Best of LinkedIn: Digital Construction CW 02/ 03

00:00:00: This episode is provided by Thomas Allgaier and Frennis, based on the most relevant LinkedIn posts about digital construction in calendar weeks O-two and O-three.

00:00:09: Frennis is a BDB market research company that supports enterprises across the construction industry with the market, customer, and competitive insights they need to navigate dynamic markets and drive customer-centric product development.

00:00:23: Welcome to the deep dive.

00:00:24: You know, I was looking back at the notes from this time last year, early twenty twenty five, and the energy in the industry was just

00:00:31: manic.

00:00:31: Oh, frantic.

00:00:32: Absolutely.

00:00:32: Everyone was screaming about how generative AI was going to replace architects by once time.

00:00:36: But looking at the feed from weeks two and three of twenty twenty six, the vibe has shifted.

00:00:43: It feels heavier.

00:00:44: but in a good way.

00:00:45: I know exactly what you mean.

00:00:46: It feels like the industry has finally exhaled.

00:00:49: We've moved past the peak of inflated expectations on the hype cycle, and we are squarely in the work phase.

00:00:56: The conversation isn't about sci-fi anymore.

00:00:57: It's not about robot dogs doing backflips in a promo video.

00:01:00: It's about what technology is actually doing on a muddy job site in the rain on a Tuesday.

00:01:05: Right.

00:01:06: It's less magic and more mechanics.

00:01:08: Which is exactly where we need to be.

00:01:09: The posts we're analyzing today from people like Shaulk Theron, Matt Lockett, Ross Griffin, they all share this common thread of demanding evidence.

00:01:19: We are seeing a pivot from look at what this could do.

00:01:22: to show me the hard outcomes.

00:01:26: So we have a massive stack to get through.

00:01:28: We've grouped these insights into four main clusters for you.

00:01:31: We're going to start with the evolution of AI into what's being called operator mode.

00:01:35: Okay.

00:01:36: Then we're getting into the weeds of digital twins, specifically why more detail is actually a trap.

00:01:41: That's

00:01:42: a controversial one, especially for the engineers listening.

00:01:44: Then we'll

00:01:45: move to the human side lean scheduling and some honestly terrifying statistics about workflow continuity.

00:01:50: And finally we'll zoom out to the market signals, funding.

00:01:53: and leadership.

00:01:54: Let's

00:01:54: dive in.

00:01:55: Okay, let's start with AI.

00:01:57: And I want to challenge you on this first concept right out of the gate.

00:02:00: Shalk Farron posted about the need for AI as an operator.

00:02:04: To me, that sounds a bit like rebranding.

00:02:07: We've had assistance in co-pilots for two years now.

00:02:10: What makes an operator different from a co-pilot?

00:02:13: Is this just semantic drift?

00:02:16: It's a fair skepticism, but I don't think it's just semantics.

00:02:19: Think about the cognitive load.

00:02:21: co-pilot or an assistant, like ChatGPT or the early construction bots, they are passive.

00:02:27: They sit there and wait for you to ask the right question.

00:02:30: You still have to drive the process.

00:02:33: Theron's argument is that construction doesn't need another portal where we have to go and do work.

00:02:38: We need systems that execute the work.

00:02:39: Yes,

00:02:40: agency, that's the word.

00:02:42: An assistant helps you draft an email about a permit application.

00:02:46: An operator, well, an operator logs into the municipal portal, fills out the fields based on the project.

00:02:51: uploads the PDF and just pings you for a final signature.

00:02:55: It's the difference between, help me do this and just get this done.

00:02:58: Okay, I see the distinction.

00:02:59: So it's shifting from augmentation to autonomy.

00:03:04: But, and this is a big but, if we're handing over the keys to the car, shouldn't we be worried about where it's driving?

00:03:11: Matt Lockett had a very strong reaction to this.

00:03:13: Yeah, Lockett is playing the role of the necessary cynic here.

00:03:17: He's looking at all these tier one contractors making these.

00:03:20: bold claims about predictive analytics and autonomous workflows.

00:03:23: And he's just asking, where is the commercial evidence?

00:03:26: He wants to see the receipts.

00:03:28: He does.

00:03:28: He warns against drowning in dashboards.

00:03:31: We have all these BI tools now, you know, red, amber, green lights everywhere.

00:03:35: But do we trust

00:03:37: them?

00:03:37: Right.

