A note from Stephen: This Insider Take is brought to you by Document Crunch. Garrett's whole argument is that the emails and daily logs you write today are the exhibits somebody reads out loud three years from now. Document Crunch works the front end of that same problem, reading the contract before the job starts so the risk is on the table in week one instead of on a projector in Exhibit 12. If you manage field ops, it's worth twenty minutes of your time.

Here is what the field rarely hears from the legal side: when a project dispute reaches a jury, the decisive factor is not the contract or the schedule. It is whether the people who ran the job are believable. The lowest ranking person on your site can swing a case worth hundreds of millions of dollars, based on what he wrote three years ago and how he carries himself on the stand today. If you lead field operations, your credibility is the asset most worth protecting, and you protect it long before anyone files a claim.

Write What You See, Not What You Think It Means

The first thing I tell every client, whether a first-year PM or a thirty-year super, is the same. Observe and report. Daily reports and emails should be descriptive, not conclusory. State the facts and leave your opinion about the facts out of it.

I had a case where a project manager in his mid twenties sent five emails at the end of some long days. Those emails became Exhibit 12 at trial, and they were bad, because he stated conclusions he had no business drawing. Once they were on a screen in front of a jury, there was no walking them back.

Email is where most of the damage happens, because people slow down, think about what they want to say, and start offering opinions they are not qualified to give. In another case, a hillside gave way when a retaining wall failed and slid into the houses below. A city employee had emailed asking whether they should keep installing a barrier product they already knew was failing. That call belonged to the soils engineer on site, not him. At trial, all I had to do was put his email next to a photograph of the damage.

The lesson is not to stop sending emails. It is to send emails that describe instead of prescribe. You are not the architect, the sub, or the material supplier, and you do not have the full picture of the job. Your job is to collect information, not to rule on it. Before you hit send, ask one question: would I want this projected on a courtroom wall?

The Record You Keep Every Day Beats the One You Build After the Fact

Evidence created after a dispute begins carries almost no weight with a jury. Evidence created as a routine part of doing business, the same way every day, is treated completely differently. California calls it the business records exception, and it exists because a document written in the same boring way before a lawsuit is inherently more reliable than one written to win an argument later.

I have tried cases with years of daily reports in red spiral-bound books, admitted without a fight, because they were contemporaneous and routine. I have also watched someone try to manufacture a record after the fact, and the court threw it out because the reliability was not there.

One superintendent I represented documented who was on site and when, every day, no exceptions. When a subcontractor submitted a change order padded with hours that never happened, we sent a single letter. Our client had notes on top of notes. They had nothing, and they folded almost immediately.

Keep your daily log as if it is the most important document on the job. On the day something goes wrong, it is.

The Witness They Could Not Shake Was the One Who Was Not Performing

Documents protect you, but the jury is still deciding whether to believe a person. I once watched an expert witness walk in with the most beautiful critical path chart you have ever seen and get taken apart on cross-examination. There was simply too much surface area to attack, and his credibility wore down over a full day of testimony.

That superintendent with nothing in his hands was the opposite. He answered like a person, not someone reciting a rehearsed script, and the jury liked him for it. Juries are regular people. They can tell the difference between someone performing and someone telling the truth faster than you would think.

So write facts instead of opinions. Keep your log as though your case depends on it, because one day it might. And when it is your turn to answer questions, be the person you actually are on the job. Your credibility is the one piece of evidence the other side cannot manufacture, and it has won me more cases than any chart ever could.

-Garrett

Chapters, quotes, and more from this episode: www.constructionconversations.com

Today's Guests:

Jeff Sample: https://www.linkedin.com/in/ironmanofit/

Episode Sponsors:

Check Out Trunk Tools: https://trunktools.com/

Check Out Document Crunch: https://www.documentcrunch.com/

A note from Stephen: This recap is brought to you by Trunk Tools, AI that reduces admin work and helps builders focus on the job. Worth saying out loud, they are also one of the six systems Elliot put into the field at Cleveland, which makes this one of the few sponsor notes where you can go read what an actual customer thinks instead of taking my word for it. Elliot's bar for a vendor is usage data on a regular cadence, no sentiment. Hold them to it.

Elliot is on the operations side, not innovation, which matters. He came up as an APM, spent his first five to seven years as a project manager on site, and inherited the technology portfolio almost by accident in 2015 when his IT director pulled him into an office to look at something that integrated with Sage 300. That became a Procore pilot on his own job, an outlet mall in Columbus, and once you run the pilot you own the rollout, and once you own the rollout you own the training. It kept snowballing until it was just part of the job.

Here’s what stood out.

1. If your name is on the pitch, you don’t get to leave the room

Most executive involvement in a pilot ends at the approval and picks back up at the ninety day readout, which is a fine way to receive information you cannot evaluate. Elliot’s rule is the opposite, and his reasoning is almost boringly practical.

“If I’m gonna put my name behind it, then I’m gonna be involved in it throughout the rollout, throughout the pilot process and make sure that if our guys are saying it’s good or bad, I’m gonna make sure that I am right there with them.”

The point is not enthusiasm, it is diagnosis. If the team comes back and says the tool is bad, he wants to know whether it is actually bad or whether they never opened it, and from a conference room those two answers are indistinguishable. Being close to the work for as long as he was is what lets him tell the difference, and it is also what lets him filter the noise before it reaches the vendor, because as he put it, an RFI is processed the exact same way today as it was twenty years ago, it is just in a system now instead of a fax.

2. Put the person who will hate it on the pilot, on purpose

This is the sharpest idea in the episode and I have not heard anyone else say it out loud. Every Cleveland pilot gets seeded across multiple demographics, and two of the seats are deliberate.

“I’ll usually try to pick somebody that I know will be a detractor from it in the pilot, and then pick somebody that I know will be down the road be a champion for it.”

The champion goes in early so they are ahead of it and can train project teams later, which is the obvious half. The detractor is the half everyone gets wrong. He will list everything wrong with the product, in detail and with feeling, and that list goes back to the tech company while the pilot is still running and the roadmap is still soft. Most companies keep their skeptics as far from the pilot as they can, which is a reliable way to find out what those people think at renewal.

3. Self-reported adoption is the least reliable instrument on the jobsite

Elliot was on site recently and asked a superintendent how the new RFI tool was going. The answer sounded great. “He goes, I think we asked 12 questions in it the other day.” So he called the rep and pulled the usage for that job.

“That guy had asked zero questions.”

The twelve were real. They just belonged to everyone else on the team. It is a small story that quietly indicts a lot of pilot reporting, and Elliot’s response to it was structural rather than personal. He now makes a point of only working with partners who will hand over usage dashboards on a regular cadence, either available to him on demand or pushed from their back end, which is how a vendor actually helps a champion do the accountability work instead of leaving him to run on sentiment.

4. Straight line time savings sell themselves. Everything else needs an offset.

Some tools you can see on the demo. Submittals were that, and the reason is that we hand the least experienced person in the trailer a two thousand page spec book and a stack of product data that no two vendors format the same way, and the training program is largely getting yelled at for what not to do. After rollout, initial submissions started going in correctly more often than not, and an architect on that first job was impressed enough to recommend Cleveland to another owner. Elliot found that out from one of his own APMs, on a webinar, well after the fact.

Not everything draws that clean a line, and he is honest about which ones don’t.

“Superintendents are very skeptical people. They’re gonna ask a question and it’s gonna give them an answer and they’re still gonna check.”

Nothing is saved until the trust arrives. So OpenSpace went out with an offset instead of a promise. Keep taking progress photos, but you are no longer required to shoot a thousand pictures a week and tag every one with a description and a location, because the weekly walk is on the drawing. Then came the use cases, like pulling it up inside the sub meeting rather than adjourning everyone to go look at the wall.

5. Six platforms in eighteen months, and a speak-now meeting

The list is not small. Microsoft Project to P6, Procore to Autodesk, plus GCPay, OpenSpace, Trunk Tools and Breadcrumb, which touches essentially everyone on the operations side. Elliot had been the biggest Procore supporter in the company for ten years, so he was moving off a platform he had personally championed for a decade.

