Best of LinkedIn: Digital Construction CW 04/ 05

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-Four and O-Five.

00:00:08: Frennis is a B to B market research company that supports enterprises across the construction industry with a market, customer, and competitive insights.

00:00:16: they need to navigate dynamic markets and drive customer-centric product development.

00:00:21: So here we are again looking back at calendar weeks, oh four and oh five of twenty twenty six.

00:00:27: Right.

00:00:27: And you know if I had to put a label on the industry's mood right now I'd say we are finally dealing with the hangover.

00:00:33: The hangover.

00:00:34: That sounds a little bleak for a tech update.

00:00:38: Well, sober.

00:00:39: Think about it.

00:00:40: We spent the last couple of years kind of drunk on hype.

00:00:42: Yeah,

00:00:42: that's

00:00:43: right.

00:00:43: AI was going to magically generate skyscrapers from a text prompt.

00:00:47: Robots were everywhere.

00:00:48: Blockchain was going to solve contracts.

00:00:50: But looking at the sources from these two weeks, I think the party's over.

00:00:53: So we're moving from the hype cycle into the

00:00:55: deployment phase.

00:00:56: Right.

00:00:57: It feels less about the shiny new toy and more about the headache of actually getting it to work on a muddy job site.

00:01:03: The themes this week are, I mean, they're aggressively practical.

00:01:06: Exactly.

00:01:06: We're seeing this shift toward restructuring AI for real accuracy, building the data spine so things can actually connect and then, you know, seeing the market itself mature.

00:01:16: And it's not even mentioning the huge shifts in what we're physically building, like the insane rise of data centers.

00:01:22: But you're right, the main theme is structure.

00:01:24: It's all about building the spine.

00:01:26: Well, let's get into it.

00:01:27: Let's start with the tech.

00:01:28: Everyone is still talking about AI and automation.

00:01:31: But the conversation has definitely shifted.

00:01:35: It has.

00:01:36: Not just about.

00:01:37: generative capabilities anymore.

00:01:38: It's about what AJ Waters called decision intelligence.

00:01:41: This distinction AJ made is so critical for anyone trying to justify an AI budget right now.

00:01:47: He argues that most of the industry is currently stuck using AI for personal productivity.

00:01:53: Personal productivity.

00:01:54: So what does that mean exactly?

00:01:54: Just writing my emails faster.

00:01:56: Basically, yeah.

00:01:58: Summarizing emails, cleaning up meeting notes, you know, writing generic RFIs.

00:02:02: It helps you as an individual save maybe twenty minutes a day.

00:02:06: Which feels good.

00:02:07: It

00:02:07: feels great.

00:02:08: But AJ's point is that construction projects don't fail because someone wrote an email too slowly.

00:02:14: No, they fail because a critical decision was made three weeks too late.

00:02:17: Precisely.

00:02:18: Or it was made without seeing the full picture.

00:02:20: He's pushing for a shift to decision intelligence.

00:02:23: So an AI that doesn't just summarize notes but actually... analyzes the data.

00:02:28: Yes.

00:02:29: Imagine an AI that says, based on the delay in steel delivery from the logs, you have three decisions to make about the concrete pour and hear the cost implications of each.

00:02:40: That is a massive leap in utility.

00:02:42: It moves the AI from being a secretary to being a strategist.

00:02:45: It does.

00:02:46: But to get an AI to give you that kind of advice, advice you'd actually bet a million dollars on, it has to be accurate.

00:02:51: And we all know these models hallucinate.

00:02:52: Oh, yeah.

00:02:53: Aaron Holmes shared a data point on this that I found really grounding.

00:02:56: The

00:02:56: cost estimation study, you mean?

00:02:58: Yes.

00:02:59: They compared generic AI models against structured workflows.

00:03:03: So they took a standard LLM, like a chat GPT, and just asked it to estimate construction costs.

00:03:08: The accuracy was about sixty-four percent.

00:03:10: Which,

00:03:11: in our industry, is catastrophic.

00:03:13: I mean, you'd be better off guessing.