00:03:38: Lockett's point is that if you trust an algorithm that doesn't understand the messy physical reality of a construction site, wet concrete, delayed trucks, angry subcontractors, you aren't just going to make a mistake.

00:03:50: You're going to make an expensive, confident mistake.

00:03:52: That's the danger of the black box.

00:03:53: If the AI says schedule is green, but you're standing in mud, who do you believe?

00:03:58: You believe your boots.

00:03:59: Always.

00:04:00: But if we want to move past cynicism to actual utility.

00:04:06: Chris Jeff Vullibich shared an example that I thought was fascinating.

00:04:08: It's about parallel AI audits in Revit.

00:04:12: Now, for anyone who's not a technical user, usually checking a three-D model is a serial process, right?

00:04:18: A human coordinator clicks through room by room.

00:04:21: Correct.

00:04:21: It's slow, it's manual, and frankly, it's mind-numbing.

00:04:25: You miss things because you're human and you get tired.

00:04:27: So how does Vullibich's approach change that?

00:04:30: This is a great example of that operator concept we were just talking about.

00:04:33: Instead of one human, they have multi-agent AI workflows running simultaneously.

00:04:38: So imagine four different AI agents auditing the model at the exact same time.

00:04:42: Okay.

00:04:43: Agent A is checking wall alignments.

00:04:44: Agent B is verifying room naming conventions.

00:04:47: Agent C is checking scope boxes.

00:04:49: Agent D is analyzing sheet views.

00:04:51: So

00:04:51: it's like having a team of four meticulous interns working in parallel instantly.

00:04:56: Precisely.

00:04:57: And the impact isn't just that it's faster.

00:04:59: It's that it catches errors before you get to the coordination meeting.

00:05:02: It saves hours per model iteration.

00:05:03: That is the hard outcome Matt Lockett is asking for.

00:05:06: It's not a dashboard.

00:05:08: It's a clean file.

00:05:09: Exactly.

00:05:09: But

00:05:10: here's the catch.

00:05:11: To run agents like that... you need good data.

00:05:13: And Henry Lee made a point that I think gets overlooked.

00:05:16: We're all obsessed with chips, Nvidia, GPUs, compute power.

00:05:22: Lee says that's not the bottleneck for

00:05:24: us.

00:05:24: No, he argues the bottleneck for physical AI.

00:05:28: That's AI that interacts with the real world is data quality.

00:05:31: You can't train a construction AI on internet text.

00:05:34: You can't train it on Reddit or Wikipedia.

00:05:36: Right.

00:05:37: It needs high quality structured data about geometry, physics, and logistics.

00:05:41: And this connects to that vivid image from Ross Griffin, the headless chicken scenario.

00:05:46: It's

00:05:46: a terrifying visual, isn't it?

00:05:48: But it's accurate.

00:05:49: If you feed these powerful operator AIs bad data, the classic garbage in, you don't just get garbage out, you get garbage out at light speed.

00:05:57: You're running around like a headless chicken, but efficiently.

00:05:59: Right.

00:06:00: So how do we keep the head on the chicken?

00:06:02: Griffin calls for connectors.

00:06:04: A new role.

00:06:05: A new breed of professional.

00:06:07: Yeah.

00:06:07: not just a software engineer and not just a bricklayer.

00:06:10: We need these hybrid professionals who can translate.

00:06:13: They need to look at the code and say that logic holds and then look at the site and say that's physically impossible.

00:06:18: They validate that data bridge.

00:06:20: It's interesting.

00:06:21: you mentioned the scale of investment needed for this because Rob Matheson dropped some numbers on the GCC region, Saudi Arabia and the UAE.

00:06:28: The scale there is just, it's different.

00:06:31: It's immense.

00:06:32: Matheson points out that while the global AI and construction market is maybe five to ten billion dollars today, the GCC is projected to outperform global adoption rates.

00:06:42: They're moving toward a fifteen to twenty five billion dollar total market faster than anyone else.

00:06:47: Because of vision twenty thirty.

00:06:48: Right.

00:06:49: In Europe or the US, we are retrofitting tech onto old processes and, you know, old regulations.

00:06:56: In the GCC, specifically with projects like Inom, they are building entire cities on a new stack from day one.

00:07:02: The mandate for innovation is coming from the very top.

00:07:05: They aren't dipping a toe in.

00:07:06: No, they're cannonballing into the pool.