What they did before the Autodesk decision is the part worth stealing. Roughly thirty people sat through three demo calls, superintendents and estimators and project managers and back office accounting, anyone who would have to touch the system. Then an internal meeting, speak now or forever hold your peace, where Elliot stood up and explained why they were looking at all, which came down to cost, an audit and an overage fee with no credit coming back the following year when they were under. Five months to the day they will be off Procore completely, with small complaints and very little resistance.

The tell that any of it worked is not in a dashboard. It is a project team with eight months left on a job asking whether they can add the safety platform nobody was ever going to make them adopt.

Conclusion

Elliot says they are in the good for now stage on pilots, with one more he would like to run if he can get a couple of people on board. What comes after is the harder problem, which is pulling data out of all six of those systems into one warehouse so the compliance checking and the WIP reporting stop eating a week of somebody’s month.

Which is the part of this that keeps nagging at me. We spent a decade arguing about whether the tools were good enough, and Cleveland’s answer is that the tools were mostly fine and the variable was always who stayed in the room. The next round of value is sitting in what all six of those systems are quietly writing down, and that one will not be won by whoever buys the best product either.

Today's Guests:

Clyde King: https://www.linkedin.com/in/clyde-king-6a95a49b/

Episode Sponsors:

Check Out Trunk Tools: https://trunktools.com/

Check Out Document Crunch: https://www.documentcrunch.com/

Welcome to a special edition of Construction Briefs.

Ever since the ChatGPT moment in 2022 I have been in and out of a lot of AI tools, seeing what they do and where they actually help. I have run Claude for 18 months, tried a couple of different OpenClaw (or whatever the current name is) instances on a VPS, and built a stack of skills and agents that drive real work, the podcast and this newsletter included.

Buzz is the first thing since 2022 that changed how I work rather than what I could produce.

I built the whole business out inside it in two days. Channels for the podcast, the newsletter, and a couple of the consulting things I have running on the side. It pulled in my Claude skills and my markdown files and made them useful in a way they had not been before.

If you have not come across it, Buzz is a new open source workspace from Block, the company behind Square and Cash App. It looks and feels like a modern team chat tool, channels and threads and messages, except the agents live inside it as participants rather than a chatbot you summon. They hold their own identities and permissions and work in the thread next to you. It is model agnostic, so you point it at whatever you already run, Claude Code or your own agents, and it sits on open protocols instead of one vendor’s rails. Free, and you can host it yourself.

I should say how it got built, because it’s relevant. I did not write agent descriptions, or set up channels, or hand-feed context into anything. I opened Claude Code, told it what I wanted Buzz to do for my business, and answered a handful of clarifying questions and it built the whole setup. Channels, agents, the context they needed, all of it. Claude Code was the implementation lead, and the part of new software that usually eats a month of evenings took an afternoon of conversation.

The moment it clicked for me was small. I called an agent to draft an episode recap. It did the thing my Claude skills already do, the recap came back in my format, fine. But it happened inside a channel I can drop my production assistant into. So they can open the same channel, see where the agent and I left off, and take it the rest of the way. They finish the publication, add the header images, and do it on their own clock while my agent keeps working on mine. Two humans and their agents working the same thread, running in parallel.

To feel why that matters you have to know how we did it before.

We run on Google Chat for the async back and forth. Alongside that I have been trying to hand over my Claude skills, the ones that encode how a recap gets written or how a post is supposed to sound. And every time I improve a skill on my end, I have to remember to pass it along, and it has to get updated on the other end, and somewhere in that relay the skill quietly drifts. The two copies stop being the same thing. It shows up in the worst possible place, which is my own voice coming back not quite sounding like me in an article, or the two of us simply not being on the same page about how something gets made.

In Buzz there is nothing to relay. The skill lives in one place and we both work off it. The drift problem does not get solved so much as it stops existing.

The part I did not expect is how different it feels to work this way. My skills stopped feeling like functions I invoke and started feeling like agents. It is closer to an open, agentic workspace than a chat window. When I called that recap, I did not first have to decide whether this was a job for Claude chat or cowork or code, or whether I was prompting it right. The skill came to me, in the channel where my work already lives, in the format where my team already is. I was just talking to something that felt like a coworker.

That is the whole thing, really. The tool met my workflow instead of making me go find it.

I want to pull this up a level, because the reason I am writing it for you and not just posting a screenshot is the open source piece underneath it.

I sell digital products and services into general contractors, so I am not a neutral party here. But the more time I spend in and around this industry’s software the more convinced I am that the value is in owning your own data and your own methods, not renting them from a walled garden. Procore and the platforms like it captured the industry’s information and then rented it back. An open workspace you shape around your own skills and files is a different arrangement. You keep the data and the methodology, and the AI works on top of what is already yours.

I would not tell you this is finished. There is real risk review to do before you run anything sensitive through it, and I am doing that work myself. But it is ready enough to have your hands on now.

And the businesses positioned to move first are not the giants. It is the small and regional contractors, the ones nimble enough to experiment and agile enough to change how they work in a quarter instead of a fiscal year. A workspace that bends to any business is worth more to a nimble regional builder than to one locked into a platform contract it cannot leave.

The timing matters more than it did two years ago. Every platform in this industry is racing to put AI on top of the data it already holds, and the data it holds is yours. Your job history, your production rates, your estimating logic, the way your people actually build. That material is about to be the most valuable thing a contractor owns, and most contractors are handing it over one integration at a time, then paying to ask questions of it.

You do not have to accept that arrangement. You can put your own methods into a workspace you control, point whatever models you want at them, and keep the thing that makes your company good at building inside your own walls.

That is why a weekend project is worth writing to you about. Not because the software is polished, but because it is the first time I have seen the alternative actually work.

Oh and did I say that its FREE!?

-Stephen

Today's Guests:

Andrew Rener: https://www.linkedin.com/in/andrew-rener-phd-pe-dbia-f-asce-956821a/

Justin Ullman: https://www.linkedin.com/in/justinullman/

Episode Sponsors:

Check Out Trunk Tools: https://trunktools.com/

Once a month, I sit down with some of the brightest minds in the industry and argue through the construction news that actually mattered. The episode is the conversation. This is the analysis.

The month in one line: the platforms are buying the inputs. Procore paid $845 million in cash for DroneDeploy and its 20 trillion square feet of jobsite imagery the week we recorded. New York decided which data centers get to exist by rationing permits. Suffolk decided it would rather build its own intelligence layer than rent one. Even the market data got contested, with the industry’s two starts trackers describing the same month in opposite directions. Every story this month is a fight over who controls the raw material, whether that raw material is visual data, megawatts, workflows, or the numbers themselves.

This briefing synthesizes the month’s construction and construction tech news with practitioner commentary from the July Reviewed as Noted roundtable. of Fieldproof, who advises contractors on AI and data strategy, returns. Joining him this month: @AJ Waters, four weeks into a new role at Procore, and , recently off the GC side, who reads deals through a capital allocation lens.

1. Procore Buys the Input Device: $845 Million in Cash for DroneDeploy

The news: On July 30, Procore announced it will acquire DroneDeploy, the reality capture and robotics platform, for $845 million in cash, expected to close later in 2026. DroneDeploy operates in more than 180 countries and holds roughly 20 trillion square feet of visual jobsite data, tens of millions of user annotations, and a record of more than 100,000 labeled safety issues. Procore brings nearly 400 million photos, more than 126 million drawings, and over 10 million RFIs, submittals, and inspections recorded annually. The deal follows Procore’s January acquisition of Datagrid and the July 23 launch of its Digital Coworker agent packages, making 2026 the year Procore assembled an AI stack piece by piece: first the agent team, then the agent catalog, now the capture layer that feeds it.

