The week the value moved below the model

The most consequential strategic story this month isn't a new model or a site-camera startup. It's Palantir, the American data and defence firm, walking into construction through the front door. On 4 June 2026, McCarthy Building Companies, one of the larger US general contractors, announced a multi-year, multi-million dollar partnership to build an AI operations suite called Pulse on Palantir's platform, connecting estimating, bidding, logistics, quality control and equipment planning into one system. Four days later, on 8 June, a new company called Cavtera launched in Ottawa with a single purpose, to commercialise Palantir Foundry for construction and field operations, built on more than 70 years of IP from Thomas Cavanagh Construction. A dedicated reseller channel forming around Foundry-for-AEC tells you this is a strategy, not a one-off deal.

Palantir positions itself as the decision layer that sits above your design tools, your ERP, your programme and your site sensors, pulling all of it into a single ontology the senior team actually looks at. AEC Magazine and the archBIM analysts spent the second half of June making the same point, that whoever owns that layer owns the project's source of truth. If value shifts from who draws the model to who understands the data coming off it, then BIM quietly becomes the base layer that other systems sit on top of. The same fortnight, Autodesk took its Forma project assistant out of beta, an action-oriented agent that drafts emails, builds reports and answers questions across a project's connected data, running on Autodesk's own construction data model and ontology. That's the same bet Palantir is making, just from inside the BIM stack rather than above it.

I wouldn't get carried away. This is North American, enterprise-scale, and aimed at firms with the data maturity to feed an ontology, which rules out most of the UK supply chain for now. Two honest cautions, though. The first is commercial. An ontology is persistent and cumulative, so the more of your operational logic you encode in it, the higher your switching costs become, which is a textbook lock-in risk your procurement people should be alive to. The second, on Palantir specifically, is reputational. Its growth has come alongside sustained controversy over its defence and immigration work, and any UK public-sector client or framework will have a view on that before you do. Worth knowing where the tool comes from before you build your reporting on it.

For your board pack: If a platform proposes to become your single source of project truth, put three questions to your board before you commit. Who owns the resulting data model, can we export it in a usable form, and what happens to our reporting if we leave. If the answers are soft, that's your answer.

We've spent a fair bit of this month on the planning fight over data centres, the design competition, the green-belt call-ins, the noise objections. New Civil Engineer's in-depth report on 24 June 2026 makes the case that we've been looking at the wrong constraint. The thing that actually decides whether a UK data centre gets built isn't the planning committee. It's whether you can plug it in and keep it cool.

Start with power, because that's where the queue is. The pipeline of demand waiting for a grid connection tripled in roughly seven months, from 41GW in November 2024 to about 125GW by June 2025, with somewhere near 50GW of that tied to data-centre schemes. Put that next to the fact that the whole country's peak electricity demand is around 45GW and you see the problem, the queue is bigger than the grid it's queuing for. On the ground that's a connection lead time of about seven years for a new 50MW site in London, per the figures NCE and others cite, and over a year even for an approved high-capacity connection. A consented site behind that wait is a slide in a deck, not a project. Ofgem has clocked it too, and its demand-connections reform, with phase one aimed specifically at the data-centre sector, is built to filter out the speculative bids clogging the line, with refundable deposits tied to milestones, commitment fees, and likely proof of planning before you even join the queue.

Then there's water, which gets far less airtime and shouldn't. A single megawatt of data centre can use up to 25.5 million litres a year for cooling, roughly the daily consumption of 300,000 people, and the Environment Agency is already warning of possible drought in 2026. The comparison only goes so far, because the better operators are moving to closed-loop and air cooling that barely touches the potable supply. But if you're a contractor or consultant pricing a data-centre job, the cooling strategy is now a commercial question as much as an engineering one, and "we'll sort it later" is not an answer a water-stressed region will accept.

The procurement filter: Next time a data-centre opportunity crosses your desk, ask two questions before you ask about planning. What's the grid connection date, in writing, and what's the cooling strategy and its water draw. If the answers are vague, you're being sold the easy half of the problem.