00:07:08: Okay, let's shift gears to our second cluster, Digital Twins and BIM.

00:07:13: And I want to stick with this idea of context.

00:07:15: Santosh Kumar Boda shared an analogy that really stuck with me.

00:07:18: He said a building without geospatial context is just an island of data.

00:07:23: It's a brilliant way to frame the silo problem.

00:07:26: Traditionally, BIM building information modeling, it stops at the fence line.

00:07:31: We model the HVAC, the walls, the furniture with incredible precision.

00:07:36: But the building doesn't exist in a vacuum.

00:07:38: It needs water from the city, power from the grid, access roads for logistics.

00:07:42: So he's arguing for GIS Geographic Information Systems integration.

00:07:46: Yes,

00:07:47: to connect the island to the mainland.

00:07:49: If you want a digital twin that is actually useful for operations, For the next fifty years of the building's life, you have to answer where is this object and what is around it, not just what is this object.

00:08:00: Right.

00:08:00: So if a water main bursts three streets away, your digital twin should know how that affects your cooling system.

00:08:06: It should.

00:08:07: But speaking of modeling, there's a trap here.

00:08:10: And honestly, this is counterintuitive.

00:08:11: We usually think more detail is better, high definition is better than standard.

00:08:16: But Florian Humeur argues that maximal detail is a trap.

00:08:20: This is one of the most important insights of the week.

00:08:22: Engineers, by nature, were completionists.

00:08:25: We love to model every bolt, screw, and washer.

00:08:28: It feels

00:08:28: accurate.

00:08:30: But humor argues that for digital twins, that accuracy is actually a liability.

00:08:35: How so?

00:08:36: Just because of file size.

00:08:37: File size, yes, but also usability.

00:08:39: He compares the Tesla approach to the typical construction approach.

00:08:44: Tesla doesn't just scan a car visually.

00:08:45: They simulate functional data aerodynamics, drag coefficients.

00:08:49: Then look at General Electric.

00:08:51: They built a digital twin farm for their wind turbines.

00:08:54: And they didn't

00:08:54: model the paint color, I'm guessing?

00:08:56: No.

00:08:56: And they didn't model the bolts on the access ladder either.

00:08:58: They modeled motor temperature and wind strength.

00:09:01: Why?

00:09:02: Because those are the variables that affect power generation.

00:09:05: By focusing on functional data over visual data, they increased energy production by twenty percent.

00:09:11: That's a hundred million dollars in value.

00:09:14: So if they had spent six months modeling the aesthetic curve of the turbine blades perfectly that would have missed the ROI.

00:09:20: Exactly.

00:09:21: You have to define the problem before you determine the level of detail, the LOD.

00:09:25: If you don't, you build a model that is heavy, expensive, and solves nothing.

00:09:29: That ties into Oliver Eichert's point about BIM and LEAN.

00:09:32: He says they only work when goals are explicit.

00:09:36: It's not about the software, it's about the transparency.

00:09:39: Right.

00:09:39: If you aren't collaborating early, all the fancy modeling in the world won't save you from a chaotic schedule.

00:09:45: Which

00:09:45: is the perfect segue to our third theme, lean, scheduling, and the workforce.

00:09:50: And I have to say, this next statistic from Amir Berman at Bildots, it actually scared me a little.

00:09:57: It explained so much about why projects fail.

00:09:59: The flow statistics?

00:10:01: Yes.

00:10:01: So Bildots analyzed residential projects to find out when flow breaks down.

00:10:06: They looked at how long it takes a trade to complete a floor.

00:10:08: They found a tipping point.

00:10:09: If trades finish a floor in a one to three week window, the first pass complete rate is seventy three percent, meaning it's done right.

00:10:18: No rework.

00:10:19: That's

00:10:19: a healthy, profitable number.

00:10:22: But if that duration stretches to four weeks, just one extra week, the completion rate drops to fifty eight percent.

00:10:29: And if it goes to five weeks.

00:10:31: It plummets to thirty-five percent.

00:10:33: That

00:10:33: is a cliff.

00:10:34: That's not a gradual decline.

00:10:35: That is a complete collapse.

00:10:37: So why does that happen?

00:10:38: Why is the difference between week three and week four so catastrophic?

00:10:42: It's entropy.

00:10:43: Think about the reality of a site.

00:10:45: If a crew is on a floor for three weeks, they have rhythm.

00:10:48: Materials are there, the plan is fresh in their minds.