Why it matters: Procore has been explicit that it wants to be a system of action, not just a system of record, and an agent can only act on what it can see. Everyone at the table opened this year saying the same thing about jobsite data capture: the input device is still broken. Buying DroneDeploy is buying an automated input method, plus the orchestration layer for drones and, eventually, ground robotics. It also lands ten months after Procore’s API terms began prohibiting marketplace partners from bulk downloading platform data for commercial purposes, including training large language models. Read together, the sequence is hard to miss: the terms fenced the data in, and the acquisitions are stocking the fenced area.

From the roundtable: The table’s first drone stories date to 2017 and 2018, a dam project measuring earthwork and a data center too big to photograph on foot. Eight years later that capture layer is an $845 million strategic asset, and the panel split on what the money actually bought. Victor’s read is capabilities and pole position: Autodesk has reached rough product parity, the big platforms are buying net new modalities rather than doing R&D, and against robotics startups raising hundreds of millions at seed, “the 845 number is actually not that crazy” from a venture math perspective. He also flagged the sleeper asset: training data for robotics, the kind of business Scale AI built for the LLM companies, sitting on top of data Procore already has. Tyler read the price itself as the tell: “I look at that as DroneDeploy probably just sold a lot of data on a lot of different job sites to Procore.” Put in the seat of a $300 million GC, his response to the deal was blunt: “I’m looking at my Procore bill every year, and I’m going, how much of this data am I putting into other people’s hands?” His answer is to pull data back internally and build an ontology layer that reads his whole business. AJ, corporate hat removed, defended the acquisition on context rather than code: “I can go build an app in my basement in four weeks, but if I’ve never set foot on a job site, I still have no context. AI is a tool, just like an arc welder is a tool.” He then put the hard question to Tyler’s build-it-yourself plan: is your data actually more secure in a homegrown stack with no security team, and pointed to the retail mega-breach that famously came in through a contractor. Nobody at the table fully answered that one.

2. New York Freezes Hyperscale Permits, and Offers the Trades a Deal

The news: Governor Hochul signed Executive Order 62 on July 14, making New York the first state to pause permitting for data centers capable of drawing 50 megawatts or more while the state prepares a Generic Environmental Impact Statement. The order sets no fixed end date; expiry is tied to the GEIS, whatever the “up to one year” framing in the announcement says. Roughly 12 gigawatts of data center load sits in the state’s interconnection queue, more than 8 gigawatts of it added in 2025 alone. The buried lede is the labor trade: the order directs a Community Investment Framework prioritizing prevailing wage, project labor agreements, local hire, and apprenticeships. Two days later, EPA moved the opposite direction, telling developers that islanded on-site power plants with no grid connection fall outside the Acid Rain Program. One arm of government slowed the grid path while another cleared the build-your-own-plant path, in the same week.

Why it matters: The stakes are quantified in AIA’s July 20 forecast: data center construction spending up 33% in 2026 and 25% in 2027, roughly 8% of all nonresidential building spending by 2027. Strip data centers out and the commercial forecast becomes a 1% decrease this year. A single asset class is the entire difference between a sector that grows and a sector that contracts, which is exactly why a single state’s permit action is now a national construction story. And because a moratorium is not a cancellation, the pause quietly consolidates the pipeline toward the developers with the deepest balance sheets, the ones who can wait out a GEIS or build their own power and skip the queue entirely.

From the roundtable: Nobody at the table read this as a win for the industry. AJ’s scorecard was that the politicians won and everyone else lost, at least at this juncture: the jobs inside these facilities are undercounted, the public underestimates how much of their daily life runs through them, and the prevailing wage provision, whatever its merits, means data centers will pull even harder on labor markets that were already losing people to them. Victor found the winners out of state: “The winners are the other states. If you’re trying to build a data center in New York and now you can go to Pennsylvania, why would you not?” As an owner, the regulatory headache is itself a site selection criterion. My own addition: whatever delay New York thinks it is imposing, triple it, because the labor that leaves for neighboring states during the pause does not come back on the state’s schedule.

3. Suffolk Embeds AI Engineers Inside Project Teams

The news: Suffolk is placing AI engineers directly inside project teams on mission-critical, healthcare, and higher-education work, running Claude and ChatGPT against RFI creation, change order documentation, procurement integration, and executive dashboards. The strategy is explicit: buy standard applications, build only bespoke tools, because Suffolk is “a construction company,” not a software firm. It is the strongest adoption signal of the month from a top-tier contractor, and it is also, roughly a year into the program, entirely unquantified. Every public statement remains forward-looking. Worth noting: Suffolk is one of the GCs named as a Trunk Tools user cut off when Procore revoked API access last fall, which makes the decision to build in-house read a little differently.

Why it matters: This is construction’s version of the forward deployed engineer, and the org design is the actual news. An engineer sitting in the trailer, hearing the real complaint at the real moment, is a structural choice any mid-size GC can imitate at a fraction of Suffolk’s cost. The model choice is the easy part and changes every six months. The discipline question is the hard part, and it is the one the panel spent twenty minutes fighting about.

From the roundtable: This was the sharpest disagreement of the episode. Tyler, who builds AI systems for contractors, came out skeptical: most of what he sees are process problems wearing AI costumes. He described a client losing an estimated $600K in revenue capacity because two tools were never connected with an out-of-the-box connector, no AI required. “If I throw AI at a crappy process, I’m getting crap faster.” He also warned where undisciplined mandates end: six-figure token bills with no verifiable return. Victor pushed back hard with the R&D math: “If you just imagine there’s five people there, 200K a head, so that’s a million bucks, double that for travel and expenses, that’s two million. If they find any sort of operational efficiency across the five billion that they do, that’s 100 percent worth it. You’ll do that trade every single day.” His position: chalk it up as R&D, scale it to ten people, and stop demanding attribution at the individual level, because the ROI shows up at the org level when a person can cover three jobs instead of one. AJ split the difference: he likes that Suffolk put the engineers where the work happens, and worries about the two things the press release skips, what these people cost and what happens when every project insists on doing things its own way. I land closer to Victor. Having lived at the project level, the zero-to-sixty on value is not hard to imagine if the people are senior enough, and Suffolk runs a tight enough ship that I would invest in this experiment. But Tyler’s closing caution deserves to hang over every one of these programs, and we will come back to it in the Outlook.

4. Two Data Vendors Described June in Opposite Directions

The news: On July 21, Dodge reported June construction starts down 19.9% to $1.42 trillion annualized, with nonresidential building down 9.1% and nonbuilding down 37.7%. Six days later, ConstructConnect reported June nonresidential building starts at a record $100.3 billion, up 32.8% from May and 35% above the twelve-month moving average, with offices and data centers accounting for more than a third of it. Same month, same country, opposite conclusions, and because there is no government series for construction starts, there is no referee. Both are proprietary trackers with different capture rules, seasonal adjustment, and scope definitions.

Why it matters: The reconciliation is probably scope, not scandal. Dodge’s headline includes nonbuilding infrastructure, which cratered, while ConstructConnect’s record is nonresidential building only. Both can be arithmetically true at once, which is somehow worse than one of them being wrong. The practical rule: name the vendor every time you cite a starts number, and when the headline and your own backlog disagree, trust your backlog. ConstructConnect’s own caveat, that the gains were “sharply concentrated rather than broad-based,” is the month’s thesis restated in vendor language.

From the roundtable: The ground truth at the table was unanimous and it was not Dodge’s. Tyler’s market read from his contractor clients: blowing up, people cannot keep up. Victor, from Ohio: data centers, manufacturing, everything going up at once, nobody can find help, boomtown. AJ has made a habit of interrogating the gloomy headlines: “Every time I see a headline that says construction starts are down, I instantly try to go figure out where... because it sure doesn’t feel like it anywhere.” His supporting evidence reached beyond data centers, pointing to Dulles airport’s newly announced $20 billion capital program as proof the big numbers are not one asset class. Tyler’s working theory for the split is project mix, that the two trackers are seeing different slices of the market flow through their systems, which is as good an explanation as any in a world where the industry’s picture of itself is assembled from vendor releases nobody audits.

