Why Your Competitor's Knowledge Is Compounding While Yours Is Walking Out the Door

This is best enjoyed using audio. To listen, click here: https://creators.spotify.com/pod/profile/bricksandbytes/episodes/Executive-Briefing-Construction-Just-Got-Its-Google-Translate-Moment-e3g1ic1

A CTO at a 3,500 person engineering firm told me this week that the construction industry just accidentally solved its oldest problem. The Tower of Babel. And most people haven't noticed yet.

The same week, a war broke out in Iran, the Strait of Hormuz came under pressure, and one of the most respected economics directors in UK construction told CEOs to stop chasing the cheapest source and start securing the most reliable one.

And somewhere underneath both of those stories, a quieter question emerged that I think matters more than either of them. What happens to the firms that capture their institutional knowledge before it retires, versus the ones that don't?

Three threads. All connected. Here's what happened.

The supply chain is a boardroom problem now

Noble Francis, Economics Director at the Construction Products Association, put out a post this week that every construction exec should read. His message to CEOs and boards was direct. This is not the stability of ten years ago. There have been major disruptions every year for the last six years, and there will continue to be major disruptions every year. Firms won't be able to predict what those disruptions will be.

His advice: fully understand your supply chain, fix contracts for the medium term where possible, and diversify sourcing even if it costs more. His line that sticks: security of supply will almost be as important as the cost of supply.

The geopolitics back him up. Iran has put shipping routes under pressure again. Oil prices are moving. And if you remember the Red Sea disruptions, the pandemic chaos, the Russia-Ukraine energy spikes, you know how fast this cascades into material pricing and delivery times.

The Daily Blueprint flagged the numbers. US nonresidential construction input prices are running at a 7.1% annualised rate, concentrated in tariff-exposed materials like copper wire, steel, and industrial controls. This isn't broad inflation. It's a targeted squeeze on specific cost lines.

Meanwhile, we've been speaking to founders building procurement software as part of a deep dive into the space. The picture is consistent. Somewhere between 400 and 600 billion dollars of US construction materials are ordered annually, most of it still managed through Excel, email, and phone calls. The average contractor is overpaying by roughly 5% on materials without knowing it. And research suggests around 40% of labour idle time on projects is caused by materials not being where they need to be. In an industry 400,000 workers short, that's not a footnote.

Read the report and feature tease: https://bricks-bytes.beehiiv.com/p/excel-runs-600b-of-construction-procurement-three-founders-want-to-change-that

Martin, one of our co-hosts, pushed back on Noble's advice during the show. His argument: diversifying to more expensive suppliers when there's no active crisis prices you out of the market. He's not wrong on a project-by-project basis. But the contractor who planned ahead is the only one who can deliver on time when something breaks. And something always breaks.

Supply chain resilience is a boardroom conversation now. Not a procurement afterthought.

Watch the episode: https://www.youtube.com/watch?v=prwLSmvdAQc&t=62s

The Tower of Babel, solved by accident

Alain Waha is the CTO at Buro Happold, a global engineering firm of around 3,500 people. He's had three careers, starting in chip design, then automotive, then construction from 2006. He's watched technology transform multiple industries from the inside.

His argument this week was the most interesting thing I've heard on this show in months. He said the construction industry has spent 30 years trying to solve the Tower of Babel problem, the idea that if everyone spoke the same digital language, everything would connect. It never worked. 90% of firms have fewer than 50 employees. Nobody was going to build or buy an enterprise system that integrates with a hundred tier-two subs.

And then he said: we just built Google Translate.

Large language models can read documents as they are. Drawings, contracts, specs, emails. They can cross-reference them and flag where things don't match. Without anyone digitising first. Without the industry agreeing on a single standard. Every other industry that digitised, banking, retail, manufacturing, forced a common language. Construction never could. But now there's a technology that translates between systems instead of requiring everyone to adopt the same one.

Early access to Alain's episode: https://bricks-bytes.beehiiv.com/p/silicon-is-the-new-steel-alain-waha-early-release

WSP's CEO echoed a similar theme during their Q4 earnings call, pushing back on what he called "AI hysteria" while posting a record 17.1 billion Canadian dollar backlog. AtkinsRealis posted record revenue too. Both are investing heavily in AI and both are saying the same thing: augmentation, not replacement. WSP formalised a "machine in the middle" policy keeping human oversight at both ends of every AI workflow.

https://www.constructiondive.com/news/wsp-cautions-ai-hysteria-earnings/813338/

Then Alain went somewhere more uncomfortable. He said the industry has been lazy about pricing. Cost your humans, multiply by time, add a margin. Cost-plus. That worked when everyone operated at similar productivity levels. But when AI makes some firms dramatically more productive than others, cost-plus gets squeezed hard. His analogy: you don't buy a laptop by the kilo. You buy it for the value it creates. The firms that win shift from selling hours to selling outcomes.

And underneath all of that sits what Alain called the GPS problem. He's a sailor who grew up knowing where north was by the sun. His son has always had GPS. His son has no sense of direction. That's fine for navigation because GPS is near-perfect. But AI is a probabilistic system. It gives you the most likely correct answer, not the definitely correct one. The human still needs to mark the work. And Alain's worry is that juniors who've only ever reviewed AI output, never written the first draft themselves, won't develop the judgment to catch when it's wrong. Last week I told you about a construction attorney whose client's counterparty used ChatGPT to redline a contract that argued against their own position. Nobody caught it until a human lawyer reviewed it. That's the GPS problem in action.

The knowledge moat, and the skills gap underneath it

Alain said something that reframed what I talked about last week. Every time someone leaves Buro Happold, the knowledge leaves. What he wants is for that knowledge to stay forever.

He pushed the idea further. Imagine an architecture firm that designs one school a year. After 20 years, they've got 20 schools of knowledge. But across the UK, hundreds of firms each have their own 20 schools. Nobody pools it. Collectively, the industry still doesn't really know what makes a great school.

If one firm captures 20 years of project data and makes it queryable, and their competitor doesn't, that gap compounds. Every project makes the system smarter. Every project the competitor does without capturing it just evaporates. Think compound interest for expertise. Same 25-year-old engineer, completely different capability depending on which firm they're at.

But here's the uncomfortable part. Even if you build the knowledge systems, who's going to use them well? The juniors replacing your retiring workforce have a different relationship with information. They've always had the answer available. What they may not have is the ability to evaluate whether the answer is right.

In the US, 41% of the construction workforce is expected to retire by 2031. 53% by 2036. In the UK, roughly a third. The knowledge crisis is actually two problems. Capturing what the current generation knows before they leave, which is what we covered last week. And making sure the next generation develops the judgment to use it. That second one is the problem almost nobody is talking about.

Quick hits

• Turner Construction posted 29.2 billion in revenue, a 40% jump, with data centres now 37% of their 44.3 billion backlog.
• Tutor Perini had their best year in 130 years.
• Autodesk's AECO segment grew 22%, outpacing manufacturing.
• Palantir's much-hyped construction web page went straight to a 404 error. Open invitation to come on the show. We're not holding our breath.

The look ahead

Alain's parting thought: current AI models don't understand the physical world. A three-year-old knows that pushing something off a table means it falls. No AI model has that yet. Autodesk's 200 million dollar bet on World Labs is aimed at exactly this, teaching AI to understand space, geometry, and physics. When that gets cracked, you're not just reading documents. You're reasoning about how buildings actually work.

That's a different world. And it's coming.

This article is the companion piece to the Bricks and Bytes Executive Weekly Briefing podcast. Listen to the full episode wherever you get your podcasts. Join the conversation.