AI Agents for Structural Design

Structural design plays a central role in a building’s environmental footprint and long-term performance. Key decisions about structural system, materialization, and load paths are often made at the project early stages and carry with them large impacts on sustainability, cost, and constructability of the final design.Yet current structural workflows remain highly manual and fragmented. Engineers spend substantial time on repetitive modeling, iterative checks, and re-formatting of information across tools. These inefficiencies reduce the number of alternatives that can be explored such that promising structural options may be overlooked, especially in the conceptual phase.AI agents offer a new interaction paradigm where instead of isolated tools, designers can collaborate with systems that understand intent, propose alternatives, and automate routine steps while keeping humans in control. Although these capabilities are rapidly emerging in other domains, their adoption in architecture, engineering, and construction (AEC) remains limited, especially for safety-critical disciplines such as structural engineering.This project aims to develop and evaluate AI agents that assist structural designers and AEC stakeholders in exploring, modeling, and comparing structural solutions at early project stages. Through close collaboration with industry partners, the project targets practical workflows that can enable more informed and sustainable structural decisions.

Scientific goals:

• Identify strategies to guide designers through alternative structural systems during early design phases supported by the use of AI models.
• Develop AI agents able to assist on repetitive modeling tasks allowing designers to focus on creative and strategic decisions.
• Evaluate how multimodal foundation models can interpret typical AEC input data to support building design.
• Demonstrate the potential of human-AI collaboration in the design process and evaluate changes in speed, breadth, and quality of structural decision-making in realistic design scenarios.