Nomic hosted a 45-minute webinar on AEC-Bench, the first open, replicable benchmark for AI agents on construction tasks. Andriy Mulyar (CEO, Nomic), Harsh Mankodiya (Research, Nomic), and Conor Lenon (Fortis Structural) explored why general-purpose AI breaks on AEC documents and how domain-specific tools change agent behavior on real drawing sets.
The session covered AEC-Bench's 196 tasks across 9 task families and 3 scope levels, a live walkthrough of agent behavior on construction drawings, and practical takeaways for evaluating AI tools in the built environment.