Lightning Round: The Next Check

We closed by calling shots on the next big construction tech acquisition before year-end. My pick: one of the emerging scheduling and planning platforms goes, because it is past time Oracle faced a true challenger and the big platforms need a planning layer to feed scenario planning and the orchestration story. Victor agreed pre-con is the weakest flank of the big suites, then picked ClearStory: universally liked, real revenue, the kind of acquisition nobody argues with, assuming the founder sells. Tyler called his shot with more confidence than evidence, by his own admission: Trunk Tools gets acquired by a construction company, not a software company, because a large GC building its own ontology and agents gets a running start, and the platform fight has left Trunk Tools without a natural home among the majors. AJ abstained on specifics, week four and all, but confirmed the rumor mill is active and put a number on it: at least one, maybe two more major acquisitions before the year is out.

Outlook

The gap to watch is between who generates construction’s data and who gets paid for it. Contractors produced the 400 million photos, the 126 million drawings, and the 20 trillion square feet of imagery that made this month’s $845 million deal make sense, and the month’s biggest strategic moves, Procore’s acquisition, Suffolk’s build-in-house bet, Tyler’s clients pulling data behind their own walls, are all answers to the same question: is the data a contractor generates on its own jobsite an asset it controls or an input it donates?

Full Episode Here:

Check Out Trunk Tools: https://trunktools.com/

So when I say I don’t believe the numbers, I’m not talking about the tools themselves. I’m talking about the story sold around them, the promise of hours saved, and how rarely it survives contact with a complex project.

Tools like OpenSpace genuinely give us time. I don’t need to walk the site every day to know what changed last week. I still should, but when I don’t have the opportunity, I can go through the capture quickly and see what’s different. That’s real value, and nearly everything in the stack is aimed at the same target, giving us more time.

But in the same exact timeframe that all of this has been happening, projects have become more complex. Regulatory environments are heavier. Clients are more sophisticated, which is good and bad, because sophistication comes with requirements, and satisfying every stakeholder creates more administrative drag. The tasks the tools took off my plate got replaced by tasks that didn’t exist before, or by things that used to get compromised on and no longer are.

The time the tools give back gets eaten

It’s a bit of a tug of war. The hours came back, and the job grew to absorb them. A pitch about time savings that doesn’t account for that is describing a project that doesn’t exist.

I know the slide. The platform saves the average user ten hours a week. I don’t know what to do with that number. Those are probably hours 41 to 50. Or 61 to 70.

And that number lands on people who have heard it before. The field teams you’re pitching have been sold the future of construction seven times in the last five years. The real fight isn’t market noise, it’s that accumulated skepticism, and it gets won back one project at a time. People feel like they need to sell the world to get in the door, but if the world is what you sold, disappointment is the only place the relationship can go.

The way through isn’t a better version of the same promise. The pitches that get traction with me show how a tool slots into the work already happening and makes it better, instead of trying to overhaul a process that, imperfect as it is, is the only thing keeping the project moving.

You can’t out-pitch a four-year feedback loop

The other number I’m wary of is traction. Used on this many projects, this much success, this much time saved. I’m always willing to hear people out, but how long have you been doing this to know that for sure?

Most of our projects run two to four years. When we pilot something, we’re piloting across twenty jobsites, and getting real success measures and metrics out of that takes a long time. Meanwhile there’s a new AEC startup every 30 seconds. The market is generating claims faster than any project cycle can verify them.

There’s also the money. We run very small margins against the liability of the entire contract, so in almost all cases a tool has to show it’s a net value to the project before anyone pays for it. And we try to stay consistent, the same approach whether the job is in Miami, Tampa, Sacramento, New York, or Boston. We don’t swap tools often, and even the changes we do make feel like a lot to the people living with them.

None of that means stop building for this industry. It means the feedback loop you’re selling into is longer than your funding cycle, and buyers like me have learned to be wary of certainty that shows up faster than the proof possibly could.

The more I learn about this stuff, the less I realize I know, and I’m always willing to hear people out. Just bring me the version of the pitch that describes my actual week. That’s the conversation I’ll make time for.

John Riggs is a Senior Project Manager at Suffolk Construction.

Today's Guest: Hugh Seaton, Managing Director, Construction | CEO Advisor on AI, Risk, and Decision Governance in the Built Environment

Check Him Out on Linkedin: https://www.linkedin.com/in/hughseaton/

Episode Sponsors: Check Out Trunk Tools: https://trunktools.com/

Welcome to this week’s Construction Brief.

Where I go through one thing thats been on my mind, a software or piece of tech to watch and a field note on a trend I am seeing from my interactions.

As I was setting my new container office in my meadow this week, I had the thought about how Construction is the worlds reminder of it’s own physicality. Software and AI tends to remove the friction of thought and experimentation. With a few clicks or intimations of your voice new digital worlds can be created and yet moving an 8’ x 20’ container 40 feet from point A to point B requires 4 grown men (a large crane) and a couple of hours.

That friction and work also brings appreciation and purpose in a way that limitless “achievement” without it does not. At the end of those two hours, that group of dudes (myself included) were able to look back at a changed landscape where we had a small hand in putting some element of order back into the world where it wasn’t previously.

This has me thinking about all of the “mundane” work we keep talking about removing from our day to day and if it might actually be removing our purpose along the way.

On My Mind - The Case for Drudgery

The strongest evidence for that suspicion is also the darkest. One of the more unsettling findings from the forced labor camps of WWII had nothing to do with how hard the work was. Prisoners put on brutal but purposeful jobs, building barracks, cutting roads, things that existed when the day was over, tended to be able to hold it together mentally. The ones ordered to haul sand from one side of the yard to the other and then haul it back, day after day, for no reason anyone would give them, came apart. The guards understood this well enough to use it deliberately. Pointless work wasn’t an inefficiency in the system, it was the punishment.

Meaning, not effort, was the variable. Which is worth sitting with as the industry lines up behind the promise of AI removing the mundane without a great idea of what its supposed to replaced with. It’s on every slide at every conference, and it’s in most of my own decks too, so this isn’t a stone thrown from outside the glass house. But mundane is not the same as meaningless, and the two are getting used interchangeably in every pitch.

Some of construction’s most mundane work is quietly load-bearing. The young estimator grinding through takeoffs is learning to read a building. The engineer processing submittals is learning how a thousand products actually fit together. The PM writing the daily log is learning sequence, and the field engineer chasing an RFI through three consultants is learning where projects actually break down. None of it is glamorous, and none of it was ever pointless. It’s the apprenticeship, disguised as drudgery. Are there ways to replace those learning events with a more efficient system? Yes but that requires an understanding of our people, workflows and have mapped the ontology of domain knowledge to enable systematic delivery.

And plenty of what we do genuinely is worth removing. Entering the same data into three systems because none of them talk to each other. Status reports assembled for meetings where nobody reads them. Chasing signatures on documents everyone has already agreed to. Hand all of that to the machines gladly, and don’t look back.

The hard part is that the work ahead isn’t sorting automatable from non-automatable, every vendor will happily do that sorting for you, and their list will be suspiciously close to whatever their product does. The real sorting is between the work that builds your people and the work that was only ever carrying water, and nobody selling you software has any idea which is which inside your organization. That one is on you.

People can endure a staggering amount of work as long as they can stand back at the end and see that something changed. Strip out the wrong mundane and what’s left will certainly be easier. Just don’t assume it still builds anything.

Tech to Watch - RhinoDox

Summary
Having just spent the top of this issue arguing you should be intentional about what you automate, let me be equally clear that reading a 10,000-page bid package at 11pm the night before the bid is due does not make the keep list. That’s the pile RhinoDox is aimed at. Specialty contractors get buried under specs, contracts, addenda, and safety manuals, and the risk that sinks the job is usually sitting on page 847 where nobody had time to look. Their AI engine, which they named LARRY, reads the whole package and points you to what you’d have missed before you price the work.

