Construction AI: the government pulls consultation out of the data centre planning queue, and the AEC software map consolidates again

Here's a change that arrived without a launch video and will reshape how the biggest jobs in the country get consented. From 24 July 2026, the Planning and Infrastructure Act removes the mandatory pre-application consultation stage for Nationally Significant Infrastructure Projects. The government confirmed it on 3 July, and the trade and legal press worked through the detail across the following fortnight. What goes is the whole statutory front end: the Statement of Community Consultation, the Preliminary Environmental Information Report, and the duty to have regard to consultation responses before you even submit. The stated prize is up to 12 months off the timeline and around £1bn saved for industry this Parliament.

Why does this land on a construction desk rather than a planning lawyer's? Because data centres are named in scope, and they can now opt into the NSIP regime rather than fight it out with a local authority. Ministers have already directed three into it: Wapseys Wood in Buckinghamshire, Ampthill Road in Bedford and New Barn Lane in Dartford. What that means on the ground is that the delivery clock on a big shed starts sooner, which is exactly the pressure the AI capacity story has been putting on the sector all year. In place of the old consultation, developers get earlier technical advice from the Planning Inspectorate, so the engagement moves from statutory-and-public to advisory-and-private.

But I wouldn't file this under pure good news, and here's the honest bit. The consultation you cut from the front doesn't evaporate. It reappears at the back, as judicial review. What used to be an objection you heard early, and could design out before you'd spent a pound, can now surface as a legal challenge once you're committed and mobilised. Campaigners have already framed the reform as removing the public's say on data centres, which tells you where the friction lands next. The comparison only goes so far, but think of it like value-engineering a fire strategy out to hit a programme date. Cheaper today, and occasionally very expensive later. The people who carry that risk aren't the ministers announcing the saving, they're the project director who has to hold a mobilised site while a challenge grinds through court.

The procurement filter: If your pipeline leans on NSIP-route data centres, ask the developer how they're handling community engagement now it's voluntary, and price the judicial-review exposure into your programme risk, not just the shorter headline consent.

While everyone watches the frontier labs, the software your site teams actually touch is quietly gathering under fewer roofs. On 14 July, Nemetschek, the German AEC group behind Archicad, Bluebeam and a dozen other brands, confirmed it had completed its acquisition of Heavy Construction Systems Specialists, HCSS, the Sugar Land, Texas firm whose software runs estimating, fleet and field operations for heavy civil and infrastructure contractors across North America. The deal was announced back in April and closed on 1 July; the mid-July confirmation is the point it became real. HCSS turned over roughly $215m in 2025, with around 21% recurring-revenue growth and a 40% EBITDA margin, the kind of numbers that tell you this isn't a distressed tuck-in.

What the structure tells you is worth reading. HCSS goes into Nemetschek's newly named Build & Construct segment, with Nemetschek holding about 72% and the previous owner, private equity house Thoma Bravo, keeping around 28% as a minority. So a European design-software group is buying deep into American heavy civil, precisely the sector, roads, bridges, earthworks, sitework, that the data centre and grid build-out leans on hardest. That's not a coincidence. The money is following where the concrete is going, and the acquirers want the estimating and field data that comes with it, because that data is what trains the next round of AI features they'll sell back to you.

I'd keep one practical eye on this if you're a customer of any mid-sized contech tool. Consolidation is good for integrated roadmaps and bad for leverage at renewal. What we've found is that the moment a tool you rely on gets acquired, two things quietly change: the pricing conversation and who owns the data you've poured in. Neither has to be a problem. But the time to read the data-portability clause is before the next renewal, not the week your estimator tells you the export button's gone. That's what it's about, not the logo on the login screen, but whether your numbers stay yours.

Worth doing: List the three contech tools your teams would struggle to replace, and check each one's data-export and contract-assignment terms this quarter. If any got acquired in the last year, that's your starting point.

Put two data points from this month side by side and you get the real state of play. ServiceTitan's 2026 report, doing the rounds these past weeks, has measurable-impact AI adoption at 38% of contractors, double last year. A ConstructConnect survey of Canadian construction leaders, out on 7 July, found the enthusiasm just as high but day-to-day use still stuck in single digits, with most firms piloting or researching rather than running anything in anger. Different samples, different countries, and yet they tell the same story. The intent is near-universal. The doing is not.

Neither survey is UK gospel. ServiceTitan sells software and leans American; the ConstructConnect read is Canadian. So take the direction, not the decimal. But the direction is consistent with everything crossing this desk, and it's the widening gap that should hold your attention rather than any single percentage. The firms that got AI onto the estimating and bid desk a year ago aren't twice as good as the ones still watching, they're just quietly faster, and in a market short of roughly half a million people, faster is the whole game.

So here's the closer, and it's the same one I'd give any week. Don't buy the tool in the room with the best demo. Find out where your own numbers are made, and how long they take today. Tie the week's three threads together: the government just sped up the front end of the biggest jobs, the software you rely on is consolidating under owners who want your data, and the evidence says your competitors are getting quicker at the bid while you read this. The through-line isn't a product. It's the estimator turning round three tenders in the time one used to take, and the PM who isn't the last to know a decision changed.

A practical step: Time your next three tender returns end to end, this week, before you change a thing. If you can't say how long a bid takes you now, no tool will ever prove it helped.