Memorandum
- From
- Derek Weston via Fast Company
- Date
- Filed
- Economy·5 min to read
- Re
Architecture Firms Build Their Own AI Tools as Data Stays Out of Reach for Big Labs
ReArchitecture Firms Build Their Own AI Tools as Data Stays Out of Reach for Big Labs
Frontier AI labs like Anthropic are not training models on architecture data because it is locked inside individual firms. Architecture firms are now hiring data scientists and building custom AI tools to stay competitive.
Architecture may be one of the hardest fields for artificial intelligence to master, and the reason is simple: the data needed to train a model is not sitting on the open internet. Unlike coding or writing, where vast amounts of text and code are freely available, the digital drawings, 3D models, and hand sketches that define an architecture project are stored in the servers and filing cabinets of individual firms. That has left the biggest AI labs, including Anthropic, largely on the sidelines, and pushed the responsibility onto the companies that actually hold the data: the architecture firms themselves.
Firms of all sizes are responding by building out their own AI capabilities. They are hiring data scientists and machine learning specialists, running internal idea competitions, and coding bespoke plugins and apps to automate highly specific tasks. Some are even developing their own narrow large language models to generate building forms and floor plans that reflect their signature style. The effort reflects a broad recognition across the industry that this is a sink-or-swim moment, as client expectations shift and the business becomes more competitive.
“Everyone has their little gold mine they’re sitting on, with all this data of past projects they’ve done,” says Faizan Zaidi, director of design technology at the architecture firm Spectorgroup. “But the question comes down to which firms are willing to build the tools on top of it.”
The AI awakening for many in the field came in 2022, when tools like Dall-E became widely available. Matthias Hollwich, cofounder of the 15-person New York-based firm HWKN, recalls that moment clearly. “Four years ago I got introduced to Dall-E, which I think was, for a lot of us, an awakening,” he says. “I decided to throw the whole office into an AI exploration for three months.” Since then, the number of AI tools has exploded, and HWKN now regularly uses them across its office, residential, and hospitality projects.
The firm’s toolkit includes NYC Zoning AI for zoning analysis and building massing, Midjourney and other tools for conceptual designs and renderings, TestFit and Forma for site plans, Maket for floor plans, and PermitProof and UpCodes for building code compliance. Hollwich says the tools have fundamentally changed how the firm approaches design, and even what it produces. “Currently, most architects are trying to use AI to optimize the process, or maybe intensify some of their designs,” he says. “What we have done after these experiments is almost take a step back and say, no, this is actually bigger than just using these tools to come up with a better way to design the buildings. There’s a new ideology that is emerging.”
That ideology includes using AI for deeper analysis during the proposal and conceptual design phase. For a recent office building proposal in Amsterdam, HWKN used AI to analyze who the future tenants might be, based on other companies and industries in the region. The tools helped the team understand where those companies are located, what types of buildings they occupy, and detailed information about their spaces, including floor plans, ceiling heights, and amenities. “We created this whole matrix, and we basically said, ‘If you design this building according to this information, you’re going to be ahead of the game because every one of these companies you would like to attract will see themselves in it,’” Hollwich says. HWKN won the project and now uses AI for other lead searches and analyses. It is also building a custom request-for-proposal response tool using Claude.
Spectorgroup, where Zaidi works, has been taking a similar path, building custom AI plugins and tools to work alongside existing design programs. One example is a tallying tool used during the creation of test fits, the example floor plans architects mock up to show clients how an office might be laid out. “It looks at your plan, does the math, which it’s really good at, and tells you you have five conference rooms, you have two pantries, you have 300 desks for people, and it spits out this legend,” Zaidi says. “That was a manual, tedious task that’s now being replaced by AI.” The firm has also explored whether AI could do the actual test fit itself, deciding where to place those rooms and desks, but so far the technology has not matched the expertise of human designers.
