Can AI tools like Claude actually perform construction takeoffs?
In this episode, Luigi, Andrew, and Mike test Fable/Claude against real construction projects with known ground truth. The results are mixed: AI is getting very strong at scope identification, especially across drawings and specs, but quantity takeoff is still harder to trust.
The conversation gets into where AI is already useful, where it breaks down, and why auditability matters. Can a model count doors, fans, and equipment? Yes. Can it reliably calculate concrete, drywall, paint, or labor-driven quantities without a human checking the work? That is where the debate starts.
The group also talks about whether subcontractors should use Claude today, why estimators still need to verify the outputs, and what this means for takeoff software, preconstruction workflows, and the future of AI in construction.
