Construction's digital maturity problem
Construction is one of the least digitalised industries on the planet. McKinsey's research has consistently placed it near the bottom of the digital maturity index, alongside mining and agriculture. Labour productivity in construction has been essentially flat for decades — while manufacturing productivity has nearly doubled over the same period.
The reasons are structural. Every project is a prototype. The workforce is fragmented across subcontractors. Margins are thin — typically 2-5% net for main contractors — which suppresses R&D investment. And the industry has a deeply embedded culture of "the way we've always done it."
But this is not a lecture about why construction is behind. You already know that. The question is: what has changed to make AI adoption viable now, when previous waves of technology (BIM mandates, 4D planning, robotic total stations) achieved only partial adoption?
Three things have changed. First, AI can now process the documents construction actually produces — 2D PDF drawings, specifications, schedules — rather than requiring the pristine BIM models that most projects still do not have. Second, the cost of AI inference has dropped by roughly 90% in two years, making it economically viable for an industry with thin margins. Third, multimodal AI models can now interpret images, which means site photographs become data.
