prompts that might have made those images. From there, they’d pull out keywords like “awnings” or “canopies” that showed up hundreds of times. This was then a way to help define key elements of that typology. It’s clever, but still pretty surface-level in terms of design thinking. But for students, it is a significant connection and usage.
In my courses, I’ve used tools like Veras for rendering. When I first introduced it about two years ago, it would change the geometry of a project, which was frustrating. Now it’s much better as a tool and no longer changes project geometries. Again you have to know and understand prompting to get good results. But it can be used in Revit, SketchUp, and other 3D modeling software as a integrated plug-in. Also you can even upload a static image and get a high-quality render based on the uploaded sketch image or view. That’s the kind of thing I see as actually useful in the process.
What I keep coming back to is that AI should be a tool you learn to control, not something that produces the final work for you. The skill is in learning how to make it do what you need, not just reacting to the novelty of what it can make. But that’s where the biggest issue lies, my department and even university doesn’t have a consistent stance on AI use, especially for higher-level work like PhDs. We’ve had discussions, but it’s still murky. And honestly, I think there’s a bit of academic bias or hesitation about directly teaching these workflows, which makes it tricky to push them forward in a formal way. But right now, at least in Architecture schools, I see AI as mostly a tool to make pretty and interesting images and not much more. Again, mostly surface level usage, but I think it is making progress, but that is still solely dependent on the Professor.
Where AI Fits in Today’s Practice jump to 33:22
Today, the use of AI in architecture feels like dozens of experiments happening in parallel – many of them uncoordinated, and most of them flying under the radar. Some are being driven by individuals. Others are coming from the top down. In either case, the shift is happening. AI use in the field is no longer hypothetical – it is real. What we don’t yet have is a consistent framework for how it gets applied across firms. Instead, adoption is scattered. Some teams are leaning in aggressively. Others are standing by, unsure if this is just another passing trend. What’s clear is that AI’s presence is no longer optional.
Broadly speaking, AI is showing up in four key areas: research, ideation, project management, and operational workflows. Among these, research and ideation remain the lowest-hanging fruit. Platforms like ChatGPT and Claude are now used to interpret code, develop narratives, and simplify dense planning language. In design studios, Midjourney and Veras are generating conceptual renderings, mood studies, and early design imagery with impressive speed and control. Small firms often use AI as a way to punch above their weight – creating visualizations, writing copy, or running quick feasibility studies. Larger firms use it as an accelerator – something that helps senior staff work faster, iterate more often, and reduce repetitive effort in early-stage tasks.
Surprisingly, the biggest productivity gains today aren’t in modeling or documentation. They’re in task compression – writing meeting notes, generating outlines, summarizing reports, and cleaning up internal communication. Tools like Notion AI and Monograph are becoming invisible assistants, built into daily workflows. AI isn’t replacing anyone. It’s just giving people more time. Ten minutes replaces forty-five. Slide decks come together faster. Emails get sharper. It may not feel revolutionary, but for many, it’s changing the pace of business. Where AI hasn’t yet made significant headway is in the production of construction documents. Yes, there are promising tools – BIM-integrated plug-ins, parametric systems, and early-stage layout automation. However, most of these are still assistive, not autonomous. Most firms remain cautious, and for good reason. CD sets require precision and carry liability. We aren’t at a point where an AI tool can draw a section detail with the same intent and judgment as a licensed professional. That time may come, but not yet.
Another theme worth noting is that AI use today is still led by individuals rather than institutions. The curious designer testing prompts after hours. The project manager editing proposal language. The marketer experimenting with image generation or automated copy. Even at larger firms, few have formal policies or workflows around AI. This user-driven approach has benefits. It encourages experimentation and fast learning. However, it also introduces gaps in consistency, raises ethical questions, and opens the door to unvetted risks. Firms will need to catch up – and eventually, they will. AI today is useful, but not magical. It speeds up repetitive work, expands creative options, and helps clarify ideas in the early stages. What it doesn’t do well is nuance, judgment, or accountability. It still requires human review. It still hallucinates. It still stumbles on ambiguity. That doesn’t make it irrelevant, it makes it a tool we’re still learning to use. As architects continue to experiment, the shape of practice will shift – not overnight, but steadily.
The Future of AI in Architecture jump to 48:16
Looking ahead, the next twelve months may mark the point where AI becomes fully embedded in the tools we already use. Rather than jumping between apps, we’ll likely begin seeing AI integrated directly into platforms like Revit, Enscape, and others. Autodesk has already started laying the groundwork. Veras and Hypar provide early previews of what this could look like. The shift won’t be loud. It will be subtle. Smarter suggestions. Faster coordination. Real-time compliance checks. We won’t be automating the drawing set just yet – but we’ll be building it differently.
Five years from now, we may see more transformative applications. AI could start to inform decision-making, not just speed it up. Generative zoning tools may offer buildable massing scenarios based on code, access, and performance goals. Proposal platforms may pull from a firm’s past projects to draft custom responses. AI-assisted design engines could evolve concepts with only a few prompts and constraints. In this world, architecture teams may start to look different. One person with the right tools could take a project much further than before. That would shift hiring, training, and even how we define project phases.
Longer-term, the questions become more complex – and more consequential. Can AI design equitable cities? Can it balance density and daylight with social infrastructure and sustainability? Could it draw directly from environmental data to propose forms that respond in real time? Maybe. However, these possibilities introduce real risks. If AI proposes a design, who is responsible? Can you stamp it? Should you disclose it to clients? These aren’t hypothetical concerns. They’re legal and ethical ones, and the profession will need to address them soon.
The evolving concept of co-piloting will become central. Many architects, myself included, see AI as an extension of our process – not a substitute. It expands capacity. It increases speed. It gives ideas a shape more quickly. But it still needs someone to guide the outcome. Others argue that AI will soon outpace the need for that guidance. Maybe we become editors instead of authors. Maybe we become curators of generative systems. That distinction may change how we define authorship, responsibility, and value in the years ahead.
Looking forward, a few key trends are worth watching. Multimodal platforms that blend text, image, and 3D in one interface. Open-source model training that allows firms to teach AI their own language. Regulation and licensing standards that address authorship and liability. These developments are not just technical. They are foundational. They will shape how architects work – and how architecture is produced.
The next decade won’t be defined by AI itself, it will be shaped by how architects choose to respond. The firms that thrive won’t be the ones chasing the newest tools, they’ll be the ones that understand how to use them with care, clarity, and creativity. AI will not replace architects, but it will force us to rethink what being an architect really means.
Would You Rather jump to 54:53
I am starting to wonder if the manner in which all of these questions is answered comes down to how easily or not you are able to monetize your answer …
Would you rather be able to speak every language on earth, or be able to speak/communicate with all animals?
All you need to now is that at one point in this discussion, I say the word “worms” 6 times in a row, and it makes complete sense.
Ep 182: How AI is Changing Architecture
The past year has made one thing clear: the role of the architect is shifting, and AI is now part of the conversation. These tools are no longer optional experiments – they are becoming integral to how we think, communicate, and produce work. Architects who stay informed and open to change will be better positioned to lead. Mastery of every tool isn’t required. However, adaptability, awareness, and judgment are becoming the new standard. The work ahead calls for more curiosity, more reflection, and a willingness to evolve with the tools that are already shaping the future of practice.
Cheers,
