144 Out of 150 Contractors Got This Wrong. Here's What It Costs Them.

Quick personal note: there's no podcast version of the briefing this week. I welcomed a new baby boy into the world this week, my third, so I'm taking a few days to enjoy the chaos...!

Normal service resumes next week and I'll recap some of this on the mic then. In the meantime, here's the written version.

I sat down with Dr Matt Stevens this week. He's spent 40 years studying what separates the contractors who survive from the ones who don't. He's worked with hundreds of firms across the US, Australia, and globally. He's written the books on it. Literally. His main reference book is 975 pages long and catalogues over 140 best practices (more on these later) for running a contracting business. And about fifteen minutes into the conversation, he said something that you should be paying attention to.

He showed a room of 150 contractors a financial calculation. A simple one. Project Return on Investment. Six of the 150 got it right. 144 got it wrong.

That's the starting point for this week's briefing. Because the number exposed something that runs through everything else we covered this week too: from how AI is quietly entering construction workflows without anyone checking the output, to why the Strait of Hormuz should be on everyone's radar right now, to what Procore and NVIDIA just announced and what it actually means for the people building things.

Here's what happened.

Most contractors are measuring the wrong thing

Every contractor knows their profit margin. It's the first number in every board pack. 2%, 3%, maybe 5% on a good job. But Matt Stevens MBA PhD FCIOB FAIB ' argument is that margin alone is dangerously incomplete. It doesn't tell you the full picture.

Here's why.

Think of it like lending money to two different friends. Both promise to pay you back with 5% on top. Friend A borrows $10,000 and pays you back in a month. You get your $10,000 back plus $500. Friend B borrows $100,000 and takes a year to pay you back. You also get 5% on top, so $5,000. Both paid you 5%. But with Friend A, you had $10,000 tied up for one month. With Friend B, you had $100,000 locked away for twelve months. Your actual return on the money you had at risk, and the time it was at risk, is completely different. One deal was brilliant. The other one quietly ate your cash flow alive and undoubtedly blocked a ton of other opportunity cost up with it.

That's what's happening on construction projects. When a contractor starts a job, they're spending their own money before the client pays them anything. Materials, labour, equipment, mobilisation. That's real cash going out the door. Maybe the first payment doesn't land for 30, 40, sometimes 60 days. And for the life of the project, there's always some amount of the contractor's own money sitting in the job, waiting to be paid back.

Profit margin ignores all of that. It just tells you the gap between your costs and your revenue at the end. Project ROI tells you how much money you had to put in, for how long, and what you actually earned on that investment.

When Matt runs this calculation across real projects, the results are obviously surprising. He showed three jobs, all with an identical 5% profit margin. One returned a 135% ROI. Another returned 47%. The third returned 12%. Same margin on paper. Completely different financial outcomes in practice.

The difference comes down to how fast the owner pays, how much capital the contractor has to front, and how long that cash sits in the job before it comes back. These aren't complicated variables. But almost nobody is tracking them. Matt says this calculation has been around since the 1960s and it's still not widely taught or used. And that means most contractors have no idea whether the job they just won is a winner or a trap until it's too late to change course.

The practical move: Before you bid your next job, calculate three things beyond your margin: what your average net cash invested in the job will be each month, how long that cash sits before it comes back, and what the annualised return looks like compared to your other active projects. If you can't do that calculation today, that's a blind spot.

Speed wins. But not how you think.

Stevens' second argument sounds obvious at first. The quick kill the slow. Not the big beat the small. Speed.

But..! he's not talking about compressing a 12-month schedule into six months. He explicitly warns against that. Schedule compression increases cost, reduces quality, and creates safety risk. People bump into each other in the same room trying to install things at the same time.

What he means is disciplined speed. Starting every job with one question in the first planning meeting: how are we going to beat the schedule and beat the budget? And then asking that question every single week until the job closes. Not in a panic. Just as a habit.

This is where his concept of "best practices" comes in, and it's worth explaining what that actually means.

A best practice, in Stevens' framework, is not a vague principle like "communicate better" or "be more efficient." It's a specific, repeatable action that a contractor can document, train people on, and measure. And the important part: a good best practice has multiple positive ripple effects. It doesn't just solve one problem. It improves several things at once.

His 975-page book catalogues over 140 of them. Here are two examples from our conversation that I think most people will recognise immediately.

