Best of LinkedIn: Digital Construction CW 26/ 27

00:00:00: This episode is provided by Thomas Allgaier and Frennus, based on the most relevant LinkedIn posts about digital construction in calendar weeks twenty-six and twenty seven.

00:00:09: Frenness is a BDB market research company that supports industrial automation and ICT companies with market intelligence across the construction industry to prioritize segments identify high value accounts and validate use cases.

00:00:22: you can find more info.

00:00:26: So we're talking about a massive issue today.

00:00:28: Every year, the global construction industry loses um...about one hundred and seventy seven billion dollars to rework.

00:00:35: Yeah!

00:00:35: Billion with a B Right

00:00:37: A massive number.

00:00:38: And it isn't because suddenly forgot how to pour concrete or you know frame a wall, it's really because our project data is trapped in these disconnected silos.

00:00:47: Exactly

00:00:47: So today for you listening we're cutting through all the noise.

00:00:50: We are looking at most critical insights from calendar weeks twenty-six and twenty seven to see how smartest minds of industry actually trying fix this.

00:00:57: Yeah

00:00:58: basically tracing that digital thread.

00:01:00: start with foundation which is data quality and AI readiness

00:01:04: The absolute core.

00:01:07: Then we'll move into AI agents and workflow automation, shift over to BIM and lifecycle intelligence.

00:01:12: And finally look at robotics in field implementation.

00:01:15: you know how this actually hits the dirt.

00:01:17: I love that.

00:01:18: And you know, if you're in the industry... ...you know exactly what that siloed data pane feels like?

00:01:22: Oh absolutely!

00:01:23: You've probably spent hours trying to cross-reference a floor plan with material schedule updated by subcontractor three days ago.

00:01:31: Of

00:01:31: course they didn't notify anyone

00:01:33: Exactly.

00:01:33: it's total nightmare.

00:01:35: But before we can even talk about all of the flashy AI tools We really have face hard truth here.

00:01:40: Yeah I mean A.I just completely fails.

00:01:43: If underlying data is trash

00:01:45: It does, it's the single most critical bottleneck.

00:01:48: We have all these incredible tools hitting the market.

00:01:50: but Rudy Mara pointed out recently that um The biggest barrier to AI in construction is not a lack of technology right?

00:01:58: Its messy data.

00:01:59: if you feed an AI unorganized drawings and inconsistent file names You don't get magical solutions.

00:02:06: You just get what he calls faster chaos

00:02:09: Faster Chaos.

00:02:10: I love that phrasing.

00:02:11: great wait let me push back on there for a second.

00:02:13: sure Isn't the entire promise of these large language models that they can, like take unstructured messy data and just make sense it for us?

00:02:21: Well.

00:02:21: That's common misconception.

00:02:23: right?

00:02:24: AI can parse text sure but think about reality.

00:02:27: a complex project okay.

00:02:29: you've got Project Info scatter across maybe five different proprietary systems And They fundamentally refused to talk with each other.

00:02:37: Marabh argues that feeding that specific brand of chaos into AI just accelerates the mistakes.

00:02:44: Oh wow, okay Yeah That makes sense.

00:02:45: So if you want an AI to summarize a clash report but your team doesn't even structure Clash reports consistently The AI is going to confidently generate and inaccurate summary exactly.

00:02:54: it Just speeds up the wrong answer.

00:02:56: which

00:02:56: brings up Florian humor's warning about digital twins.

00:02:59: Because you know Digital twin is supposed to be this perfect replica right?

00:03:03: A living digital model yeah.

00:03:05: But humor says that without explicit metadata, a digital twin just becomes this unsearchable black hole.

00:03:12: A Black Hole?

00:03:13: That is such a good way to put it!

00:03:15: You pour information in but you can never actually get anything out.

00:03:19: so he lists five pillars.

00:03:21: every ingestion schema needs context lineage relationships governance and discoverability.

00:03:27: Those are non-negotiable.

00:03:29: And let's just um pause on relationships for a second because Guido Machiochi explained this perfectly with the simplest example possible

00:03:37: A door.

00:03:38: Okay, walk me through the door example.

00:03:39: So on a real project... ...a human knows that a door-on-a-floor plan The same door in an Excel schedule and the door in a PDF spec sheet are all the exact same physical object.

00:03:50: Because we have that intuitive context, right?

00:03:52: Precisely.

00:03:53: But AI flattened structured AEC data.

00:03:56: Standard AI pipelines just lose those threads entirely.

00:03:59: They don't understand the relationship between two D drawing And text schedule.

