Architecture and engineering professionals are reassessing everyday tasks such as specification writing and 3D rendering thanks to artificial intelligence language learning models (LLMs), like Open AI's ChatGPT and Google Bard, and text-to-image tools, like DALL-E and Midjourney. Designers are wondering which aspects of the industry can or cannot be automated—and which should not be. Will AI and machine learning tools present more opportunities or more risks? And what is their role in the age of AI?
Yehia Madkour, a Vancouver, British Columbia–based principal at architecture firm Perkins&Will, sees potential for designers who are open to new ideas. “It’s a matter of understanding … how you can leverage and empower yourself as a designer to use these tools to create better architecture and design—and liberate yourself as well.”
Designing and engineering opportunities for AI
Iterating design concepts by hand can take upwards of several days. Meanwhile, AI can, within minutes or seconds, generate a multitude of options from which an architect could then choose and build upon to meet their aesthetic criteria, as well as a developer’s pro-forma. This approach augments traditional design methods by allowing designers to articulate their ideas more effectively and efficiently, Madkour explains: “It is a matter of defining what the value is, where do we stand as designers, and what can we contribute within the different changing tools that we have.”
Laura Paciano, vice president of marketing at Dallas-based TestFit, has high hopes for AI in the design process. The company’s generative design platform uses AI to analyze programmatic and zoning requirements for early-state multifamily residential projects, outputting multiple design options from which architects and developers can identify optimal solutions. “This is going to get you 80% of the way in seconds,” Paciano says.
For a nominal subscription fee, today's industry professionals can access generative design tools like TestFit and Ark, the latter of which focuses on multifamily housing in New York. As a result, the industry’s attitude toward AI has changed for the better, Madkour says. “We’ve seen a major shift in sentiment.”
For Mike Lawless, a St. Louis–based senior director of innovation at engineering firm IMEG Corp., integrating AI and ML into the design workflow makes sense for rote tasks such as quantifying mechanical equipment and identifying duct layouts. “How do you get to the optimal solution?” he poses. “Iterate through thousands of options [generated automatically] and come up with the best one, or just choose from the best three you can come up with?” (Disclosure: I work for IMEG and collaborate with Lawless’s team, but I do not report to him.)
The key, Lawless continues, is to treat AI as an assistant. “Think of it as augmented intelligence instead of artificial intelligence, where it’s augmenting our skill sets and giving us the data we need to make good decisions.”
AI might also push the industry toward greater standardization, he hopes. “If AI and ML are able to drive productivity, then does that drive us toward standard [building and energy] codes? Because everybody will want to realize that efficiency within our industry.”
Everyday to novel AI applications
As LLM tools like ChatGPT are released to the public, architecture and engineering (AE) companies have begun testing them in administrative capacities—authoring emails, project proposals, and marketing materials—as well as in client-facing capacities.
Perkins&Will is developing a system that utilizes existing AI engines to listen to conversations, such as those between an architect and a client, and then generate images related to the conversation in real time, says Thomas Kerns, a digital innovation strategist based at the company’s Chicago office. The technology connects seven ML models “in a thoughtful way to produce one experience,” he says, and it translates “speech to text, text to concept, concept to sentiment, sentiment to prompt, prompt to image.” For clients unaccustomed to verbally expressing design ideas, this type of tool can improve communication with their architect.
AI also has applications in cost estimating and project budget tracking. With good data, AI can predict material quantities and add elements, such as plumbing fixtures, to a building information model (BIM) automatically. “If you can predict how many water closets or sinks you’re going to have, then you can automate getting a lot of those [BIM objects] into your model too,” Lawless says. The end result, he hopes, is a “much more accurate pricing model.”
But all these technologies hinge on the availability of good data, which takes time and effort to gather. “If you’re counting on people to enter data manually, you’re not going to have a good data set,” Lawless says. “[Data collection] has to be automated, and AI and ML are part of that automation.”
This data can be specific to one project or to an entire city. The goal of Ark’s platform, according to founder and architect Dity Ayalon, is to “help architects design better and faster by automating the repetitive tasks.” Ark automatically incorporates location-specific data, such as local building codes and ordinances, into its generative design platform, allowing architects to focus on design optimization.
