AI tools for architecture: Designers explore what’s possible

So, how did we use AI tools for architecture in our design competition project?

• We used AI for massing, modeling, and orientation: We set up the site in a realistic context to understand nearby buildings and daylight orientation. AI tools helped us experiment and iterate on various possibilities for the building’s shape. And it allowed us to test them for sustainability. We even used AI tools to convert 2D images into 3D models, which is a new approach for architecture.
  
• We used AI for functional programming and generating floor plans: We used AI to determine the appropriate square footage for the headquarters. Our team validated AI’s recommendations for the program against space utilization by real aerospace companies. We ran employee profiles through a large language learning model (LLM) platform to align our program with user experience. And we used an AI-powered tool to automate and iterate floor plan layouts.
• We used AI for interior materials selection: We trained an AI tool on the interior design for various airport projects that feature a mass timber structure. We used the AI image generator to see how various material palettes would relate to the mass timber in the workplace interior.
  
• We used AI for parametric façade design: Parametric design creates patterns based on a user-defined algorithm. It can take a team a week to program and create a parametric façade. The fictional company’s origami-inspired logo inspired our design concept and facade design. Our team leveraged AI to generate scripts that we used to create a complex parametric façade in our 3D model. In this way, we integrated AI-powered design with our BIM workflow. This approach reduced the time required for this task from about 17 hours to just 2 hours. Many of our design teams already apply AI in this way in their daily workflows.
  
• We used AI for design visualization and rendering: Detailed, photorealistic renderings can take many hours to produce and update. AI tools helped us quickly produce realistic renderings. This made it easier for us to iterate and evaluate our design ideas. AI lets us perform minor editing tasks quickly, enhance details, add people to our renderings, and test various furniture layouts. We validated the 3D environment for user experience with virtual reality goggles.

What lessons did we learn about using AI tools for architecture?

What did we learn about AI in IdeaXchange design competition? We can summarize the lessons of AI in the following AI design mantras. They are, like our AI-powered design process, a work in progress that we will continually refine.

1. Approach AI as a tool, not a decision-maker: Understanding where and when to apply which AI tool is important. And it’s still work. Creating prompts, optimizing workflows, choosing inputs, sorting through the output, retouching imagery. All of that takes time. There’s still a lot of work done outside of AI that defines the project outcome. We can use AI to augment our design capabilities, but it doesn’t replace human creativity or choice. Designers remain the ultimate decision-makers.

2. Overcome fear through use: For some of us, the first emotion when it comes to AI is fear. We were able to overcome fear through experimentation, play, and curiosity about the power of AI. Using it to iterate and refine our project helped demystify AI. We built confidence in our ability to guide the design process and achieve satisfying results. We’re thrilled to have the chance to push this technology to the limit.

3. Experiment, iterate, and play: Success with AI comes from approaching it with curiosity, testing what is possible, and learning from it. Again, and again. Some AI tools are new to us; we need to try them out to see where they are most valuable.

Design is an iterative process. Using AI at this stage doesn’t change that. Years ago, we were playing in Midjourney and finding the results unsatisfactory; the quality was not what we required for work. But we kept on playing. Now, as technology is evolving, we are ready to push AI tools to places where they bear fruit. AI is very good at generating multiple options. Practically speaking, AI speeds up the iterative design process. As we sharpen our design intention, it allows us to make hundreds of edits and re-export the imagery.

4. Choose the right AI tool for the task: A good designer knows that they are better at some things than others. We are always looking for tools that can do the things we don’t like to do or to take on arduous, time-consuming tasks.

For example, we love creating renderings, but we don’t always love that they can take 60 hours to generate. What if we could create a quality rendering in 3 hours? That’s the kind of tool we want in our toolkit. On the IdeaXchange project, we were able to dial in astounding results using AI’s image-generating capabilities in a fraction of the time it would take to render them in conventional technology.

Maybe some AI tools for architecture aren’t ready for day-to-day use? Recognizing which AI tool fits a specific design challenge is key; not every tool is suitable, and part of the process is reducing our tech stack to what works best. Example: This competition gave us the opportunity to research certain applications. We tried and discovered that one well-known app was a clunker. It was not useful for our purposes. And we’re able to share our results with designers. That way, we can spend more time with better tools.

5. Combine AI with other powerful design tools: It’s inevitable that we will find ways to combine different AI tools for architecture to find more efficient workflows. In IdeaXchange, we discovered that using AI in the design process was extra powerful when used in conjunction with the latest and greatest in our digital design toolbox. When we take this intentional, professional approach, we can constrain and control the AI tools to produce output with design utility.

For example, we created the building’s complex origami-inspired façade by using various AI tools in conjunction with Grasshopper, an algorithmic design tool. To get the most out of AI, we had to make sure our model was Revit-friendly. We had to ensure that the models we were building with help from AI could be used in our downstream workflow. We wanted to be able to apply our sustainability modeling tools to the models built in AI.

6. Find solutions through collaboration and share them: The value of our experimentation is multiplied when we tell our colleagues what we’ve learned. In IdeaXchange, we shared both our discoveries and failures. That helps other team members learn about tools and workflows along the way. We plan to share more best practices for AI tools for architecture as we develop them.

7. Remember that creative freedom fuels AI experimentation: The value of internal competition can’t be underestimated. This competition’s format allowed us to take risks and explore AI tools for architecture in ways that would be impractical with a current project for an existing client. The freedom to create, explore, and iterate was only limited by our time. It had promising results.

The project helped shift team perceptions, showing that AI can be a positive, creative force in design rather than something to fear. Hands-on experimentation helped us understand the limitations and strengths of AI tools. Through experimentation in the competition, our team moved from apprehension to confident adoption.

8. Apply your AI design experience in your day-to-day practice: What to do with all this fresh experience? Don’t lock it away, use it, share it—thoughtfully and appropriately.

Once we discovered positive applications for AI in design and saw its reliability and potential for refinement, we soon found we were integrating it into real projects. This is perhaps inevitable. We always look for the best methods and solutions.

One example? Since the competition, we have used AI tools for architecture to improve our workflow efficiency in response to requests for proposals. We applied AI to speed up the process of generating visualizations for RFPs, allowing our team to complete multiple proposals in a short timeframe and include high-quality renderings in the materials we submitted. This gives potential clients a more immediate understanding of our design vision

Of course, that vision is not the same as the final design. It will take months or years of design development for us to deliver a final, buildable design package. However, with AI tools, we can quickly test the design idea and get a client’s reaction. We can share a visual representation more quickly and accurately than before. This gives them something to evaluate and react to before we commit to a design. It gives the client room to think about what they need in far more detail.

Embracing change but staying true

In IdeaXchange, we learned that these new AI tools for architecture and design can be powerful. But we’ve also been reminded that technological change is nothing new in the design discipline. We can embrace progress while staying true to our sense of good design. With that in mind, we can adapt to change and benefit from using AI tools.

Our IdeaXchange experience has given us the confidence and experience to apply AI in our workflows; speed up processes like massing, programming, and design visualization; and achieve high-quality design. We won’t look back.

IdeaXchange Team 7: Joseph Bastone, Hans Davis, Edson Figueiredo, Ekaxi Hernandez, Erin Kilberg, and David Lee