What it is
RhinoDox is an AI bid intelligence platform for specialty contractors, built around an analysis engine called LARRY that reads full bid packages and flags project risk before you commit to bid.

Why it matters

LARRY claims to read up to 10,000 pages in about four minutes, running 100s of criteria across eight risk categories. Each one encodes something the trades already know goes wrong: go/no-go disqualifiers like project labor agreements, commercial terms like retention held past substantial completion, contradictions buried in the specs, non-scope cost drivers like a required full-time safety rep, prevailing wage compliance, schedule feasibility, project-type requirements, and even the GC’s track record from your own bid history. Findings point back to the page, section, and spec number they came from, which matters more than the speed, because an AI risk flag you can’t verify is just another opinion in the room.

It also works the other direction. Once your response is drafted, LARRY reads it back against the spec and shows you where what you wrote and what they asked for diverge — not just what the package says, but whether what you wrote actually answers it.

The company says it has processed $25+ billion in bids and offers an Autodesk Building Connected Bid Board Pro Integration

Who it’s for

Specialty contractors and subs. Specifically the estimators and precon leads drowning in bid documents on deadline, and the principals making go/no-go calls without time to read what they’re signing up for.

Check it out

https://www.rhinodox.com

Field Notes - Ontology Is a Data Strategy in a Tuxedo

Three separate conversations this month, none of them connected to each other, and the word “ontology” came up in all three. That is not a word construction people used a year ago, and the reason everyone is using it now is Palantir. Since the McCarthy partnership announcement in June, Foundry implementations are spinning out of contractors, startups are building knowledge graphs out of contract documents, and every enterprise pitch deck suddenly has a slide with nouns connected by arrows.

So here’s the demystification. An ontology is the nouns and verbs of your business, your crews, trucks, bids, and change orders defined once as data objects, plus how they relate to each other. Which is a fancy way of saying you need a data strategy. The word is new. The homework is not.

The part getting skipped in the excitement is that your ontology should map to your competitive advantage, not to a vendor’s template. If you win work on self-perform labor productivity, your labor data is the crown jewel and deserves the definition work first. If you win on prefab, it’s your assemblies. A generic ontology produces a generic advantage, which is to say none.

And you don’t need Palantir money to start:

1. Write down the ten nouns your business actually runs on (crews, equipment, estimates, change orders, whatever wins you work) and note where each one lives today. For most contractors that’s three systems and a spreadsheet nobody admits to.

2. Pick the one noun closest to why clients hire you, and give it one definition, one ID, and one owner. Just the one. Resist the urge to boil the ocean in month one.

3. From here forward, make “can I get my data out cleanly” a purchase requirement for any new software. The ontology era will punish tools that hold your nouns hostage.

The AI everyone is being promised runs on this homework. The contractors doing it quietly now will look like geniuses in three years, and the rest will be buying an ontology off the shelf, which is another way of saying they’ll be renting someone else’s understanding of their own business.

See ya in a few weeks.

-Stephen

Today's Guest:

Tim Smith: https://www.linkedin.com/in/tim-smith-63693a300/

Episode Sponsors:

Check Out Trunk Tools: https://trunktools.com/

Check Out Document Crunch: https://www.documentcrunch.com/

Angelo is a mechanical engineer who will tell you he didn’t choose construction, construction chose him. He graduated into the late-2000s recession, couldn’t find consulting work, and took a project coordinator job at an MEP firm in eastern Canada to pay the bills. Twenty years later he’s Director of Operations at Mattina Mechanical, an Ontario MEP contractor, with stops on the owner side and the GC side in between. He wrote a book called The Human Side of Construction, and that title has become the frame for everything he does: the argument that the industry’s hardest problems are people problems wearing technical costumes.

Here’s what stood out from our conversation.

1. You Need Five New Hires to Replace One Retiree

The workforce math starts familiar and gets worse. Roughly 10 million people work in US construction, and Angelo cites reports putting 30 to 40 percent of them at retirement age within five to ten years, call it 400,000 people a year walking out the door. About 250,000 new apprentices start each year, and he pegs completion at 30 to 35 percent. Federal apprenticeship data suggests construction completions run closer to 25 percent, so if anything he’s being generous to the industry.

But the head count isn’t what keeps him up at night. The gap analysis everyone publishes assumes one new entrant replaces one person leaving, and Angelo doesn’t buy it. “That person with 30 years experience, they’re going to work more productive. They’re going to solve problems faster. They’re going to know what to do. And that person coming in, you’re probably going to need five of those to replace just the productivity of the person leaving.”

Which reframes the whole crisis. “The real crux of the problem is the quality of information, getting that asset knowledge that’s out of people’s heads into some sort of system or some mentorship framework or whatever to help bring these younger people up to speed faster.”

He isn’t waiting for someone else to build that system either. After a Toronto tech incubator passed on his mentorship platform idea, he put 40 to 60 evening hours into Lovable and shipped a working prototype anyway.

2. People Want to Be Led, Not Managed

Angelo has worked at a six-person shop where he chased receivables for the owner, and at corporations big enough to have layers between him and anyone with a vision. His read on what changes with scale is blunt. Past a certain size, companies get run by lawyers and accountants, because that’s what a business entity that size requires. The structure isn’t the villain. What the structure does to people is.

“When people get lumped into cogs in a machine and just following a process and they become just managed, not led, that’s where things can kind of break down.”

His example of management at its worst is precise: “nobody wants to be told, hey, you know, Steven, you missed step six in this 10 step process, even though the output was correct.” People want to be in touch with the person pulling the strings. They want to be part of a vision, not a workflow.

The distance compounds it. The farther corporate gets from the boots on the ground, the harder it is to understand what those people need, and Angelo calls avoiding that disconnect the biggest challenge facing large construction companies.

3. AI Fixes the Quantity Problem and Whiffs on Quality

Angelo isn’t an AI skeptic. He’s just specific about which of construction’s two labor problems it actually addresses. Fewer workers processing more information, AI helps with that today. Agentic workflows can issue RFIs and hunt drawing discrepancies fast, but “if you don’t prompt it, if you don’t start that process, it doesn’t know what to do.” The tools amplify people who already know what they’re looking for.

“Tech definitely plays a role in the quantity, because as the demand grows and the supply is still limited, tech’s gonna fill that gap. But it’s missing that element of quality, I think.”

The quality he means is the tacit judgment a 30-year super carries, the stuff that lives in lunchroom conversations and never makes it into any system. Capturing and transferring that knowledge is something he says the current tools haven’t touched, which connects the AI hype cycle right back to the retirement math. The technology that’s supposed to offset the exodus can’t yet hold what the people leaving actually know.

4. The Technology Is the Specialist Now

For a generation the career advice was to specialize, and Angelo thinks AI quietly flipped it. He’s a systems thinker by trade. On a cooling system you’ve got a chiller, a cooling tower, fan coils, and what always fascinated him wasn’t how the chiller works, it’s how the pieces fit together. That instinct is now a career strategy.

“The technology is the specialist now. It is, but it’s only providing us the puzzle pieces. It’s up to us, where I see the future going, to put those pieces together in the right configuration to make the full picture.”

The corollary is that domain knowledge just got more valuable, not less. “We need builders, Steven. Anybody now can use ChatGPT to answer questions, but the information isn’t available in models, at least what I’ve seen, to solve our construction problems alone.” Understanding how trades interact on site, how jobs actually come together, the logistics nobody writes down. None of that is getting taken over by a prompt anytime soon, and the person who pairs it with the tools holds the winning hand.

5. Mental Health Hits the Bottom Line, and Nobody Has Quantified It

Angelo brought heavy numbers to this part of the conversation, so I checked them against the latest data. CPWR and NABTU’s most recent release counts roughly 5,000 construction worker suicides and 11,300 overdose deaths in 2024. Both are declining, overdoses sharply, and both still dwarf jobsite fatalities. CDC data has construction workers several times more likely to die by suicide than from a jobsite injury.