This aligns with the expectations of San Antonio–based Lake|Flato Architects, which uses AI to improve its efficiency and to augment its staff's design abilities. Dan Stine, director of technology and leader of the firm’s internal research program, says having people—and not computers—in charge of the process of designing spaces for people is critical.
Professional replacement
Losing jobs to automation is a concern whenever groundbreaking technology is released. Tasks such as report writing, meeting scheduling, and email correspondence will eventually be assigned to AI rather than design staff, but IMEG software engineering team lead Steve Germano believes, other responsibilities will remain immune. Experience and artistic creation, he says, are “very hard for the computer to replicate.”
Madkour sees AI as an opportunity for those who can wield it. “It’s not really about losing jobs,” he says. “It will be about losing jobs to people who use AI.”
And the change might happen sooner than later. “I’m convinced that in five to 10 years,” Kerns says, “the practice and profession of architecture will not look the same as it does today.”
Even if AI tools replace some of the labor required for administrative and technical tasks, such as bookkeeping, spec writing, and image rendering, deploying them will still require talented professionals. Paciano likes to quote TestFit’s chief revenue officer Kyle Bernhardt: “[AI] is like the Ironman suit. It’s great, but you still need Tony Stark in there to work it.”
When AI quirks become risks
AI technology has several hurdles to overcome before it achieves widespread adoption. AI-generated “hallucinations”—false answers like the one that occurred during Google’s public debut of Bard—are a significant limitation and require guardrails. AI is “going to make things up that are … inherently wrong,” IMEG’s Germano warns. Professional liability for design decisions is already challenging, and the responsibility will likely continue to fall on the architect or engineer, not the AI or tech companies. “Now that we rely on so much tech, who’s there to do the due diligence, to understand the ‘gotchas’?” Germano says. “It’s the licensed engineer.”
Copyright is another issue. In February, stock image provider Getty Images sued text-to-image provider Stability AI for using 12 million images from Getty to train its Stable Diffusion platform without permission. The copyright violations were discovered when many Stable Diffusion output images contained the Getty Images watermark.
Similar scenarios can arise in architecture. An AI-generated rendering of a project could incorporate copyrighted material from another design, and blaming AI for the mistake may fall flat in court.
Legalities surrounding copyright and AI will require time to iron out, Lake|Flato’s Stine acknowledges. “It’s important for architects to stay informed about the legal landscape surrounding AI-generated content and adapt accordingly.”
Some traditionalists and technology holdouts view AI as yet another tool that distracts emerging architects and engineers from learning and accumulating firsthand knowledge. The more the industry relies on technology, IMEG’s Germano posits, “the less knowledge we gain and the less need we have to learn how things actually work.”
However, AI can be a way to break barriers among designers at different experience levels, Perkins&Will’s Kerns notes. With AI as their assistant, a junior employee can research a problem or context faster, as well as get help in generating appropriate questions for the senior person. “It will allow somebody [new to] the domain of architecture to have a conversation with somebody at a senior level,” Kerns says. “That’s a big game-changer.”
The rise of new AI-focused skills
While the risks of AI are real and evolving, some designers are not hesitating to hone the skills to work better with the technology. Prompt engineering, or interacting with AI systems to achieve effective industry-specific outputs efficiently, is one knack that may become quickly in demand.
A poorly worded prompt can result in misleading or incorrect results. However, a good prompt engineer can provide the necessary context and guardrails to ensure the AI system does not veer off into make-believe. Prompt engineering will become a skill taught by instructors and orchestrators, IMEG’s Germano expects. “Industry-specific engineers or architects that know their industry and their domain and learn how to interact with the computers efficiently” will be sought after for their ability to channel their experience into prompts that yield results, he says.
The industry has weathered shifts in technological paradigms before. New tools and workflows, from hand-drafting to CAD and BIM, have been introduced throughout the history of architecture and engineering. The AE professionals who adapt to technological developments can expect to remain relevant in the field. “To stay ahead in the industry," says Lake|Flato's Stine, "it’s crucial to embrace AI as a powerful tool that can help us reach new levels of creativity and efficiency.”
This story has been updated since first publication.
For more news, projects, and profiles in the smart buildings ecosystem, subscribe to the SBT newsletter and follow us on LinkedIn, Twitter, and Facebook.