00:04:04: So basically trying to run AI on generic data is like, um... Trying to run a high-speed bullet train onto dirt road.

00:04:10: That's exactly what it's like!

00:04:11: It just gonna derail.

00:04:13: And this ties right into what Li Xiaowang highlighted about Palantir and McCarthy.

00:04:16: They have the new Pulse AI operating system

00:04:19: Yes The Pulse System

00:04:20: Right, and Palantire isn't bringing AI models.

00:04:23: they are bringing what we call Data Ontology.

00:04:26: It's a connected datagraph to solve that exact hundred and seventy seven billion dollar rework problem.

00:04:32: It maps the relationship so data isn't siloed anymore.

00:04:36: Exactly, they are laying down actual tracks for the train can run.

00:04:40: So now we have those tracks laid.

00:04:42: what do trains actually look like today?

00:04:44: Well, this is where we get into AI agents and workflow automation.

00:04:48: And Vragan Paramanantham had this really fascinating conversation with the CEO of OpenSpace, Givon Kalaneethi... Oh

00:04:56: yeah I saw that!

00:04:57: They discussed his concept called Jagged Intelligence.

00:04:59: Jagged intelligence?

00:05:00: What does it mean exactly in

00:05:02: practice?!

00:05:02: It means AI can be those genius polymath one second writing complex code…and then literally have a logic for a child next.

00:05:11: That's terrifying for a project manager.

00:05:13: Right,

00:05:13: and Rodin broke down the math on this which is just wild if you have an AI that is ninety-five percent reliable

00:05:18: Which sounds great honestly.

00:05:20: Right!

00:05:21: Ninety five percent sounds amazing but If it's doing a five step workflow autonomously...that success rate drops to seventy seven percent.

00:05:29: Oh wow And at twenty steps It plummets To thirty six percent.

00:05:33: Wait

00:05:33: really?

00:05:34: Thirty-six percent.

00:05:35: That's exactly why this industry needs hyper specific bounded agents, not just some everything agent that promises to run your whole job site.

00:05:43: Exactly!

00:05:44: Reliability beats magic every single time

00:05:47: Like an issue review agent who only does one thing.

00:05:49: really well

00:05:49: Yes And we're starting.

00:05:51: see those bounded tools pop up.

00:05:52: Irving Resendiz posted about the Atenea Rivet plugin and it is so accessible.

00:05:58: Accessible how?

00:06:00: Like easy to install.

00:06:01: It installs in under five minutes, and you don't need to know any scripting.

00:06:04: You just use plain language to generate schedules and tags right inside Revit.

00:06:07: That's huge!

00:06:09: And then David Campbell showed what happens when you start chaining these tools together with EZ-DD.

00:06:14: He uses things called sub MCPs.

00:06:16: Right the model context protocol

00:06:18: Yeah Inside Civil-DD and Revit.

00:06:20: So the AI can actually pull a site plan from ACC, import it and verify against field data all from one single natural language prompt.

00:06:30: It's orchestrating multiple tools simultaneously, which is a total game changer...

00:06:35: it is!

00:06:35: ...and Amanal Samahi with build metrics AI is taking this even further, he's saying the days of passive dashboards are completely over.

00:06:44: Oh totally.

00:06:44: a passive dashboard is basically rear view mirror exactly.

00:06:47: his system doesn't just tell you that your over budget.

00:06:49: it actually predicts the impact prioritizes recovery actions and then assign them directly to stakeholders.

00:06:55: so its actively managing problem not reporting which makes for perfect transition.

00:07:00: with AI handling all heavy lifting is shifting.

00:07:05: It really is, it's moving away from just pretty three-D pictures to actual constructability and lifecycle value.

00:07:11: And Jay Vaishnav made this crucial distinction.

00:07:13: he asked Is your model coordinated in the software?

00:07:16: Or has it coordinated for construction?

00:07:18: That is such an important question because a clash remodel is totally useless if it ignores real world installation sequences

00:07:26: Right or like material procurement lead times or The fact that a human actually needs physical clearance to turn a wrench

00:07:33: exactly If the worker can't reach the pipe, then digital model doesn't matter.

00:07:39: It's like a perfectly spell-checked essay!

00:07:42: The spelling is flawless and grammar perfect but if plot makes absolutely no sense... It's still a terrible story.

00:07:49: I love that analogy, it's so true.

00:07:51: and Lugman Comben had a viral post about why people struggle to execute this.

00:07:56: he broke down seven different BIM career paths.