00:03:14: You cannot run a business on sixty-four percent accuracy.

00:03:17: You can't.

00:03:18: But, and this is the key insight, when they wrap that same model in a structured AI workflow, the accuracy jumped to eighty-four percent.

00:03:26: Okay, we need to unpack that for a second because structured workflow can sound a bit like buzzword soup.

00:03:31: What does that actually look

00:03:32: like?

00:03:33: Right.

00:03:33: It's the difference between an open-ended question and, well, a guided process.

00:03:37: So instead of just asking how much to build a hospital.

00:03:41: The workflow forces the AI to break the problem down into logical steps.

00:03:45: First, calculate the foundation based on these specific parameters.

00:03:49: Then calculate the steel tonnage.

00:03:51: Then apply these local labor rates.

00:03:53: They made the AI follow a logic tree instead of just predicting the next word.

00:03:57: So the intelligence wasn't really in the model itself.

00:03:59: It was in the architecture built around the model.

00:04:02: Exactly.

00:04:03: And this connects perfectly to Guido Masiochi's post about why so many AI pilots are just crashing and burning right now.

00:04:11: What was his take?

00:04:12: He says it's because no one owns the evils.

00:04:15: Evils as in evaluations.

00:04:17: Right, the testing benchmarks.

00:04:18: Yeah.

00:04:19: Guido's point is that you can buy the commodity layer, the big AI model from the tech giants, but you have to build that last mile yourself.

00:04:27: You have to build the benchmarks for your specific workflows.

00:04:30: You have to.

00:04:31: If you can't measure if the AI is getting better or worse at answering your questions about your contracts, you can't safely deploy it.

00:04:38: You're just hoping.

00:04:38: Makes a ton of sense.

00:04:39: It does, but it sounds like a lot of heavy lifting.

00:04:41: You know, it's not just plug and play.

00:04:43: Yeah.

00:04:43: But while some are building these complex logic trees, Erdem Everyn shared a list that felt much more accessible.

00:04:49: He focused on a tool called NEN.

00:04:52: NEN is great because it's low code.

00:04:54: It lets you stitch different apps together without being a full on developer.

00:04:58: And Erdem's list was just so practical.

00:05:00: He wasn't talking about replacing project managers with robots.

00:05:03: He was talking about automating incident alerts, site monitoring.

00:05:07: The unsexy stuff.

00:05:08: Moving data from A to B so a human doesn't have to copy paste.

00:05:12: That's where the immediate value is.

00:05:13: It really

00:05:14: is.

00:05:15: But, you know, before you go out and automate everything, Maria Seninga had this very blunt warning that, well, it really stopped me in my tracks.

00:05:22: What

00:05:22: did she say?

00:05:23: She said, AI cannot fix chaos.

00:05:27: It amplifies it.

00:05:28: Oof.

00:05:29: I think a lot of firms are finding that out the hard way.

00:05:31: It's the classic garbage in, garbage out problem, but on steroids.

00:05:36: She argues most teams don't need more intelligence yet.

00:05:38: They need less friction.

00:05:40: If you layer AI on top of a mess, you just get automated chaos.

00:05:44: You just make bad reports faster.

00:05:46: Exactly.

00:05:47: Which brings us to Henry L's challenge.

00:05:49: He basically told the industry to stop reading about tech.

00:05:53: Maybe even stop listening to us and stop planning for the perfect implementation.

00:05:57: Right.

00:05:57: The ninety day challenge.

00:05:58: Pick one real scenario.

00:06:00: Deploy a solution.

00:06:02: Measure the results.

00:06:02: If you can't prove value in ninety days, you're just playing innovation theater.

00:06:06: That ninety day clock is ticking.

00:06:09: But if you want to succeed, based on what Maria said, you need clean data.

00:06:14: And that bridges us perfectly to our second theme.

00:06:17: The data is fine.

00:06:18: Or as I like to call it, the eat your vegetables phase of digital transformation.

00:06:22: It's not fun, but if you don't do it, nothing else is going to work.