His critique of the standard corporate response is that it’s a checkbox. “Companies, I’ve seen a lot of people have like an EAP, right? Employee Assistance Program. To me, that’s just to check the box.” Published utilization studies back him up, traditional EAP engagement runs about 3 to 8 percent of eligible employees. His hunch on the rest: struggling in silence.

Then he does something most human-side advocates won’t do. He makes the ROI case. A worker who isn’t mentally present is a safety risk to himself and everyone around him, lost-time injuries hit your TRIR, and TRIR hits your insurance premiums and your eligibility with safety-conscious owners. “It hits the bottom line. And right now that’s not quantified... what’s the cost of not doing something? That’s what you got to think.”

His line on responsibility is careful, and worth keeping. Leaders can’t fix everybody’s problems, but “we have a responsibility as leaders and good people to give people the tools they need to fix their problems, not solve them for them.”

The Landing

Angelo opened the conversation with the lesson from his first week in Robert’s truck, and when I asked what he’d tell his younger self, he closed with the same one. “You don’t know everything, but neither does the guy next to you.” It sounds like a poster until you notice it’s the argument underneath everything else he said. Knowledge transfer, mentorship, leading instead of managing, asking the question you’re afraid sounds stupid.

When I asked what keeps him going across the day job, the books, and the platform, he didn’t hesitate. “I want to help other people learn. I want to make their lives better. I want to make their careers better.” This industry has plenty of people building tools. It has far fewer thinking hard about the humans who have to use them, and the retirement math says we need both.

You can find Angelo on LinkedIn, or at angelosuntres.com, where the guides alone are worth the visit.

Today's Guest:

Libby Magliolo: https://www.linkedin.com/in/libbymagliolo/

Episode Sponsors:

Check Out Trunk Tools: https://trunktools.com/

Check Out Document Crunch: https://www.documentcrunch.com/

You’re reading the first Executive Briefing, the newest piece of Construction Briefs. Every story that moved construction and construction tech this month, what the numbers say, what it means for the people building, and what three practitioners closest to the work said about it on the Reviewed as Noted roundtable.

The month in one line: a soft market with a hot tech layer. Single-family demand is stalling and heat enforcement is tightening while venture capital and AI keep pouring onto the jobsite, and the industry’s biggest demand driver, data centers, is now getting blocked before it breaks ground.

This briefing synthesizes the month’s construction and construction tech news with practitioner commentary from the June Reviewed as Noted roundtable: Karl Sorensen, who oversees construction technology innovation at DAVIS Construction; Tyler Campbell of Fieldproof, who advises contractors on AI and data strategy; and Matt Graves, who works on the trades side and advises construction tech startups.

1. The Toolbelt Generation Finally Shows Up

The news: After a decade of workforce hand-wringing, the demand signal flipped. Construction-trades program enrollment rose 23%, vocational community college enrollment climbed 16% to its highest level since tracking began in 2018, apprenticeships grew more than 11%, and roughly 42% of Gen Z is working in or pursuing skilled-trade work. The commonly cited driver is cost: a four-year degree now tops $100,000 while trade training finishes in under a year and leads straight to a job.

Why it matters: This is the first genuine supply-side good news on labor in years, but enrollment is not retention. The industry has historically lost a large share of new entrants in their first few years, so the surge only becomes a workforce if the first job is good enough to keep them. The recruiting window is open now, and it favors contractors who build trade school and university relationships this year.

From the roundtable: The group added a driver the enrollment studies don’t capture: AI anxiety is doing the recruiting. Matt argued that students are watching entry-level white-collar work get automated and concluding that hands-on work has a longer shelf life. Tyler, who taught a project management class at UGA this year, watched students realize the assistant PM role they trained for changed underneath them before graduation: “I’m worried the kids are getting bad advice about what a good future career looks like. The reality is that shifted even over three years.” The hiring side confirms the surge is real. Karl reported DAVIS’s internship program is up 53%, supported by course-credit programs with Maryland, Virginia Tech, and George Mason that place students inside live operations, and noted his screening filter is problem-solving ability, not AI fluency, since the tooling changes faster than the fundamentals.

2. Builder Confidence Is Stuck at Foreclosure-Crisis Levels

The news: The NAHB/Wells Fargo Housing Market Index fell to 35 in June, the 14th consecutive month below 40, a streak last seen during the 2011 to 2012 foreclosure crisis. Buyer traffic sat at a deeply soft 25. Builders are buying demand: 35% cut prices in June (average reduction 6%) and 62% used sales incentives, the 15th straight month at or above 60%.

Why it matters: This is a grind, not a crash, and a grind is arguably more corrosive. When residential stalls, trades and suppliers chase commercial and infrastructure work, tightening bid lists everywhere. The open question is whether the problem is the price of the house or the price of the mortgage, because construction can only fix one of those. In a market where builders cut 6% and traffic still doesn’t move, reliability and predictable delivery become the differentiators that win work.

3. OSHA Is Enforcing a Heat Rule That Doesn’t Exist

The news: The federal heat standard remains stalled and is unlikely to be finalized under this administration, but on April 10 OSHA renewed its National Emphasis Program on heat hazards for five years, through 2031. Contractors face guaranteed heat inspections with no binding standard, and citations flow through the General Duty Clause. The draft rule’s triggers, 80°F initial and 90°F high-heat, would require written plans, rest breaks, water, and acclimatization protocols.

Why it matters: Certainty about enforcement, uncertainty about the rule. A written heat plan with scheduled water-rest-shade and an acclimatization ramp is cheap, defensible insurance, and “we didn’t have a rule to follow” is not a defense under the General Duty Clause. Multi-state contractors are already navigating a patchwork, since California, Washington, Oregon, and Minnesota run their own standards.

From the roundtable: Karl’s field detail is instructive: DAVIS already runs project-specific heat stress plans keyed to heat index, temperature plus humidity, which can be stricter than the draft rule’s dry-temperature triggers. His concern was structural, not operational: “It is very curious to see a governing body wield the penalties of breaking something that is not quite a rule yet.” Tyler flagged that wearable vendors are using the enforcement wave as their wedge onto the jobsite, positioning monitoring as automated compliance documentation. Matt followed the money and asked who is lobbying for mandated monitoring, with the group landing on insurers as the likeliest beneficiary. His caution connects this story back to the workforce surge: “You want to keep kids in the trades? Yeah, go ahead and put a dog collar on and track everything.” Monitoring tech may produce a good-faith record, but surveillance framing carries a real retention cost.

4. The Money Is Back, and It Wants AI on the Jobsite

The news: Six contech startups raised a combined $126 million early in 2026: Fyld’s $41 million Series B for AI that turns jobsite video into safety and risk insight, Sensera’s $27 million for site monitoring, and XBuild’s $19 million (with Andreessen Horowitz participating) for AI estimating that produces a cost estimate in about 15 minutes. Ten more startups raised in the single week ending June 15, and the largest June check, roughly $52.9 million to Dublin-based CameraMatics, went to fleet and driver risk, adjacent to the jobsite rather than on it. The pattern: investors are funding operational plumbing tied to measurable outcomes, not moonshots.

Why it matters: Estimating, reality capture, and monitoring are where both the capital and the product maturity now sit, which makes them the rational place to spend a pilot budget. But today’s funding is next year’s acquisition pipeline, so pricing and lock-in questions belong in every pilot conversation. And the deeper question is whether venture-scale returns exist in categories where AI is simultaneously collapsing the cost of building software.

From the roundtable: The practitioner reception was notably cooler than the funding pace. Karl, whose job includes piloting exactly these tools: “None of these products interest me, because I think we have solutions for all these things already.” On XBuild’s 15-minute estimate, his instinct was to peel back what’s actually in it, because an estimate is a leveraged bet where bidding high loses the job and bidding low buys it. Tyler pressed the structural point: AI is making software cheap to replicate in-house. He described an estimator who built a working dashboard for his whole team using Claude skills, and a client paying $50K a year for what amounted to a database and a dashboard. His view is that hardware startups now have better longevity odds than software ones, and that data security, where a startup’s estimating tool sends your cost data, deserves more scrutiny than it gets.

