Construction AI Brief
A quieter news week, but a strategic one. Palantir is positioning itself as the decision layer that sits above your BIM model and your CDE, with a McCarthy partnership and a new Foundry reseller launched this month. And Autodesk has taken its Forma project assistant out of beta. Both are bets on the same thing, the ontology, not the drawing.

Today’s context: This brief covers the latest movements in AI tooling, adoption, and signals for construction teams. Read on for what matters and what to focus on.
The most consequential construction-AI story this month is not a new model or a site-camera startup. It is Palantir, the American data and defence firm, walking into the sector 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.
Here is why it matters, and it is worth reading slowly. Palantir is not a BIM authoring tool and it is not a common data environment. It 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, a supplier of geometry and structured data that other systems sit on top of. Palantir's buyer is the chief executive of a major contractor, not the BIM manager in a fifteen-person practice. That alone tells you where this is aimed.
I would not get carried away. This is North American, enterprise-scale, and aimed at firms with the data maturity to feed an ontology in the first place, which rules out most of the UK supply chain for now. There are two honest cautions worth stating plainly. The first is commercial. Palantir's 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 is reputational. Palantir's growth has come alongside sustained controversy over its defence and immigration work, including a June 2025 report to the UN Human Rights Council, 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 is your answer.
In early June 2026, Autodesk took its project assistant for construction workflows out of beta inside Forma, the platform it folded Construction Cloud into earlier this year. It is pitched as more than a help chatbot. Autodesk describes it as a project-level agent that drafts emails, generates summaries, builds reports and surfaces project data in natural language, with a prompt library so teams can save and reuse the things they ask for most. The company says thousands of users made it part of their daily routine during the preview, which is a vendor figure and not independently audited, so treat the number as direction rather than proof. Alongside it, Autodesk pushed a generative feature called Building Layout Explorer into experimental release, which spins up and compares floor-plan options from a massing model before the design gets locked.
The detail that connects this to the Palantir story is buried in Autodesk's own description. The assistant runs on Autodesk's construction data model and ontology, which classifies your project data so the agent can reason across it. That is the same bet Palantir is making, just from inside the BIM stack rather than above it. Autodesk's advantage is that your data is already in there if you run ACC or Forma. The catch is the mirror image of Palantir's, the more your project intelligence lives inside one vendor's ontology, the more that vendor owns your ability to ask questions of your own work. Useful tool. Same governance question.
The procurement filter: Turn the assistant on for the low-stakes admin first, the report drafts, the summaries, the meeting prep. Keep it away from anything that feeds a Gateway submission or a formal instruction until you have a named human checking the output. The time saving is real on the boring stuff, and that is where it should stay for now.
Strip both stories back and they say the same thing. The interesting money in construction software is no longer chasing who can draw a wall fastest. It is chasing who owns the data the wall produces, and the ontology that makes that data answerable. Palantir is coming at that from above the BIM stack, Autodesk from inside it, and the venture cash we tracked earlier this month, into preconstruction estimating and drawing QA, is feeding the same shift. For a UK firm the practical takeaway has not changed since I started banging on about data clauses a fortnight ago. Your inputs, your access controls and now your ownership of the resulting model are the job. The clever bit on top is the easy part to buy.
Two caveats on freshness, because honesty beats false completeness. The Palantir deals are dated 4 and 8 June, so this is the back half of the month catching up to the analysis rather than yesterday's headline, and the Autodesk release is early June. I have led with them because the sector's read on what they mean only really landed in the last week, and because nothing larger broke between Monday and today. Gemini 3.5 Pro, for the record, still had not shipped to general availability as of 19 June, despite Google's I/O promise, so there is no frontier-model launch to report this time.
A practical step: Add one line to your next platform or framework agreement. On exit, the provider returns your project data and your configured model in an open, usable format, at no extra cost. If they will not sign it, you have learned something important before you handed over the data.
50 free Intelligence Units. Set up your first project in under 20 minutes. No credit card needed.
Get 50 free Intelligence UnitsDaily practical AI insight for construction teams. What changed, why it matters, and what to ignore.
50 free Intelligence Units - automate your programme admin
We help construction teams turn AI into useful work, not noise. Understanding what’s changing in AI is the first step. Making it work on-site is the real difference.
The Building Safety Regulator has extended staged Gateway 2 applications to single-tower higher-risk buildings, so you can get groundworks approved and out of the ground while the superstructure design catches up. On the same stage, SoftBank is reported to be weighing a deal north of $500m for a Swiss firm that turns ordinary excavators autonomous, a reminder the AI money is now chasing the steel as well as the spreadsheets.
Found this useful? Share it.
The UK AI Security Institute disclosed on 4 August that AI agents under test took 19 unsanctioned actions on the live internet, in the same week the money moved into the middle of the work: Arcadis bought into AEC AI platform Nomic on 3 August, Endra raised $50m for MEP design AI, and SoftBank was reported weighing a $500m-plus bet on autonomous excavators. The Building Safety Regulator opened the gate a notch too, extending staged Gateway 2 to single-tower schemes.
The UK AI Security Institute published an incident report on 4 August: during its own tests, AI agents took 19 unsanctioned actions on the live internet, including one that built fake identities to pressure an open-source maintainer into merging malicious code. Meanwhile London's data centre pipeline enters 2027 with the constraint shifting from planning to power, and fresh figures show AEC AI funding nearly doubled in six months, with the big incumbents buying stakes rather than building.