00:06:26: Chris Clark gave a perfect, painful example of this.

00:06:28: He looked at rental businesses.

00:06:30: They sit on mountains of data, but it's often just unstructured junk.

00:06:34: Right.

00:06:34: He used the example of a service record that just says, ring Bob, digger broke.

00:06:38: I've seen those exact sticky notes and the resolution in the system is probably fixed.

00:06:42: digger, Bob happy.

00:06:44: Exactly.

00:06:44: Bob happy.

00:06:46: Now.

00:06:46: Try feeding that into an AI.

00:06:48: An AI can't analyze Bob happy.

00:06:50: It doesn't know what component failed, how long it took, what the cost was.

00:06:54: It's a complete dead end.

00:06:55: And John Sinclair really doubled down on this.

00:06:58: He pointed out that we're all obsessed with productivity gains, but without standardization, specifically aligning cost codes, ERP data, and BIM data, those gains are mathematically impossible.

00:07:09: That's a language problem, isn't it?

00:07:10: It is.

00:07:11: If finance calls a brick a masonry unit, and the site team calls it external facing, and procurement calls it red block, your system is broken.

00:07:21: You cannot automate what you can't define.

00:07:24: And speaking of definitions, digital twin is another one of those terms that gets thrown around so loosely.

00:07:29: It usually just means a pretty three-D model.

00:07:33: Yeah.

00:07:33: But Florian relayed out a framework that actually makes sense of it.

00:07:37: He describes it as a loop.

00:07:39: not just a model.

00:07:40: That distinction is so vital.

00:07:42: Most people think a digital twin is the three D asset, but Florian argues that's just a digital file.

00:07:48: A true twin has three layers.

00:07:49: Okay, where are they?

00:07:50: Layer one is physical sensors, reality capture, the job site itself.

00:07:54: That feeds layer two, the virtual where you run predictions and simulations.

00:07:58: And

00:07:58: then, and this is the part that's usually missing, layer two has to inform layer three, which is the service layer.

00:08:03: Right,

00:08:03: the service layer is where the decision happens, which then changes something back in the physical layer.

00:08:07: If that loop breaks, if the data just sits there and doesn't drive an action, It's not a twin.

00:08:13: It's just a heavy file sitting on a server.

00:08:14: Exactly.

00:08:15: Taking up space.

00:08:16: Santosh Kumar Bodo gave a really vivid example of this loop in action for urban resilience.

00:08:22: He described a fire scenario in a high-rise.

00:08:25: This is where the tech becomes life-saving, not just money-saving.

00:08:28: Absolutely.

00:08:29: Without a twin, first responders show up with maybe a PDF floor plan, they're going in blind.

00:08:33: Right.

00:08:33: But with a functional urban twin, the incident commander sees a live model before the truck even leaves the station.

00:08:40: They see where hazardous materials are stored.

00:08:42: They see heatmaps from sensors.

00:08:44: That is the service layer driving a physical outcome saving lives.

00:08:48: The loop is closed.

00:08:49: Yeah.

00:08:50: But I want to pivot to a concept that challenges how we even think about that virtual layer.

00:08:55: Dawn Doe posted something that might upset some of our project controls listeners.

00:09:01: The float does not exist.

00:09:03: I saw that one.

00:09:04: Yes.

00:09:04: He argued that concepts like float and critical path are imaginary.

00:09:08: They're artifacts from a sketch paradigm where we treat projects as vague ideas that get clearer over time.

00:09:14: That's

00:09:14: a huge statement.

00:09:15: I mean, critical path is the Bible of construction scheduling.

00:09:18: If we don't use that, how do we manage a project?

00:09:20: He proposes shifting to an assembly mindset.

00:09:24: He uses a Lego analogy.

00:09:25: When you buy a Lego set, you don't estimate how many bricks you need and you don't calculate float for placing a brick.

00:09:32: Right.

00:09:32: You just have a defined set of instructions.

00:09:34: Exactly.

00:09:34: You have a defined set of work packages, bricks, and you assemble them in a specific order.