5. Palantir vs. the Startups

The news: McCarthy, one of the largest U.S. builders, announced a multi-year strategic partnership with Palantir on June 4 to run operations on Pulse, an AI suite built on Palantir’s AIP and a “McCarthy Ontology” encoding more than 160 years of how the firm builds. It is framed as an operating partnership, not a pilot. The counter-position raised money the same week: Enlaye closed a $5 million seed, led by Glasswing Ventures and co-led by Link Ventures, to be a stack-agnostic contract and risk layer that works across whatever software a contractor already runs.

Why it matters: This is the defining strategic fork in construction AI: does the winning intelligence layer live embedded inside one platform’s ontology, or stay a neutral specialist that survives the next software-stack change? For the mid-size GC that cannot fund a McCarthy-scale build, the practical answer is neither vendor. It is keeping project data clean and portable so the firm can plug into whichever layer wins. The question to ask every AI vendor is who owns the data model and how hard it is to leave, because that is where lock-in hides.

From the roundtable: Tyler reviewed the full Pulse presentation and called what McCarthy built genuinely impressive, but was precise about the layers: the visible product is dashboarding, and the real asset is the ontology underneath, which is not a Palantir invention but simply a formal map of how a company’s data connects and what it means. “Ontology is going to become the term. You’re going to hear it ad nauseum along with AI.” Karl raised the diligence question that should precede any deal of this shape: “If you have a relationship like that, how do you get out of it?” Even among three technologists, nobody could confidently answer who owns Pulse. Tyler’s larger position, and his closing take of the episode, is that owned AI wins: the ontology a contractor builds inside its own walls compounds, while the rented one inherits the exit problem, which is why he’s seeing on-prem data strategies regain appeal.

6. Announced, Not Built: Data Centers Hit a Wall

The news: Data Center Watch found 75-plus U.S. projects worth about $130 billion were blocked or delayed in Q1 2026 alone, matching all of 2025 in a single quarter. Active opposition groups more than doubled to 833 across 49 states, with Maryland, Ohio, and Texas the hottest, and more than 300 bills hit statehouses in the first six weeks of the year.

Why it matters: For contractors tied to AI infrastructure, the constraint set just expanded from power and labor to social license, permitting, and courts. An announced project is not a funded start, so community posture and permitting status now belong in the diligence stack before staffing up. The work does not vanish when the box stalls; it shifts into the civil, power, and water scope around it, and geography increasingly determines how much announced pipeline actually converts.

From the roundtable: Karl builds in Northern Virginia, the densest data center market in the country, and his account made the friction concrete: developers arrive in small towns whose fire ladders cannot reach the proposed building height, and the negotiation becomes a new fire station, new infrastructure, new life for an old coal town. Some communities take that deal gladly. Others do not want the hum near a school at any price. His summary of the tension: everyone wants to use AI, but “when it comes to put that in my backyard, not a chance.” The group’s read is that incentive structures are improving under pressure but remain misaligned with the communities absorbing the impact.

7. Nuclear Goes Vertical for the AI Build-Out

The news: TerraPower broke ground on its Natrium plant in Kemmerer, Wyoming on April 23, the first U.S. commercial advanced (non-light-water) reactor construction start in more than 40 years. The 345 MW sodium-cooled design employs about 1,600 construction workers. Big tech has signed for more than 10 gigawatts of new nuclear, and Meta holds a deal for up to eight Natrium reactors, the first targeted around 2032.

Why it matters: The power demand behind Story 6 is generating its own construction category: advanced nuclear and SMR sites, prefab-heavy, often in rural labor pools. SMR economics target $4,000 to $7,000 per kilowatt against $8,000 to $12,000 for recent large builds, which rewards firms with factory-build and module-assembly capacity, the same muscle modular construction demands. The honest caveat: nuclear’s track record on schedule and cost is neither fast nor modular, and Natrium’s 2030 completion target will be the test. Wyoming’s first-ever reactor exists because of AI, which says everything about where the next decade of megaproject backlog is forming.

8. The Backlog Number Is Hiding a Split

The news: ABC’s Construction Backlog Indicator rebounded to 8.8 months in April, explicitly buoyed by data center and infrastructure work, after a winter low around 8.0. But firms under $30 million in revenue fell to roughly 6.9 months, their lowest in more than four years.

Why it matters: A single sector is now strong enough to move the national number, which means the headline can look healthy while most contractors feel a downturn. For firms without mega-project access, the play is a defensible niche rather than chasing a shrinking pool of commercial work down on price. The connection to Stories 1 and 6 is direct: data centers giveth backlog and taketh away labor, and now face their own conversion risk. Manage to your actual pipeline, not the average.

Outlook

The gap to watch is between the capital layer and the field. June delivered $126 million in fresh venture funding, a Palantir-scale platform commitment, and record trade-school enrollment, and yet the practitioners closest to the work remain unconvinced the current tooling moves the number that matters. Matt’s assessment stands as the month’s open challenge: “As much AI as we infuse into the industry, and all the new tech and money invested and everything else, productivity’s not increasing at all.”

Karl’s advice to the contech community doubles as the industry’s filter for the second half of 2026: spend 55 minutes of the hour on the problem before touching a solution, because “not every problem needs a tech solution. Sometimes it just needs better ownership.” Watch three things in July: whether the data center blocking pace holds into Q2 numbers, whether small-firm backlog stabilizes or keeps sliding toward consolidation territory, and how quickly “ontology” shows up in every contech pitch deck.

Full Episode Here:

Today's guests:

Karl Sorensen: https://www.linkedin.com/in/karlsorensen/

Matt Graves: https://www.linkedin.com/in/matt-graves-pmp/

Tyler Campbell: https://www.linkedin.com/in/tylerscottcampbell/

There is real skepticism about the value venture capital brings to construction, and I understand it. I have spent time on both sides of the table, first in the headquarters of a large builder, and now running a construction-focused fund. That vantage is what shapes how I think about where money helps, where technology actually sticks, and why building the product was never the hard part.

Start with the skepticism itself, because it is fair. Venture is a very small amount of capital relative to the dollars moving through our industry, and it carries an unusually high tolerance for risk. It is the equity invested in ideas and entrepreneurs to try the things that would not otherwise get tried.

In a sense, then, venture capital in construction is research and development for an industry that does not experiment widely on its own. Without that capital, the best ideas remain good conversations that never scale. Venture is what moves an idea from conversation to action. It is how you get gentle disruption in an industry that does not disrupt itself easily, and it most often comes from a team that pairs technical know-how with real construction experience.

Building and scaling tech are interrelated, yet uniquely different in construction

That pairing matters because building the product is rarely what kills a startup. Getting onto a jobsite, doing exactly what you said you would do, and earning the next introduction site after site, that is the hard part. Construction is fragmented. If you have done one project, you have done one project. The teams change, the buildings change, and every pilot begins from zero trust.

That is why go-to-market in construction has to be as innovative as the product itself. I have often pointed to the model in Singapore, where the government subsidizes pilots so a builder can try new technology without absorbing the full cost and friction of the experiment. If it works, people are happy to pay. If it does not, no real harm is done. Removing that friction is itself a go-to-market strategy, and it is hard to overstate how difficult that friction is to overcome.

The same fragmentation raises a harder question once a product exists, which is whether it is actually a business. Since 2025, that distinction has come into sharp focus. AI has made it dramatically cheaper to build a first version of almost anything, but cheaper MVPs raise the bar rather than lower it. A business is not “I built my first widget, now let’s sell it.” A business is happy customers, priced correctly, served consistently, with a product that keeps getting better. The companies worth backing are living, growing organizations that are always pushing forward, not a finished widget in search of a buyer.