00:10:51: Once you cross that line into week four or five, it means something has gone wrong.

00:10:55: Maybe materials are missing, maybe they got pulled to another area.

00:10:58: Now you have stop-start work.

00:11:00: And cognitive load increases.

00:11:01: You forget what you did two weeks ago.

00:11:03: Exactly.

00:11:04: You have return visits, someone damages work that was already done.

00:11:07: It's a clear early warning sign for any project manager.

00:11:11: If a floor is dragging into week four, don't just monitor it.

00:11:14: Sound the alarm.

00:11:15: But why are we dragging in the first place?

00:11:17: Yeah.

00:11:17: AJ Waters had a really sharp take on this.

00:11:20: He pointed out that, statistically, the U.S.

00:11:23: is actually world-class at tracking cost and schedule.

00:11:27: Yeah.

00:11:27: We are better at it than almost anyone else.

00:11:29: We love our spreadsheets.

00:11:31: We are great at measuring the disaster.

00:11:33: Right.

00:11:33: We measure the crash in high definition.

00:11:36: But only two percent of projects finish on time and on budget.

00:11:40: It's the paradox of measurement.

00:11:42: We measure everything, but we fix nothing.

00:11:45: Waters identifies the root cause as scope clarity.

00:11:49: We don't understand what we are actually building early enough.

00:11:51: We rush into construction to show progress, but the scope isn't locked.

00:11:55: So disputes are inevitable.

00:11:56: Disputes don't disappear just because we have a schedule.

00:11:59: They just wait.

00:11:59: They wait in the background until they turn into a shutdown or a lawsuit.

00:12:03: We prioritize speed over certainty in the early phases and we pay for it with interest at the end.

00:12:08: So we have scope creep, we have schedule entropy, the natural reaction is buy more tech to fix it.

00:12:13: Yeah.

00:12:13: But Aditya Thakhar offered a filter for that, which I think every listener should write down.

00:12:18: It's a great heuristic to avoid tool sprawl.

00:12:20: He asked three questions before buying new tech.

00:12:23: Number one.

00:12:24: Does it reduce coordination costs?

00:12:26: Number two, does it create a reusable data asset?

00:12:30: And number three, this is the kicker.

00:12:32: Does it improve predictability within thirty to sixty days?

00:12:36: That timeline is aggressive, but it's necessary.

00:12:38: Thirty to sixty days.

00:12:41: In construction, we love to talk about ROI.

00:12:43: that happens, what?

00:12:45: Three years from now, Thakur is saying, no.

00:12:47: If the impact is eventual, it's usually imaginary.

00:12:51: If impact is eventual, it's usually imaginary.

00:12:52: That belongs on a bumper sticker.

00:12:54: It does.

00:12:55: But let's connect this back to the people using these tools.

00:12:57: Christopher Lingjesa made a point that resonates deeply.

00:13:01: He says this isn't a digital transformation, it's a workforce evolution.

00:13:05: What's the distinction there?

00:13:06: Transformation implies we simply change the systems, upgrade the software, change the server.

00:13:11: Evolution implies the people change.

00:13:13: He argues we need to stop treating foremen like data entry clerks.

00:13:16: That's the administrative tax.

00:13:18: Exactly.

00:13:19: We take our best builders.

00:13:21: people who have twenty years of experience with concrete and steel.

00:13:24: And we force them to spend fifteen hours a week fiddling with an iPad, entering data into a field they don't understand.

00:13:31: We need to remove that tax so they can focus on the craft.

00:13:33: And Ryan Hines backed this up from the sales side.

00:13:35: He noted that construction experience is becoming a mandatory requirement for contact rules.

00:13:41: You can't just be a soft salesperson anymore.

00:13:44: No, because if you don't understand the friction on the site, if you don't know what an RFI actually feels like when it delays a pour, you can't sell the solution.

00:13:53: It goes back to Ross Griffin's connectors.

00:13:55: The industry is demanding people who bridge the gap between code and concrete.

00:14:00: Okay, let's move to our final cluster, the ecosystem.

00:14:04: What is the market actually telling us right now about where the money is going?

00:14:07: Well, according to Patrick Hellerman's Q for report, the market is taking a breath.

00:14:11: It has stabilized.

00:14:12: We saw about five hundred and fifty million dollars raised in Q four twenty twenty five.

00:14:17: So not the explosion of twenty twenty one but not a drought either.

