Every electrical foreman knows the meeting. Three trades crowded around a shared screen at 5:30 on a Thursday, someone's cold slice of pizza going stiff on the conference table, and a federated model that will not stop clashing a conduit rack with a sprinkler main no matter how many times it gets nudged six inches to the left. Somebody always says "just push it up a level," and somebody else always says that level is already full. The meeting that was supposed to take an hour eats the evening, and everyone drives home having resolved exactly one clash out of forty.

That meeting exists because building the coordinated model in the first place, especially on a job with a million square feet of gear, conduit, and cable tray, has always been slow, manual work done by people who would rather be doing almost anything else. This week brought a concrete number on how much of that manual grind can actually be handed to AI, and it came out of one of the least glamorous corners of construction: electrical modeling on a hyperscale data center.

Augmenta, the AI powered virtual design automation company, and E-J Electric Installation Co., a Long Island City based electrical contractor and an ENR Top 50 firm, announced this week that E-J completed initial model population on a data center project of more than one million square feet 8.5 times faster than its usual pace, turning roughly 693 hours of modeling work into 82 hours. "Model population" is the unglamorous but essential step where every piece of gear, every run of conduit, every cable tray gets placed into the coordinated model before clash detection and construction documents can even begin, the exact work that normally eats weeks before a single MEP coordination meeting can start being useful. Tom Kregel, executive vice president of E-J Electric, called it "a step change in efficiency and quality," while Augmenta co-founder and CEO Francesco Iorio framed the result as proof of what happens "when unparalleled industry experience like E-J Electric's is paired with frontier technology like Augmenta."

The two companies are not new partners improvising a press release. E-J Electric and Augmenta signed a five year strategic partnership back in April, explicitly built around using Augmenta's platform, which AEC Magazine has described as agentic AI applied to electrical design, to analyze a project site's free space and obstructions and generate a coordinated model that accounts for labor and material cost rather than just geometry. This week's number is the first hard evidence from inside that partnership that the pitch translates into an actual jobsite result, on an actual project, not a demo reel.

The timing is not incidental. Data center construction spending in the United States hit a $50.7 billion annualized rate in April 2026, according to Census Bureau data, up from $39.8 billion a year earlier, a jump of roughly 27 percent that has made data centers the single largest category of private office construction in the country. That is exactly the kind of demand curve that turns a slow manual process into an existential bottleneck rather than a mere annoyance. Owners racing to get racks powered and cooled do not have patience for a modeling phase measured in weeks, and electrical contractors who can compress that phase have a real edge in both scheduling and, just as important, in how confidently they can price the work in the first place.

That pricing angle is the part worth sitting with a beat longer. A model that populates in 82 hours instead of 693 does not just get built faster, it gets estimated faster and with fewer of the assumptions that turn into change orders six months later. The gap between "we think this is what the gear layout costs" and "we know because the model already reflects it" is precisely the gap that turns a bid into either a win or a painful lesson, and it is the same gap agentic tools are starting to close on the proposal and estimating side of the business generally, not just inside Revit. Electrical contractors have spent years being the trade everyone blames when the schedule slips because coordination took too long. A tool that turns weeks of population into a couple of days does not eliminate the coordination meeting entirely, clashes are still clashes, but it does mean the meeting can start from a real model instead of a half finished one, which is a genuinely different conversation to walk into.

None of this means every electrical contractor on every project is about to see an 8.5x speedup, one hyperscale job with a committed technology partner is not an industry average. But it is a specific, named, sourced result on a specific project type that is currently absorbing more construction dollars than almost anything else in the country, and it points at the same direction the rest of preconstruction has been drifting all year: the slow, manual, unglamorous middle of the process, not the flashy renderings on either end of it, is where AI is actually starting to earn its keep.

Sources