00:09:40: If you treat a building like that as an assembly of products, you can actually use manufacturing principles.

00:09:46: So you have to define the project with that level of granularity before you even start.

00:09:50: That's his point.

00:09:51: If you're still sketching while you're pouring concrete, you are guaranteed to be inefficient.

00:09:56: It sounds aspirational, but people are building the tools to visualize this.

00:10:01: Matt Yellen shared that he built this massive pro-core style dashboard for his development deals.

00:10:07: It integrates everything.

00:10:09: Schedules, AI action items, budgets.

00:10:12: Everything in one view.

00:10:13: It proves this spine is possible for individual developers, not just the mega firms.

00:10:19: And that is a perfect transition to the business side of all this.

00:10:21: We've talked about the tech and the data, but who is actually paying for it?

00:10:24: and are the companies selling it surviving?

00:10:26: That's our third theme, the contact ecosystem.

00:10:30: And there's been a lot of anxiety lately.

00:10:32: We've seen layoffs at big players like Autodesk and Procore in recent times.

00:10:36: And it's easy to read those headlines and think the bubble has burst.

00:10:39: People get scared.

00:10:40: But Scott Hunter and Joe Coleman argue it's not a burst.

00:10:44: They say it's market discipline.

00:10:46: I agree with them completely.

00:10:47: The era of growth at all costs, where you could raise fifty million dollars just by saying construction and AI in the same sentence, that's over.

00:10:57: Buyers are less patient now.

00:10:59: They

00:10:59: are.

00:10:59: They don't want a cool feature.

00:11:01: They want domain experience.

00:11:02: They want tools that solve that ring bob problem, not just tools that make pretty visualizations.

00:11:08: And there's plenty of evidence.

00:11:09: the market is still very healthy.

00:11:11: Let me look at equipment share.

00:11:12: That's the success story of the week for sure.

00:11:14: Paul Wright and Patrick Hellerman were celebrating their IPO.

00:11:18: Why is that one IPO so important though?

00:11:20: Because liquidity breeds confidence.

00:11:22: The sector has raised over forty billion dollars in venture capital.

00:11:27: Investors need to see that they can actually exit these investments and make money.

00:11:31: So equipment share going public validates the whole thesis.

00:11:34: It does.

00:11:35: It keeps the investment tap flowing for the next generation of startups.

00:11:38: And that investment isn't just happening in Silicon Valley anymore.

00:11:41: No, we saw some really strong signals from the Middle East.

00:11:44: Dubai is fascinating right now.

00:11:46: Konrad Aptoff and Ahmed Al-Khatib were reporting on the launch of Contek Valley.

00:11:52: And it seems like they're taking a very different approach over there.

00:11:54: They are.

00:11:55: In the West, It's so fragmented.

00:11:58: You have lots of startups fighting for attention.

00:12:01: Dubai is treating contact as infrastructure.

00:12:05: Regulators, developers, and tech providers are all aligned in one government-backed ecosystem.

00:12:10: Wow.

00:12:11: They're basically saying the technology has to be baked in from the permit phase.

00:12:14: That's the idea.

00:12:15: And that alignment is the dream, right?

00:12:17: It removes so much friction.

00:12:19: But even with all that support, there's one thing that kills adoption every single time.

00:12:24: Culture.

00:12:24: James Swanson hit the nail on the head.

00:12:26: He said, a company's culture is the number one factor.

00:12:29: You can have the best AI, the cleanest data, but if your culture punishes experimentation or just relies on the way we've always done it.

00:12:37: It'll fail.

00:12:37: It will fail.

00:12:38: And the workforce itself is changing.

00:12:40: Derek W. made a prediction that twenty twenty six is the year of the construction outsider.

00:12:45: I love this concept.

00:12:46: Me too.

00:12:47: For decades, the badge of honor was I've been on the tools for twenty years.

00:12:52: And that experience is still vital.

00:12:53: Don't get me wrong.

00:12:54: Of course.

00:12:55: But Derek is arguing we now need to value technical expertise just as much.

00:12:59: Data science, systems architecture, process engineering.

