Construction AI Brief
A Toronto firm, Provision, has put AI agents onto the one preconstruction job most teams quietly skip, roughing out scope of work before the bid goes in, and a Californian contractor says it's halved the time per bid. Underneath all this, the plumbing that lets an agent reach into your existing software, the Model Context Protocol, had its biggest revision yet on 28 July, which is duller and more important than it sounds.

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.
Here's the job that quietly loses money on jobs you don't even win. When a tender lands, you've got hundreds of sheets of drawings, thousands of pages of specs and preliminaries, and two or three weeks to turn it round. So the full scope of work, the careful read that catches what's missing and who's meant to be doing it, is the thing that gets half done or skipped. "People love to bid jobs," ProWest Constructors' president Michael DeMarie told ENR on 4 August 2026. "But scoping is a tedious job."
ProWest, a public-works contractor out of Wildomar in California, put AI agents from a Toronto firm called Provision onto exactly that. You upload the plans and the project documents, and the agents rough out the scope, flag the gaps in what each subcontractor is meant to cover, and draft the RFIs. DeMarie says it's taken them from about two weeks to one week per bid. That's a contractor's own figure, not an independent one, so I'd treat it as a ceiling on the scoping task rather than a promise about your whole tender. But even a partial win here frees your senior people to spend their time on scope interpretation and risk, which is where tenders are actually won and lost.
The bit worth pausing on is where the tool sits. Provision's co-founder Luigi La Corte is blunt that it isn't trying to be your estimating system. "It would be a hard time getting a general contractor away from their Excel spreadsheets, so we're not trying to make our own Excel," he told ENR. "We're going on top of it." It reads files from BuildingConnected, Procore and, yes, Excel. And the most telling detail in the whole piece: ProWest didn't first use it to bid new work. They used it on jobs they'd already won, to get their subcontracts out the door before prices shifted underneath them. That's the skipped scope biting at the far end of the job, not the near one. What that means for the person doing it is simple. The scope that gets skipped is where the bid, and then the job, quietly leaks. A tool that does the tedious first pass and lets a human sign it off is worth more than any demo about robots.
Every one of these agents has to reach into software it doesn't own. The scoping agent has to read your Procore. Your assistant has to pull a document out of Autodesk. The thing that lets it do that without a bespoke integration for every pair of tools is a standard called the Model Context Protocol, and on 28 July it shipped its 2026-07-28 revision, the largest since it launched. Most of the work went into two unglamorous places: making the connections more reliable at scale, and tightening how an agent proves who it is, aligning it with OAuth 2.1 and OpenID Connect.
I know. Authentication protocols are not why anyone got into construction. But sit with what it's actually deciding. When you let an agent into a live project, the question that keeps a sensible BIM manager up isn't what the agent can do, it's what it's allowed to touch, and whether you can show afterwards which door it went through. Standardising the log-in is standardising exactly that. It's the difference between a new starter with a named pass and a set of keys left under the mat.
And there's a portability point riding underneath. Autodesk is opening a path for certified third-party MCP servers to be called inside its own Assistant, and it's already got MCP into Autodesk Construction Cloud and a Revit MCP in technical preview with Revit 2027. Read one way, that's welcome openness. Read as a builder, it's worth asking which way the pipe runs: can your agent reach your data through a standard you could point at a different tool next year, or only through a pipe the platform holds. The comparison only goes so far, but the wiring behind the wall is usually what decides whether the lights work. Ask to see it before you sign.
Put the two together and there's a pattern. The AI that's earning its keep this month isn't the walking camera or the frontier model. It's moving into the boring middle of the job, the scoping, the checking, the reading for what's missing, the work people skip because there's never time. And the plumbing that lets those agents plug into what you already run is standardising at the same moment, which decides whether they widen your options or narrow them.
So the discipline holds, whichever tool you're weighing. Pilot it on a job you already know the answer to. Name the person who owns it. Keep a record of what it touched. And ask the one question that separates a tool you can leave from one you can't.
The procurement filter: before you buy any agent, ask the vendor to show you how it reaches your project data, through an open standard you could repoint at another product, or a private connection only they hold. If they can't answer plainly, you've learned something.
Source: AI Agents Aim to Speed Preconstruction Scoping in Bid Preparation (ENR, 4 August 2026) →
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.