Construction AI Brief
The Bloomberg rumour this brief carried a week ago has landed: Gravis Robotics closed a $200m round from SoftBank on 17 August, at roughly a billion-dollar valuation, and it comes with an eight-million-pound UK job to make Flannery's excavators drive themselves. On the same day Bedrock put fully autonomous diggers on live US infrastructure. And Anthropic made the bit that lets an agent learn your way of doing things generally available, which matters more to your paperwork 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.
A week ago this brief passed on a Bloomberg line that SoftBank was weighing a big cheque for a Zurich firm that makes excavators drive themselves, and told you to hold it loosely because it was a rumour with someone else's name on it. Hold it no longer. On 17 August 2026 Gravis Robotics confirmed a $200m Series A led by SoftBank, which the trade and tech press are calling the largest Series A construction robotics has seen. The round takes Gravis, a spin-out of ETH Zurich founded in 2022, to roughly a $1bn valuation. Europe's newest robotics unicorn, and it makes diggers.
Here's the bit that matters for a reader in Warrington or Wembley rather than Palo Alto. Gravis is leading an £8m UK government-backed CAM Pathfinder project with Flannery Plant Hire, the biggest operated plant hire firm in the country, to retrofit excavator fleets with its Gravis Rack. What that does is turn autonomy into something you rent. Not a machine you buy and a system you integrate and a team you hire, but a digger on a hire schedule that happens not to need someone in the seat. The Rack clips onto Caterpillar, John Deere, JCB and Hitachi, and Gravis claims a 30 per cent productivity gain over manual operation. That last figure is the company's own, so treat it as a claim, not a fact. I'm not sure 30 per cent survives contact with wet clay and a tight dig profile, but the direction is no longer in doubt.
And it's not one firm talking. The same day, 17 August, Bedrock Robotics said its excavators are now working fully autonomously on live US infrastructure, including a water treatment site in Nevada with Sundt and a 1.2m cubic yard civil job with Zachry, with Bedrock saying it's raised $350m to date. So on one day you had the biggest raise the sector's seen and a competitor putting the thing on real ground. What that means for the person running an earthworks package is simple enough. The question this year isn't whether an autonomous excavator exists. It's what it costs to hire one against a manned one, and whether your ground and your programme suit it. That's a bid decision, and it's landed on your desk earlier than most of us expected.
While the diggers grabbed the headlines, Anthropic did something quieter on 19 August 2026 that's worth a construction reader's attention. It made two features generally available on the Claude platform, taking them out of beta: Agent Skills, and a Files API. Both sound like plumbing. Both change what you can safely hand an agent on a project.
A Skill is a reusable pack of instructions that teaches an agent how you do one particular thing. Your ITP wording. The way your firm writes a progress report. The clauses you always check in a RAMS. Rather than re-explaining your standard every time, you teach it once and the agent carries it. What making that generally available does is move it off a beta footing, so the tools you buy, or the ones your own IT builds, sit on something stable rather than something that might change under them next quarter. It's the same move Procore made a fortnight ago when it shipped Skills into its own agents. When two firms at opposite ends of the market reach for the same idea in the same month, that's the idea becoming the standard.
The Files API is the less glamorous twin, and arguably the more important one. It's how documents go in and out of an agent, now with proper versioning, expiry and permissions instead of a beta workaround. Think of it like the difference between letting a subbie wander your site with no induction and giving them a signed-in pass that says where they can go and when it lapses. The comparison only goes so far, but the point holds: when an agent reads your drawings and writes your registers, the two things that decide whether it's safe are what you taught it and what it was allowed to touch. Skills and files are exactly those two surfaces, and both just grew up. For the site manager whose name is on the record, that's the reassurance worth asking your software supplier to demonstrate.
Step back and the week tells a tidy story. The big cheque, $200m, went into the machine that moves the dirt. The quiet release went into the agent that does the paperwork. One end of construction and the other, funded on the same three days. And both land in the same place: a UK project team deciding what to trust with a real job.
The discipline doesn't change with the headline. Whether it's an autonomous excavator on your earthworks or an agent drafting your registers, you buy the answer, not the demo. Pilot it against a package you've already priced or a report you've already written by hand, and check whether it agrees with you before you let it run. The autonomy is a hire decision and the agent is a governance one, but the test is the same: prove it on something you already know the answer to.
Today's action: On your next earthworks package, ask your plant supplier for a straight price on an autonomous excavator against a manned one. Even if the answer is no this time, you'll know the number, and you'll be asking it a year before your competitor does.
Source: New construction robots gain traction on jobsites (Construction Dive) →
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.
RICS published its 2026 AI survey on 18 August: two thirds of construction professionals now use AI in some form, but only 19 per cent use it regularly, and the barrier that grew fastest was security. On the same week the contech cheques went to spatial data and to the transformer shortage stalling data centres, not to the office admin most contractors think AI is for.
Found this useful? Share it.
Gravis Robotics closed a $200m SoftBank round on 17 August with an £8m UK job to make Flannery's excavators drive themselves, and RICS published survey numbers the next day showing two thirds of construction people now use AI but only 19 per cent rely on it. Every capability story this week ran into the same brake, which is people and trust rather than technology.
OpenAI put a date on how long defenders have before capable open-weight models arm attackers as well as everyone else, in the same week two of those open models landed cheap enough to run on a site laptop. Procore packaged agents into its suite and built a dashboard to meter the tokens. And the AI data-centre boom keeps running into the one thing money can't conjure fast: skilled trades.