Can AI help us be better designers?

Image generated with ComfyUI & FLUX.

Artificial intelligence is rapidly transforming the way we approach design. In recent years, AI tools have flooded the market; some proving genuinely impactful, others fading as short-lived trends. As with any technological shift, only a few will endure: those that offer true value and integrate meaningfully into our workflows.

Many accessible, low-friction tools are being adopted in relatively straightforward design contexts such as home refurbishments, interior styling, or small-scale interventions. These solutions often succeed because they are quick, easy to use, and well-suited to the tasks at hand.

However, such tools are rarely sufficient for more ambitious design challenges, where innovation, precision, and conceptual depth are essential. In those cases, it becomes crucial to identify which technologies genuinely support design excellence, and which risk leading us toward oversimplified or generic outcomes. We must remain critical, adopting only those tools that enhance our process rather than dilute it.

"We must remain critical, adopting only those tools that enhance our process rather than dilute it."

Broadly speaking, the AI tools most relevant to architectural design today fall into three categories:

• Language Models: tools such as ChatGPT assist in articulating design intent, structuring workflows, drafting briefs, and developing presentation texts. They help organize thinking and refine ideas during early conceptual phases.

• Image Generation Models: platforms like Midjourney, and open-source models like FLUX or Stable Diffusion are increasingly used to create reference imagery, visualize early concepts, or enhance in-progress and final presentations. These tools accelerate ideation and communication, though they should complement, not replace critical design judgment.

• AI Design Assistants: tools like Spacemaker or Finch3D are capable of generating massing studies, conducting environmental simulations, and even producing detailed drawings. In many ways, they represent a new generation of intelligent, responsive design platforms, extending the capabilities of traditional BIM tools.

AI should not be seen as a replacement for human creativity, but as a set of tools that, when used wisely, can elevate the quality and efficiency of the design process. The challenge ahead lies in remaining intentional: embracing the technologies that sharpen our practice, while resisting those that offer only convenience at the expense of design integrity.

"AI should not be seen as a replacement for human creativity, rather more of a productivity booster."

My research focuses on how image generation tools can be effectively integrated into architectural workflows. While web-based platforms offer elegant solutions, they remain closed systems with limited flexibility. To go beyond surface-level use, designers need to understand these tools and tailor them to specific goals.

Image generated with ComfyUI & FLUX.

I’ve pursued this using ComfyUI and FLUX in a local setup, which offers greater control, scalability, and creative independence.

A local setup enables custom workflows, ideal to fit the process of image generation to our design intends. Furthermore, as AI outputs often need to be constantly refined, local setups allow for that long iteration process without the cost barriers of web platforms. Techniques like training LoRAs also allow for model fine-tuning, enabling designers to create distinct visual styles and avoid generic outputs.

"To move beyond surface-level application, we must take ownership of the tools: understand how they work and adapt them to serve specific design goals."

And let's not forget, privacy and autonomy are also vital. Local or private servers give users more control, avoiding the unpredictability of commercial services.

Of course, they do present challenges as well: steep learning curve, hardware costs and the need to keep up with rapid developments. Results can be inconsistent without fine-tuning. Still, for those willing to invest the time, the long-term benefits often outweigh the initial effort.