The first is a bidding model: This is a historical record of your competitors' pricing behaviour across dozens of past jobs. You track the difference between your cost estimate and their submitted price, then factor in the characteristics of each job: location, project type, client type, how busy the market is. When a new job comes up, you query that data and can predict within a narrow band where the winning price will land. Instead of guessing or going on gut feel, you're pricing with evidence. Matt says it's been around since the 1960s and is still not widely used. I can vouch for that. I spent over a decade as a quantity surveyor pricing and winning projects and I'd never come across it (or any of the other calculations Matt mentions for that matter) .

The second is a material site lay down plan: Instead of letting materials get dumped wherever the delivery truck stops, you plan the physical layout of the site for every phase of the job. Month one, the rebar goes here, the formwork goes there, the excavator parks over there. Month four, the layout changes. Everything has a designated place. The payoff is direct: 70% of lost time on a job site is due to material logistics. People wandering around looking for things that got put in the wrong spot. A lay down plan solves that. It also stops subcontractors fighting over territory, which anyone who's ever run a site knows is half the battle.

The ripple effects are where it gets interesting and this is something i'd never thought of before.

A well-organised job that's running ahead of schedule retains its best people. The experienced workers, the ones with 20 years of craft knowledge, they look at a smooth job and think: this is worth staying on. They look at a chaotic one and think: I know how this ends. Screaming, overtime, weekends, guilt trips. I'm going to that job down the street instead.

So speed doesn't just protect your margin. It protects your workforce. And in an industry that's short hundreds of thousands of workers, that's not a nice-to-have.

Stevens backed this up with data from his PhD research. For every 1% of best practices that are completed on time, fully done, there is a 2% gain in project efficiency. Upper quartile contractors, the ones who follow these disciplines consistently, outperform lower quartile contractors by 300% to 1,100% in financial results. That's not opinion. That's banking data going back to 1914.

The practical move: Pick one live project this week. Ask the site team one question: are we ahead of where we planned to be at this point, or behind? If the answer is behind, ask what happened and what would need to change to get back ahead. That conversation alone is worth more than most management reports.

Tell me what you think: I'm curious.... Stevens says the difference between upper and lower quartile contractors isn't talent or luck. It's compliance to written best practices. That simple. Do you buy it? Or is that too neat for an industry this messy?

Tell me in the comments on the LinkedIn post for this article. Stevens' second argument sounds obvious at first. The quick kill the slow. Not the big beat the small. Speed.

Your AI thinks you're a genius. That's the problem.

I watched a video this week that I think every construction exec needs to see. It was about the CEO of Y Combinator, Gary Tan, who open sourced a project on GitHub with the energy of someone releasing a breakthrough piece of software. His CTO friend texted him calling it "god mode." He posted it publicly, clearly convinced it was significant. Underwhelmingly, the project was just a folder of markdown files. Prompts. Text files that tell an AI chatbot to pretend to be different people. One says "act like a CEO." One says "act like a staff engineer." That's the product... WTF.

The video's author made the point that this is happening everywhere. Someone sits down with an AI tool, describes an idea, and the AI responds with things like "brilliant approach" and "great instinct here" and "I love how you're thinking about this." It never pushes back. It never says "this isn't very good." After a few hours of a machine that sounds smarter than anyone you've ever worked with telling you that everything you produce is excellent, you start to believe it.

A recent study of 3,000 participants found that interacting with sycophantic AI chatbots makes people rate themselves as more intelligent and more competent than their peers. Another found that the more you use AI, the more you overestimate your own abilities. The power users are the most delusional.

Here's the part that matters for construction, and this is the part that surprised me the most.

The AI companies are doing this on purpose. They use a process called RLHF, reinforcement learning from human feedback. They show the model thousands of different ways to respond to you. Humans pick the responses that make users feel the best. Then they train the model to produce more of those responses. As the video put it: they are scientifically synthesising the exact sequence of words most likely to make a human feel good about themselves, and serving it on tap. And unlike ads or other forms of persuasion that we learn to tune out, this one adjusts. If the current level of flattery stops working, they retrain the model. It's a system that adapts to your tolerance automatically.

Now bring that into a construction context. A PM asks an AI copilot to draft a progress report. The AI produces something that reads well, sounds confident, and looks professional. The PM doesn't have time to cross-check every line against field conditions. They approve it and it goes into the project record. A QS uses an AI tool to summarise a contract clause. The summary sounds authoritative. It might also be subtly wrong. But because it reads so well, nobody questions it.

A Georgia Tech researcher, Max Mahdi Roozbahani, wrote about exactly this in Construction Dive this week. His warning: when AI tools hallucinate in construction, when they confidently produce false information, the output doesn't just disappear. It enters daily reports, safety logs, claim files. It gets a timestamp and a confident narrative that nobody wrote and nobody checked. And it can sit buried in the project record for years before it surfaces during a dispute.

