Amplifying Creativity: The Role of AI in our Design Process

As impactful as AI is becoming, innovation must be balanced with an awareness of the inherent biases within these recent technologies. Early in our discovery phase, we published guidelines to help those curious about experimenting with AI. The main message was “trust but verify.” This is because widely available AI tools are often designed to satisfy the user, generating “hallucinated” results—outputs that seem plausible but are incorrect.

We learned that AI models are trained on historical and perhaps outdated data sets and from various sources across the Internet, which may perpetuate obsolete design norms and biases. We ask our designers to seek feedback from their teammates and disclose to our clients the areas in our design process that use AI. With clear guidelines and transparency, we can harness AI to enhance our projects without compromising quality or ethical standards.

Building upon our firm’s history, culture, and evolving AI guidelines, we began exploring, experimenting, and sharing our findings. This helped us craft a strategy for developing AI-enabled applications and design workflows. By focusing on practical, readily achievable solutions, we have avoided much of the inflated expectations surrounding these technologies and have become more confident in developing solutions that help our design teams to, as Pixar co-founder Ed Catmull said, “be wrong as fast as they can.”

We have built several AI tools, including an iterative rendering application. Unlike conventional rendering engines and early text-prompt generative AI tools, this tool produces images ranging from stylized illustrations to photorealistic renderings generated from simple massing models, including site context. It allows the designer to make incremental changes to specific aspects of the generated image without requiring a wholesale change—a limitation of earlier generative AI tools. This iterative function increases the tool’s value while accelerating early conceptual design explorations.

Another tool, affectionately called Chato (Portuguese for “boring”), performs data assurance and validation tasks that are often considered tedious. Chato enhances design data by comparing, correcting, and aligning room, department, and space type names across various sources—including space programs, Building Information Models (BIMs), and additional data sources like cost, material, and carbon information. Chato has the potential to unlock BIM’s long-held promises by addressing the “garbage in—garbage out” problem.

For instance, Chato replaces the repetitive and mundane tasks of completing program variance reports, counting space names on PDFs, and manually updating Excel reports. By automating these tasks, Chato frees our designers to apply their creative talents more effectively to solve our client’s challenges.

Rework can be minimized by ensuring data accuracy with Chato. Automating these mundane tasks increases efficiency and enhances job satisfaction by allowing our designers to engage more deeply with the creative aspects of their work.