The most important thing happening in AEC right now is not the tools. It is what happens to workflows when time and data distance collapse.
McKinsey recently argued that the biggest gains from agentic AI come not from point solutions but from reimagining entire end-to-end workflows. Not automating a task. Rethinking why the task existed in that form at all.
In AEC, the delivery process was always fragmented by necessity. Design, engineering, procurement, construction, and operations were separated not just by organisational boundaries but by time gaps and data handoffs. A coordination problem so deep it shaped the structure of every firm and every contract.
Those gaps are not permanent features of the work. There were constraints of the medium.
When AI compresses the time between a decision and its downstream consequences — when data no longer loses fidelity across handoffs — the logic of those boundaries weakens.
What replaces them is starting to come into view, at least in other industries, and it will come faster for AEC as well.
Engineers won’t no longer just producing outputs. They are going to build the automation that produces outputs. The most valuable technical professionals in AEC over the next decade will not be the ones who are best at running the workflow. They will be the ones who design it, encode it, and hand it to an agent.
Project managers are going to shift position. Not inside the loop managing handoffs. Above the loop — setting intent, coordinating between agent workflows, exercising judgment at the points where the work requires it.
McKinsey’s framing is useful here: human above the loop does not mean hands-off. It means repositioning human judgment from execution to oversight. Squads of agents running the core process, with humans accountable for the decisions that matter.
This is where Andrew McAfee’s observation differs in AEC from other industries. The learning cadence has compressed. What used to take years of iteration now takes months. Firms that start encoding their workflows now will not just be faster, they will be compounding knowledge while their competitors are still debating adoption.
The structural question this creates is the interesting one.
If time and data distance collapse across the delivery process, does integration follow? Does one entity absorb the stack vertically?
The more likely answer is a network. Not one firm owning design through operations, but a set of specialised entities: designers, engineers, contractors, platform operators, coordinating more tightly than was previously possible. Agent-to-agent handoffs are replacing the slow human intermediaries that populated every boundary.
Vertical delivery without vertical ownership. Coherence through shared data layers and interoperable workflows, not through a consolidated hierarchy.
The organisational models being tested at Meta and Block — smaller teams, player-coaches, flatter structures — are symptoms of something broader. You need people who can set direction, build automation, and hold accountability for outcomes.
The distance is collapsing. The question is whether the industry redesigns around that or waits until the new configurations appear without it.