Construction AI: a bricklaying robot firm raises on delivered walls, not demos, and the golden thread meets its product data problem

Monumental announced a $32m Series B on 15 July, led by Khosla Ventures with existing backers Plural and Hummingbird returning. The company raised $25m in February 2024. Co-founder and chief executive Salar al Khafaji, who sold his previous company Silk to Palantir in 2016, describes that first round as "basically still an R&D round" and this one as raised off actual success. The numbers behind that claim: more than 150 robots in the field, brickwork delivered on over 100 homes, plus a school, a community centre, a sports centre, a hotel, commercial units and hundreds of metres of Amsterdam's canal walls. Nearly half of the completed homes were finished in the three months before the raise. That curve is the thing to watch, not the headline figure.

But the reason this matters for a UK contractor isn't the hardware. It's the commercial model. Monumental doesn't sell or lease robots. It bids brickwork like a specialist trade, quotes per thousand bricks or per square metre at rates close to local masonry pricing, and gets paid for finished walls. If the robots can't complete an agreed scope, human crews finish it at Monumental's own cost. What that means on site is that the customer never holds the technology risk, which is precisely the trap the previous generation fell into. FBR's truck-mounted Hadrian X asked contractors to commit nearly $6m per unit for a machine nobody had seen deliver. Al Khafaji's own diagnosis is partly about timing, since the computer vision and localisation stack simply didn't exist a decade ago, and partly about go-to-market. General contractors don't want to buy unproven kit. They want to subcontract work and manage subcontracts. So Monumental made itself look like something contractors already know how to hire.

The UK is the primary commercial focus for the next six months, with a local team already in place and delivery happening through VINCI business units including Taylor Woodrow. The labour arithmetic is doing a lot of the selling. The Home Builders Federation estimates the UK needs at least 20,000 more bricklayers to reach the 1.5 million homes target, against roughly 1,990 completed bricklaying apprenticeships in 2024. On speed, al Khafaji is refreshingly unbothered: a single set of robots averages between half and two-thirds of a human mason's daily output, with peak days at double or triple. The answer to a deadline is fleet size, not machine speed. Heritage work with decorative arches gets a polite no. I'm not sure the economics survive a wet February on a tight urban site with three trades stacked behind you, but the delivery record is real and the risk allocation is genuinely sensible.

Today's action: if a robotic trade lands on your tender list this year, ask three things before you look at the rate: who carries the technology risk, what happens to the programme if autonomy drops, and who is legally the contractor for that scope.

GS1 UK and Barbour ABI published a joint report on 14 July putting the cost of poor construction product information at up to £3.8bn a year. Not poor software. Not poor process. Poor product data: fragmented, inconsistent, hard to share, and frequently impossible to tie back to a single product identity across the organisations that specified, bought, installed and now maintain it.

The compliance numbers in it are the ones worth quoting in a board paper. 98% of professionals say they are aware of the Building Safety Act, and 21% say they are fully prepared to meet its requirements, nearly three years after the legislation came into force. Nine in ten say they are aware of the golden thread. 14% say they fully understand it. And 84% report products being substituted on site, which is the ordinary reality of construction and also the single hardest thing to trace once the building is occupied. 52% say a lack of digitalisation costs them revenue, rising to 69% at larger organisations. Three in five say inefficiency in managing product information is holding them back, and 48% call their current approach disorganised. Iain Walker, GS1 UK's director of industry engagement, put the risk plainly: without fixing product data, the golden thread becomes a repository of documents rather than a trusted record of what was actually specified, installed, substituted and maintained.

That last sentence is the whole item. It's a bit like a beautifully organised filing cabinet where half the labels were written by someone guessing. The cabinet is fine. The labels are the problem, and no amount of retrieval software rescues you from labels that don't match what's inside. So if you are buying an AI tool to search your building records this year, the question to ask isn't how good the search is. It's whether the underlying records identify the actual installed product, including the substitution the site manager approved on a Tuesday afternoon in 2023 because the specified one was on a twelve-week lead time. The sector is 99% SMEs, most of whom cannot fund a data programme, so this doesn't get solved by exhortation.

For your board pack: pick one completed higher-risk building and try to trace three installed products back to a manufacturer identifier and a current declaration of performance. Time it. That number tells you where you actually stand on the golden thread, regardless of what the compliance slide says.

I covered Google's slipping frontier model on 19 July, when Bloomberg had just reported that coding and reliability performance was holding it back. Here's the update, and it's a non-event, which is the point. The widely reported 17 July target passed. As of 20 July there is still no model card, no general availability, and no gemini-3.5-pro entry in Google's public API model list. Reporting now points to August. Bloomberg's 16 July piece said DeepMind had scrapped and rebuilt the base model after it fell short of internal quality goals on hallucination rates and real-world reliability. Gemini 3.5 Pro was announced at Google I/O on 19 May with a June general availability target. That's three slips in two months.

Two things follow, and neither is about Google. First, a lab that holds a model back over hallucination rates is behaving well, and I'd rather have that than a rushed release feeding a compliance workflow. Second, and less comfortably, if your procurement process or your product roadmap has a line item that depends on a model that hasn't shipped, you have taken on a delivery risk you don't control and cannot manage. That happens more than people admit, usually in the form of a pilot scoped around a capability someone saw in a keynote.

The procurement filter: if a supplier's answer to a capability gap is a model or feature that isn't generally available yet, treat it as absent. Price the deal on what runs today.

A robot that lays brick and a report about product labelling look like different stories. They aren't. Both are about whether the record of what happened on site is good enough to trust. Monumental's robots run off a 3D model with an exact brick count and a full build plan produced before anyone arrives, and its operators place localisation stickers on day one so the machines know where they are. That is a project where what was built and what was planned stay tied together by construction. The GS1 report describes the opposite condition: a sector where 84% of people watch products get substituted and almost nobody can trace it afterwards.

So the standing discipline hasn't changed. Whatever you buy this year, the value comes from the quality of the record underneath it, and the person who benefits is the site manager who no longer spends a Friday afternoon reconstructing what went in a riser eighteen months ago. That's what it's about.

A practical step: name one person who owns product identity on your next project, and put it in the appointment, not the aspiration.