Construction AI: the tender that reads itself, and the scanner that sees inside the slab

Here's a launch that lands on the part of the job nobody films for the showreel. On 25 August, Volve, an Oslo startup founded in 2024 by Herman Smith, Abyl Ikshanov and Alf Jørgen Dovland, raised $3m (NOK 30m) in seed funding led by Skyfall Ventures, with Norrsken Evolve, OBOS Ventures, Antler and StartupLab along for the ride and OBOS's former chief executive Daniel Kjørberg Siraj in as an angel. That last name matters more than it looks: OBOS is one of the biggest developers in the Nordics, so this isn't a lab toy, it's money from people who buy tendering for a living.

What Volve does is read your tender and contract documents and turn them into a structured map of scope, requirements and risk. Instead of an estimator ploughing through a few hundred pages of PDF to work out what's actually being asked for, the platform pulls out the scope, flags the gaps and the risky clauses, and helps the team make the call every contractor makes badly under time pressure: is this one worth bidding, or do we walk away now and save the effort. It's already used by main contractors and public-sector clients across the Nordics, and the company says it's building its presence in the UK and continental Europe. The prize it's chasing is a €1.6 trillion European construction market that, as one write-up put it, still runs on PDFs.

I'd temper the excitement in one place. A tool that reads a tender is only as good as the pack it's given, and public tenders in the UK are a mess of addenda, clarifications and drawings that contradict the spec. I'm not sure any model reads all of that cleanly on day one, and I'd want to see it on a live OJEU-style job before I trusted the risk flags. But the direction is dead right, because the go or no-go decision is where contractors quietly lose the most money, chasing bids they can't win and mispricing the ones they do. Get that call right and everything downstream gets easier. That's the bit that matters, and it lands on the estimator and the bid manager, the people who currently carry it in their heads on a Friday afternoon.

The other raise this week I keep thinking about is Birdsview, another Norwegian company, which pulled in €3.7m for non-destructive concrete scanning. It runs ground-penetrating radar and lidar over a structure and turns the sensor data into a 3D model showing the rebar position, the slab thickness, and the cracks, voids and corrosion hiding inside it, all without touching the concrete.

If you've ever managed inspection on an existing structure, you'll feel why this is more than a gadget. The old way is to drill a core down through the slab, test that plug in a lab, and extrapolate the result across everything you didn't drill. It works, but it's a single point pretending to speak for the whole floor, and it misses the localised trouble, the void under one bay, the corrosion round a badly detailed joint, the thin patch where the pour went wrong. Scanning lets you read a much larger area, and there's no core hole to backfill afterwards, which on a live road or a working floor plate is the difference between a quick survey and a closure. The comparison only goes so far, but it's one blood sample versus an actual scan of the patient.

The caveat, same as ever with this kind of kit, is that radar and lidar hand you data, not judgement. Somebody who understands reinforced concrete still has to read the model and decide what it means for the structure. But it puts far more of the slab in front of that person, and it does it without a drill, a generator and a lane closure. For anyone dealing with ageing concrete, bridges, car parks, 1970s frames, that's a real change to how a condition survey gets done.

Here's the through-line, and it's the one most contractors miss when they weigh up a tool. We argue about AI in terms of hours saved, an afternoon off the takeoff, a Friday back on the bid. Fair enough. But there's a second return that's easier to bank, and it lives at renewal.

In the US, the broker Shepherd runs a programme called Shepherd Savings that knocks up to 25% off insurance premiums for contractors who use approved site tech, Procore, Autodesk and Raken on the project management side, OpenSpace and DroneDeploy on reality capture, telematics and sensors on top. The logic is simple and it's evidence-led: Shepherd found that firms running OpenSpace had up to 50% fewer loss events and materially lower loss rates once they adopted it (those are the insurer's figures, so treat them as a claim, not a law of nature). And in the week to 24 August, a startup called RockRose Risk raised $12.5m to push the same idea further, wrapping risk assessment, the mitigation work and the insurance into one vertically integrated offer. The comparison only goes so far, but it rhymes with what challenger insurers did to motor cover: price the risk off real data instead of a rating table, and reward the people who lower it.

None of this is live in the UK in the same packaged way yet, so don't ring your broker expecting a menu. But the mechanism travels, and it reframes the whole adoption question. If a tool cuts your claims, the data it throws off is worth money to an underwriter, and that's a harder number to argue with than "it saved us time". That's what it's about: the site manager keeping a clean visual record isn't just admin, it's evidence, and evidence prices risk.

A practical step: at your next renewal, ask your broker whether any of the tech you already run counts towards your premium. If the answer's a blank look, you've found a conversation worth starting.