Leen: For interiors, AI is still evolving because our work is very detail-driven. It’s a little more complicated to apply than in architecture, but we’re exploring ways to use it more effectively.

Sherif: For architecture, we use a combination of publicly available AI models like those from Google, Microsoft, and OpenAI — and specialized tools for design and video generation, such as Kling or Veo. For computational design, we rely on Rhino and Grasshopper for rapid iterations and testing project feasibility, including built-up area and other constraints. Revit and Dynamo also play important roles in form development and performance-based analysis.

We run simulations for climate, solar radiation, daylight availability, wind, and views. All of this data informs the design, shaping how forms develop and how spaces are used. For instance, at Marjan Island, wind analysis influenced building angles to protect outdoor areas, while views analysis helped select forms that maximized key sightlines.

The AI-assisted workflow is iterative. It starts with mental ideation, then moves to brainstorming with a language model like ChatterBot or Gemini. Next, images are generated using AI tools such as Nano Banana Pro, and the prompts are refined based on these visuals. This loop — thinking, AI feedback, and sketching or modeling — enhances both conceptual design and execution, similar to the traditional mind-and-hand process, but faster and guided by data.