Here's the number worth pinning to the wall. In the 12 weeks to 30 May 2026, the Building Safety Regulator made 358 Gateway 2 decisions on higher-risk buildings, and around three-quarters were approved. That sounds like progress, and it is, set against the logjam that defined the back end of 2025. But it hides the real problem, which sits one stage earlier. Build UK reckons close to 30% of submissions are still found invalid at validation, which means they bounce before the regulator assesses the substance at all. You don't get marked down on your fire strategy. You get told the application wasn't put together in the required way, and you start again.

That gap between a 75% approval rate and a 30% rejection-at-the-door rate is the whole story. The schemes failing validation aren't necessarily unsafe. They're incomplete, or assembled wrong, or missing a document the regulator needs before it can even open the file. On 4 June 2026 the Construction Leadership Council and Build UK refreshed the Gateway 2 guidance suite, and tellingly the updates centre on the Fire and Emergency File and the Building Regulations Control Statement, two of the documents people most reliably get wrong. Build UK's validation checklist runs to 19 items. Most of the rejections come down to not working through it.

This is where AI earns its place, and where it doesn't. A model is genuinely useful for assembling a complete, consistently formatted application pack and flagging the gaps before submission. It's not useful, and is frankly dangerous, as the author of the safety judgement itself. Use the tools to stop yourself failing on a technicality, then keep a competent human accountable for the content. My own bias, for what it's worth, is that the firms who treat validation as a tedious clerical box-tick are the ones who keep losing weeks to it.

The discipline: Before your next Gateway 2 submission, run Build UK's items 1 to 19 as a final walkthrough, and have one named person confirm the Fire and Emergency File and the Building Regulations Control Statement are complete. A validation rejection is the most avoidable delay in the regime.

For two years the construction-AI funding headlines were about the visible end of the job, cameras on poles, drawing review, MEP design. This fortnight the money moved somewhere far less glamorous and arguably more useful, the part where you decide what to bid for and then write the thing.

On 16 June 2026, ContraVault AI raised US$3.1m in a pre-Series A led by Chiratae Ventures with Titan Capital. The Indian startup builds what it calls procurement intelligence, software that shortlists opportunities, reads the tender documents, flags the risky clauses and helps assemble the bid pack. It says it's been trained on more than a million tenders and runs across 13 modules, from a Go/No-Go analyser to clause editing, with a customer list that's Indian infrastructure rather than UK contracting (all vendor-reported). It wasn't alone. The same week, Swiss startup Scait closed an oversubscribed high six-figure pre-seed for an AI tendering platform, and Italy's Soource raised €3m, with the line worth remembering, that procurement is moving from a "copilot" model where the AI helps the human to an "autopilot" model where it does the task. That's three or four bets in a fortnight on the same idea.

What that means on a UK desk is simple enough. The bid team, the people who lose evenings to PQQs and quality questions under the Procurement Act 2023, are now squarely in the sights of the venture money. I'm not sure any of these tools survives contact with a British public-sector portal on day one, none is UK-built and none is trained on our procurement regime. But the direction's right, and the bid-no-bid call is exactly the kind of structured, document-heavy judgement these things are good at supporting.

The procurement filter: Next time you sit a bid-no-bid meeting, run the same tender through one of these tools as a shadow exercise and compare its read of the risk clauses to your own. If it catches something your team missed, that's the business case. If it invents risk that isn't there, you've learned that too, cheaply.

The second funding cluster worth your time is code compliance, and the interesting move is where it now sits. Inside the model. On 11 June 2026, Denver-based Kestrel Labs launched what it bills as the first AI compliance platform built natively inside Revit, alongside a US$2.15m pre-seed. One click runs a full compliance check in about 30 seconds, the vendor says, with every flag tied to a specific model element and cited to the exact code section. The same week, a Berlin startup called Baumind came out of stealth doing a strikingly similar job, reading the regulations, checking the 3D model against them, then generating the geometric correction for a human to approve.