Brought to you by FieldProof

Everybody in construction is being told they need to use AI. Almost nobody is showing operators how to actually do it. FieldProof runs a live, hands-on AI boot camp built specifically for construction professionals, estimators, PMs, superintendents, and BIM teams. Hands-on, not theory. If you want to get ahead of this instead of behind it, start with a cohort.

bootcamp.fieldproof.io

For the buyer, start with the workflow

That discipline cuts both ways, because the contractor on the other side of the table has homework too. Before a firm spends a dollar on technology, it should understand its own workflow completely. The first-principles questions sound almost too simple, things like “what am I actually trying to accomplish?” and “where in that workflow does it make sense to exchange human effort for technology?” Yet they pay off time and again. The best construction technology companies think the same way, which is why the strongest startups emphasize outcomes, not just tools. That outcome focus, and the empathy to sit alongside the contractor as a partner rather than a vendor, is what separates the companies that last.

The coming physical shift

All of that is software. Further out, the change becomes physical. The tower crane was invented around 1950 and was slow to catch on, until the point arrived when no one could imagine building without it. New physical tools in construction have always been adopted slowly. But robotics is becoming cheaper and more capable, and we are heading toward that same tower crane moment with a more diverse set of purpose-built tools, for drilling, for repetitive installations, and for the measurement we perform over and over. Not tomorrow, but it is coming, and it is worth watching closely.

Technology can improve the work, but it cannot replace trust

None of this, though, displaces the thing that actually makes projects happen. I do not believe a technological magic bullet erodes the human influence at the center of this industry. What good technology does is clear the routine work out of the way, so the best people can spend their time where judgment and relationships actually decide the outcome. The Master Builder may one day be possible, but not on the near-term horizon. Everything from large-scale token factories to custom homes still relies on professional judgment and interconnected, bespoke supply chains to deliver on time and on budget. That relationship-driven success, unique to every project team, is what makes construction a remarkable industry to work in, and to keep innovating for.

Recent episode with Aaron:

Aaron Toppston is Managing Partner at GS Futures, an early-stage venture fund focused on construction and real estate.

Today's Guest: Elliot Christiansen

Elliot Christiansen: https://www.linkedin.com/in/elliot-christiansen/

Episode Sponsors:

Check Out Trunk Tools: https://trunktools.com/

Check Out ClearStory: https://www.clearstory.build/

Check Out Document Crunch: https://www.documentcrunch.com/

Welcome to this weeks Construction Brief.

We’re in the dog days of summer, or is that July/August…all I know is I start sweating when I even think about going outside.

I am building an office! Better said I am moving my office back into a trailer…you know back to my roots. All I will be missing is a superintendent yelling at me about a scope gap I missed. Below is a picture of said office, its a 20’ container that was converted to a hunting lodge that I am converting into a office and podcast studio (sans yelling superintendent).

As a part of this build I am having to build a deck which has me thinking about some of John Fish’s comments about the return of the master builder in digital form. In my brief tenure as an “owner” managing the build of this deck, its quite enjoyable working with someone who can design and build. Just saying.

On My Mind - You Can’t Build from the Drawings

You can’t build from the drawings. Say it out loud and it sounds obvious, but it exposes how many layers of our design process exist to produce a document nobody can actually build from.

Start with what architects and engineers are actually for. They take an owner’s concept and put parameters around it, forcing the hard questions about how the building has to perform in a real world environment. That is necessary, valuable work, and it aligns desired outcomes before a shovel hits dirt. The problem is we don’t let it end there. The people who are good at problem discovery and programming are rarely the same people who should be figuring out the best way to waterproof two converging angles of roofing, and yet we keep them on the clock, designing toward some ambiguous finish line we call Construction Documents.

I learned this standing on a roof, looking at a detail where two slopes came together in a clean intersection on the sheet that did not exist in the framing we’d already built. The intent was obvious. The means to actually waterproof it were nowhere on the page. We wrote the RFI, we waited, and we eventually solved it the way it was always going to be solved, by the people standing on the roof.

That wait is the part we get wrong. We pour enormous energy into engineering the RFI out of existence, chasing a set of documents so complete that no question survives contact with the field. That’s the wrong target. The RFI is not waste, it’s the moment design finally meets reality, and the only real inefficiency is that we’ve staged it to happen after the documents are “done,” after the schedule assumed it wouldn’t, after the cost of answering it has tripled.

So stop trying to eliminate it. Incentivize it, and pull it forward. If the RFI is where intent becomes constructable, then it belongs inside the design process, not bolted onto the end as a defect report. Picture the master builders in the room while the design is still soft, paid to ask the converging-roof-angle question before the framing goes up instead of after. The question is identical either way. The only variable is whether we treat it as collaboration or as a change order.

I understand the counter. Design-Build, IPD, and the other delivery models were built specifically to kill the misaligned incentives that create this gap, and I’ve built under each of them. They each carry real flaws that lead right back to the same question, which is why we insist on over-designing to intent when that intent still has to be translated to something buildable by the people holding the tools.

Maybe the master builders were never supposed to receive a finished design. Maybe they were supposed to finish it.

Tech to Watch - Primepoint AI

AI attempts at drawings so far has been a chatbot bolted onto a pile of PDFs, useful right up until it confidently invents a spec section that doesn't exist. Primepoint is taking a different swing. Instead of treating your documents as a wall of text to search, it reads the drawings themselves, the linework, the tags, the cross references, and connects them into a model it can actually reason about. The pitch lands close to home this week, because the entire problem with drawings is that the intent on the sheet and the reality in the field rarely line up, and Primepoint is built to catch that gap before the field does.

What it is

Primepoint is a construction AI platform that reads drawings, specs, and schedules as connected systems, then flags conflicts, drafts RFIs, and answers plain-English project questions with every response traced back to the exact drawing detail or spec section.

Why it matters

Its Constructability Review scans for conflicts between drawings and specifications during preconstruction and generates traceable RFIs off what it finds, which is the upstream move the industry keeps saying it wants and rarely staffs for. Because every answer links back to a source, you aren’t taking the model’s word for it, which is the right response to the hallucination problem that makes most teams rightly nervous about AI on contract documents. Sundt’s group controls manager claims it cut document navigation time by up to 50 percent. Treat that as a customer figure rather than an audited result, but the direction is credible, and the named contractors already on it, Sundt, Fortis, and W.E. O’Neil, are not small shops experimenting at the margins.

Who it’s for

Who isn’t it for…if you interact with drawings, this tool is for you. GCs and their preconstruction, VDC, and project engineering teams, the people buried in submittals, RFIs, and revision sets. Controls managers watching rework eat margin. Anyone whose job is to find the conflict on the sheet before the framing crew finds it for them.

Check it out

https://www.primepoint.ai

Field Notes - The Owner Who Merged OAC and Coordination

This section is where I highlight one thing I am hearing in the industry that might be a leading indicator for a future trend.

A hyperscaler representative mentioned, almost in passing, that they’d started running their OAC and their coordination meetings as a single meeting. It was rough at the start, the way any change to a standing ritual is, but what it bought them was transparency. The owner, the architect, and the trades were all looking at the same model at the same time, and the model started getting treated as the source of truth instead of one more reference nobody fully trusted.

Yes, the meeting runs longer. What it quietly killed was the string of follow-on meetings that used to exist only because the people who carry intent and the people who carry constructability were never in the same room at the same time. Put them there, in front of the model, and most of the decisions that used to spawn their own meeting just get made.

The reflex in this industry is always toward fewer meetings and shorter agendas, as if the cost were the time on the calendar. It isn’t. The cost is decision latency, and one heavier meeting with everyone who matters in it beats four lean ones that each end in “let me check and get back to you.”

It’s one data point from one owner. But hyperscalers have the program volume to rewrite a ritual and make it stick, and when they change how the room works, the contractors chasing that work tend to change with it.

See you all in a couple of weeks.

- Stephen

Today's Guest: Sean Johnson

Sean Johnson: https://www.linkedin.com/in/seanjohnson/

Episode Sponsors:

Check Out Trunk Tools: https://trunktools.com/

Check Out ClearStory: https://www.clearstory.build/

Check Out Document Crunch: https://www.documentcrunch.com/