AI in AEC isn’t about chatbots; it’s about orchestrating specialist agents around a machine-readable “building state” to compound and scale professional expertise, writes Paul O’Carroll, CEO of Arcol
The conversation around AI in AEC is often distracted by two extremes. One is the “Lovable” future, where design is reduced to a consumer-grade chatbot that spits out “good enough” buildings from simple prompts. The other is the “Cursor” future — a professional-grade, agent-first environment where AI is deeply embedded in domain logic to compound, not replace, human expertise.
We believe the Cursor model is the only viable path for the industry. However, achieving it requires a fundamental shift in how we work. The real unlock for AEC isn’t about better chatbots; it’s about orchestration. Buildings are not just geometry; they are high stakes coordination problems. Today, a single structural shift triggers a manual, fragmented ripple effect through MEP, cost plans, and procurement. This web of consequences is currently navigated through “dark data” — the institutional memory and lessons learned trapped in email threads or the heads of senior partners. To move forward, we must stop treating the building as a static file and start treating it as a live, machine-readable “Building State.”
Discover what’s new in technology for architecture, engineering and construction — read the latest edition of AEC Magazine
👉 Subscribe FREE here
The multi-agent ecosystem
At Arcol we’ve spent a lot of time trying to solve the human-to-human collaboration problem. How can multiple people interact with the same model at the same time? Now we are aiming that work at agents. The infrastructure we built is a prerequisite to enable a user to collaborate with dozens of agents solving discrete problems in real time. Without this you are just re-creating the fragmented workflows that plague AEC today. An AI future in AEC will be dictated by the orchestration of the new tools and agents.
Imagine a workflow where, as you focus on the human experience of a space, a Compliance Agent continuously audits life-safety requirements. A Sustainability Agent suggests carbon-neutral material swaps in real-time, while a Pre-con Agent maps every wall shift to a live cost plan
Imagine a workflow where, as you focus on the human experience of a space, a Compliance Agent continuously audits life-safety requirements. A Sustainability Agent suggests carbon-neutral material swaps in real-time, while a Pre-con Agent maps every wall shift to a live cost plan.
In Arcol these agents don’t compete for the steering wheel, they are deployed by an orchestrator thoughtfully. They collaborate under your direction, seeing the ripple effects of every decision through a shared data model. If the Sustainability Agent suggests a change in the building envelope, the Pre-con and Structural agents immediately flag the cost and coordination impacts. This is collaboration in real-time, moving the industry away from “post-mortem” reviews at the end of a design phase.
Scaling intelligence
The value of this shift isn’t just efficiency; it’s the compounding of expertise. In legacy software, lessons learned on one project are rarely captured for the next. In an orchestrated environment, a firm’s unique detailing logic and preferred assemblies are encoded and executable. This changes the economic leverage of the industry. Today, firms sell hours. But when expertise is orchestrated via agents, firms can begin selling outcomes — guaranteed constructability, instant feasibility, and hyper-optimised performance.
We don’t believe in a future where everyone is an architect. We believe in a future where the “drudgery of coordination” is offloaded to the platform, leaving the professional to focus on what matters: judgement. Modelling won’t be the job. Orchestrating the outcome will be.
Paul O’Carroll will be presenting at AEC Magazine’s NXT BLD 2026 conference in London 13-14 May.