Put that next to the UK problem it looks built for. Nearly a third of Gateway 2 applications for higher-risk buildings still get bounced before the Building Safety Regulator assesses the safety case, because something in the documentation fails validation. A tool that catches the compliance gaps inside Revit, before submission, is aimed squarely at that pain.

But, and it's a big but, read which code these tools actually speak. Kestrel checks against US codes. Baumind is built in Germany. Neither is reading Approved Document B or the rest of the UK Building Regulations out of the box. The spell-checker comparison only goes so far, but it's the right one, a spell-checker set to American English will happily wave through "color" and "thru" all day. A compliance bot trained on the International Building Code will do the same with rules that simply aren't ours. The capability is real. The localisation is the whole job.

For your board pack: If a compliance-in-BIM tool comes across your desk, the first question isn't accuracy or speed, it's jurisdiction. Ask the vendor, in writing, which code editions it's validated against and whether UK Approved Documents are covered. If the answer is "on the roadmap", it's a demo, not a tool.

A short, honest follow-up to a thread we've been pulling on. For weeks the trade chatter, and this brief, flagged Google's Gemini 3.5 Pro as the model to watch, a June release built for the long-document reasoning a golden-thread pack demands. As of last week, 3.5 Pro still hadn't landed in general availability. What did land, on 22 June 2026, was Gemini 2.5 Pro with Deep Think, made available immediately on the Gemini API, AI Studio and Vertex AI, with a 2-million-token context window and a step-up reasoning mode that Google's own benchmarks put at the front of the pack on science, maths and code.

So the capability arrived, just wearing a different badge than the one everyone was waiting for. That's worth saying plainly rather than pretending the prediction was bang on. What Deep Think actually does, for our purposes, is hold a very large document set in its head and reason across it in one pass, which is exactly the shape of the problem when you're checking an O&M manual against a safety case, or interrogating a fat specification for the clause that contradicts the drawings. The 2-million-token window is the headline, but the reasoning mode is the part that matters, because long context without good reasoning just gives you a confident summary of the wrong thing.

I'd voice the same caution I always do, and mean it. A model that's brilliant on a benchmark can still miss the one buried inconsistency that costs you on site, and Deep Think reportedly runs at around four times the standard token cost, so it's not the thing to point at every email. Use it where the document load is genuinely heavy and the stakes justify the spend, test it first on a job you've already closed where you know the right answer, and keep the competent person on the sign-off.

A practical step: Pick one document-heavy task your team dreads, an O&M review or a spec cross-check, and run it through Deep Think against an answer you already trust. If it finds what your reviewer found, you've got a time-saver. If it invents a problem, you've learned that cheaply, before it touched a live job.

A figure to keep you honest about the gap between the press releases and the site. RICS survey work this year still has around 45% of construction firms reporting no AI use at all, with roughly a third more stuck in early pilots (survey figures, so read them as direction rather than gospel). The more encouraging number from the same research is that about 38% of contractors now report measurable results from AI, up from 17% a year earlier. So the ambition for continuous, whole-life, machine-readable project information is arriving faster than most firms can absorb it. The standard wants it, the platforms sell it, the models can read it. The hold-up is us. The firms that close the gap won't be the ones with the cleverest model, they'll be the ones whose information is structured enough to feed it.

Worth doing: Spend the rest of the summer on information hygiene, not tool selection. A clean Information Production Schedule, built to where ISO 19650 is heading, will outlast whichever model tops the leaderboard in August.

One standing item worth restating in a week about ownership and controls. The RICS professional standard on responsible use of AI has been mandatory since 9 March 2026, and its core principle is blunt, AI assists, it doesn't replace, and the surveyor stays accountable for every piece of advice regardless of the tools used. Apply that to this week's stories and the action is concrete. Let AI assemble and check your Gateway pack, but sign it as a human. Let an agent read your project data, but only inside controls you set and can audit. Let a bid tool flag the risky clauses, but make the bid-no-bid call yourself.

The practical bit: Adopt one rule this quarter. Any AI output that goes to a regulator, a client or into the golden thread carries a named human sign-off. It costs nothing and it's the difference between a tool that helps and a liability that hides.