00:14:21: Exactly.

00:14:21: It's a normalized healthy market.

00:14:23: The tourist capital has left.

00:14:25: The investors who are left actually understand the industry.

00:14:28: And some of those deals are for things that aren't exactly sexy but are incredibly important.

00:14:34: Jesse Landry highlighted the story of city logics.

00:14:37: This is a perfect example of.

00:14:38: hard outcomes.

00:14:39: Right.

00:14:40: CDLogix isn't about flying robots or three-D printed houses.

00:14:43: It's about road data.

00:14:45: Potholes, municipal budgets.

00:14:47: And CRH Ventures invested in them.

00:14:49: Why?

00:14:50: Because it's infrastructure intelligence that scales.

00:14:53: It helps cities manage their assets better.

00:14:55: It's unglamorous, but it's essential.

00:14:57: That is where the smart money is going.

00:14:59: Solutions that solve boring, expensive

00:15:01: problems.

00:15:02: Speaking of boring, but essential, let's talk about compliance.

00:15:05: Christina Polinova shared some news about Autodesk achieving the BSI KiteMark certification for ISO.

00:15:12: Now I know ISO nineteen six fifty sounds like the cure for insomnia.

00:15:16: It

00:15:16: does doesn't it?

00:15:17: But stick with us on this.

00:15:18: No, why does it matter?

00:15:19: It matters because it's a signal of maturity.

00:15:22: ISO nineteen six fifty is the international standard for managing information over the whole life cycle of a billed asset.

00:15:29: The BSI kite mark is the gold standard of trust for Autodesk.

00:15:33: to get this specifically related to the UK's Building Safety Act It signals that the Wild West era of data is over.

00:15:41: You can't just say you did the work.

00:15:42: You have to prove the data is accurate, traceable, and secure.

00:15:46: It's a bad accountability.

00:15:47: It is, and trust.

00:15:48: And trust brings us to our final leadership note.

00:15:51: Corey Cutler gave a shout out to Tracy Van Dalsum, and the core message was about kindness.

00:15:55: It sounds soft, doesn't it?

00:15:57: In an industry defined by concrete steel and contracts, kindness feels out of place.

00:16:02: What

00:16:02: is it?

00:16:02: Not

00:16:03: at all.

00:16:03: In a high-pressure environment, kindness is a competitive advantage.

00:16:07: When a project is over budget and behind schedule, panic destroys value.

00:16:12: People hide mistakes because they are afraid of being yelled at.

00:16:14: And hiding mistakes leads to that three-week cliff we talked about earlier.

00:16:18: Exactly.

00:16:19: Calm leadership kind leadership creates psychological safety.

00:16:23: It allows people to report bad news early when it's still fixable.

00:16:27: The leader who stays calm is the one who actually solves the problem.

00:16:30: So

00:16:30: let's synthesize this.

00:16:32: We've covered a lot of ground.

00:16:33: We've gone from AI operators executing tasks to the trap of maximum detail in digital twins to the entropy of construction schedules and finally to a market that values stability and trust.

00:16:45: I think the synthesis here is maturation.

00:16:47: We are growing up.

00:16:48: AI is moving from a magic trick to a blue collar operator.

00:16:52: Digital twins are moving from pretty, three-D models to functional tools that solve specific physics problems and the workforce.

00:17:00: The workforce is demanding tools that actually help them build, not just tools that help them report on building.

00:17:05: We are scraping away the noise.

00:17:06: Exactly.

00:17:07: We are looking for the signal.

00:17:08: I want to leave you with a final thought from James Swanson.

00:17:11: He predicts that AI washing isn't going away.

00:17:15: Companies will keep slapping the AI label on everything.

00:17:18: Of course.

00:17:19: It's the buzzword of the decade.

00:17:21: But his challenge is, can you spot the difference?

00:17:24: When you look at a tool, Can you tell if it's just a fancy label or if it's an operator that will actually change your outcome, improve your predictability within thirty days?

00:17:34: That's the litmus test.

00:17:35: Don't tell me it's AI.

00:17:37: Tell me it works.

00:17:38: If you enjoy this deep dive, new episodes drop every two weeks.

00:17:41: Also check out our other editions on smart manufacturing and digital power tools.

00:17:45: Thanks for listening.

00:17:46: Keep

00:17:47: learning.

00:17:47: Subscribe so you don't miss the next one.