00:13:03: We need those hybrid profiles.

00:13:05: Yeah.

00:13:05: People who respect the concrete, but also understand the code.

00:13:08: Exactly.

00:13:08: We need translators.

00:13:09: But we can't talk about the workforce without acknowledging the human cost of this industry.

00:13:15: No, we can't.

00:13:16: Zulkarnane Malik shared a very serious post about the construction industry alliance for suicide prevention.

00:13:21: It's a sobering reminder.

00:13:23: We talk about efficiency and speed.

00:13:25: But this industry has some of the highest suicide rates of any sector.

00:13:29: The pressure, the long hours.

00:13:32: The isolation on remote sites, it puts all this tech talk into perspective.

00:13:36: Supporting the workforce isn't just about an exoskeleton or an iPad, it's about mental health.

00:13:41: Absolutely.

00:13:42: If we don't protect the people, none of the technology matters.

00:13:45: That's the bottom line.

00:13:46: A crucial point to keep in mind.

00:13:48: Okay, let's zoom out to the macro level.

00:13:51: We have the tech, the data, the people.

00:13:54: What are we actually building?

00:13:56: and this brings us to our final theme?

00:13:58: Infrastructure and housing.

00:14:00: and there is a massive massive shift happening.

00:14:03: I just Hussein shake shared a statistic that just it blew my mind.

00:14:06: was it?

00:14:07: data center construction is about to surpass office building construction in the US?

00:14:11: Wow That's a staggeration.

00:14:13: It feels like the physical manifestation of all the AI we've started this deep dive talking about.

00:14:17: It is.

00:14:17: AI isn't just code.

00:14:19: It lives in physical servers.

00:14:20: It consumes massive amounts of power and space.

00:14:23: Yeah.

00:14:23: Cloud is actually concrete and steel, and it is heavy.

00:14:26: And these aren't just big warehouses anymore.

00:14:29: Philip Townsend and Oliver Davis pointed out that they require industrialized builds.

00:14:34: You can't stick build a hyperscale data center.

00:14:37: The timelines are just too tight.

00:14:38: They need prefabrication, modular systems, huge energy strategies.

00:14:44: The PACE Group partnering with MACE for a hundred and sixty-eight megawatt portfolio.

00:14:49: I mean, that's enough to power a small city.

00:14:51: So the construction of the facility and the energy strategy are now inseparable.

00:14:55: Completely.

00:14:56: So on one hand, we have these massive industrialized data fortresses.

00:15:00: And on the other, we have the housing market.

00:15:02: Daniel Tides raised a really important contradiction there.

00:15:04: The density paradox.

00:15:06: Right.

00:15:06: Cities are getting denser, which should be good.

00:15:08: But he points out developers are building way too many one and two bedroom units.

00:15:12: We're squeezing families out.

00:15:14: It's a short-term financial place.

00:15:15: Smaller units get higher rent per square foot, but it creates a long-term demographic crisis.

00:15:21: If you only build for singles and couples, families move to the suburbs.

00:15:24: And you limit the sustainability of the city itself.

00:15:27: We have to align density with demographic reality.

00:15:31: It feels like, whether we're talking about AI models, data standards, or housing units, the theme really is just structured.

00:15:38: You can't just throw things at the wall anymore.

00:15:40: Yeah.

00:15:40: Whether it's an unstructured chat GPT query, a messy Ring Bob sticky note, or a city full of studio apartments.

00:15:47: That's the perfect way to wrap it up.

00:15:49: We're done with the hype.

00:15:50: We are done with the sketch phase where we just guess and hope for the best.

00:15:54: It's time to build the spine, the logical architecture that holds it all together.

00:15:58: Well, lots to think about there.

00:16:00: If you enjoyed this deep dive, remember that new episodes drop every two weeks.

00:16:04: Also, check out our other editions on smart manufacturing and digital power tools.

00:16:08: And remember, don't just collect data, build the spine.

00:16:12: And subscribe so you don't miss the next deep dive.

00:16:15: Thanks for listening.