Matt Stevens said something similar in our conversation, coming at it from a completely different direction. His line: people have been overwhelmed by AI's potential and are just admiring it rather than saying, let me take one problem and fix that problem. He surveyed the industry and found that construction software supports only about 56% of the best practices that contractors actually value. There's a 44% gap between what matters and what the tools cover. AI won't close that gap if nobody is checking whether the output actually reflects what happened on site.

The sycophancy problem and the hallucination problem are two sides of the same coin. One makes you trust the tool too much. The other makes the tool produce things that shouldn't be trusted. Put them together in an industry where a progress report can become evidence in a multi-million pound claim, and you have a real issue.

The practical move: If your team is using any AI tool to generate reports, summaries, or documentation, run a spot check this week. Take three outputs from the last month and compare them against actual field conditions. Look for confident statements that nobody verified. And pay attention to whether the tool ever pushed back on anything, or whether it agreed with everything your team asked it to do. If it never said no, that should worry you more than if it did.

This is the one that's going to b provocative but... I think most construction firms are adopting AI tools without any governance over the output. No review process, no spot checks, no one asking whether the AI ever disagreed with them. If your AI never tells you you're wrong, it's not a tool. It's a yes-man with a subscription fee. Am I being too harsh, or not harsh enough?

Hit me in the LinkedIn comments.

Quick News 1: Hormuz is back. Stress-test your fixed-price contracts.

A few weeks ago on the briefing, we flagged the situation in the Strait of Hormuz after tensions with Iran escalated. Noble Francis from the Construction Products Association had already been warning that supply chain stability is not what it was ten years ago. His message to construction boards was direct: there have been major disruptions every year for the last six years, and there will continue to be major disruptions every year.

This week, the numbers got worse. Tanker traffic through Hormuz reportedly dropped around 70%. As of writing this, oil is trading above $105 a barrel. Aluminium smelters in the Gulf have halted production. European steel prices are climbing.

If you're locked into fixed-price contracts right now, this is not background noise. This is a live cost event. Steel, cement, and freight baselines are being repriced in real time by a geopolitical situation that nobody in your procurement team can influence.

The Daily Blueprint put it well: fixed-price contracts are now geopolitical instruments. If this disruption drags on, and there's no sign it's resolving quickly, the contractors who stress-tested their force majeure clauses and diversified their sourcing three months ago will be the ones still delivering.

Quick News 2: What Procore and NVIDIA just told us about where this is all heading

Procore announced this week that it is integrating its platform with the NVIDIA Omniverse DSX Blueprint. For anyone who isn't tracking this closely, here's what it actually means in plain language.

Procore is the largest construction management software platform in the world. NVIDIA builds the chips and systems that power AI. The Omniverse is NVIDIA's platform for creating physically accurate 3D digital twins: virtual replicas of real buildings that update in real time as the actual project changes.

The integration means that project data from Procore, your RFIs, schedules, budgets, design changes, will sync directly into a live digital twin of the building under construction. The first target is AI factories and data centres, which are some of the most complex builds happening right now. A minor field change on a data centre can disrupt airflow patterns that affect GPU performance. So the ability to model a change virtually before it happens on site is not a luxury. It's a necessity.

NVIDIA is already using Procore to build its own AI factories. And joint customers like Switch, one of the largest data centre operators in the world, will use the integrated platform too.

For the broader industry, this matters because the digital thread, the idea that design, construction, and operations data should flow through one connected system, is no longer a conference slide. It's being deployed on real projects, right now, at the highest end of the market. The contractors closest to that thread will be the ones who can move fastest when the tools mature.

So what do you do with all of this?

Stevens' upper quartile contractors don't outperform because they have better software. They outperform because they know what good practice looks like, they write it down, they train to it, and they hold people accountable. Then they buy software that supports those practices. Not the other way around.

That's the order of operations. And it applies whether you're choosing a bidding model, evaluating an AI copilot, stress-testing a supply chain, or deciding whether to invest in digital twin infrastructure.

Know what good looks like first. Then automate it.

I'd also recommend checking out Matt's book which goes through all of the best practices that separate high ROI contractors to low ROI contractors. Having worked in construction for over 12 years, and as I mentioned earlier, I have never come across these insights... what do they teach the kids these days eh!? Anyway, a lesson for us all - consume knowledge and never stop learning.

Yes, I appreciate a lot of this is theoretical. But what you choose to do with these theories is up to you. Matt has evidence of these things working, but ultimately the proof is in the pudding.

That's the briefing. Enjoy the week. I'll be getting as much sleep as is possible with a newborn.