00:07:59: oh i saw that most people just get stuck at bim coordinator.

00:08:01: yeah

00:08:02: they get completely bottlenecked there because they don't even realize that roles like vdc manager or information manager actually exist.

00:08:09: right the people who tie the model to the actual site.

00:08:11: logistic exactly

00:08:12: the broader lifecycle rules

00:08:14: then life cycle intelligence mandated now globally.

00:08:18: Really?

00:08:19: Where are we seeing that?

00:08:20: Well,

00:08:20: Ngoc Bin shared an update on Vietnam's New Decree to Seventeen.

00:08:25: They're making BIM absolutely mandatory for grade two and above projects.

00:08:29: Wow!

00:08:30: Mandatory Yeah.

00:08:32: And MGSudin is pushing for similar mandates in West Bengal.

00:08:36: It's not just a save money.

00:08:37: it literally prevents structural failures and reduce sight accidents.

00:08:42: That

00:08:42: makes sense.

00:08:43: you simulate the build of life

00:08:45: Exactly.

00:08:46: But this leads us to the ultimate question, right?

00:08:49: How does all of these brilliant digital planning actually survive contact with mud and dirt in a physical job site.

00:08:56: The physical reality is it's an ultimate test!

00:08:59: And Jason Smith had some great insights on drones for that exact issue.

00:09:03: Drones used to be just expensive hobby sites.

00:09:05: honestly Right

00:09:06: but not anymore.

00:09:07: They are actively eliminating guesswork.

00:09:09: Smith says they're saving project managers an average of fourteen hours a week by verifying what he calls phantom progress.

00:09:16: Phantom Progress, so like when a sub says that they moved the dirt but didn't?

00:09:19: Exactly!

00:09:20: The drone flies over and gets undeniable volumetric measurements.

00:09:23: That's incredible.

00:09:25: And Akshath Reddy is taking it even further with TravTech.

00:09:28: at Georgia Tech They deployed Boston Dynamics spot Robot

00:09:32: The robot dog

00:09:33: Yeah, the robot dog.

00:09:34: They paired it with drones AI and the live BIM model to create this fully autonomous, live-site monitoring ecosystem.

00:09:42: That is just brilliant engineering.

00:09:44: but you know getting a robot onto a traditional job site isn't really an engineering problem.

00:09:50: No what is it then?

00:09:51: It's

00:09:51: business problems!

00:09:52: Contractors hate risk Which is why Michael Sharon's post about automated architecture, or AURR was so smart.

00:10:00: Oh

00:10:00: they build those robotic micro factories for home building right?

00:10:03: Yes

00:10:04: and Bridget Hipwell from AUIR explained how to actually get these change-averse contractors To hire their robots.

00:10:10: Do they pitch them some massive tech revolution?

00:10:13: Not at

00:10:13: all.

00:10:13: They don't reinvent the wheel.

00:10:14: They use the exact same standard contract The builder already uses For human subcontractors.

00:10:19: Well

00:10:20: that's smart...They

00:10:21: literally just add an appendix.

00:10:23: That is pure psychology.

00:10:25: I mean, tech companies always fail in construction because they try to force their own software playbooks on an industry that buys things very slowly...

00:10:34: Exactly.

00:10:35: ...adapting the industries' existing paperwork.

00:10:38: just sliding a robot into a subcontractor slot?

00:10:41: That's genius!

00:10:42: It

00:10:42: removes all of friction.

00:10:44: Well, before we wrap up I want to leave the listener with a final thought.

00:10:47: We've covered a lot of ground but there's a bigger picture here.

00:10:49: There really is and i wanna leave you The Listener With A Thought based on a post by Felix Irumbin.

00:10:55: He pointed out that construction industry suffers from this massive resource inversion.

00:11:00: Resource inversion?

00:11:01: What does it mean for project life cycle?

00:11:03: well think about It!

00:11:04: We allocate our best people Our strictest governance And are most intensive systems entirely To the execution phase Right

00:11:12: when were actually building.

00:11:13: But during execution, our ability to actually influence the outcome is near zero.

00:11:18: Moving a concrete wall costs of fortune.

00:11:21: Yeah it's too late by then

00:11:22: Exactly.

00:11:23: but the data shows that A three-to five percent investment in front end planning yields up to a twenty percent return.

00:11:30: Wow!

00:11:30: Twenty percent.

00:11:32: So the provocative question for you was this Are you managing the concept phase where this project is actually decided?

00:11:39: Or are you just managing the execution phase, when decisions are merely

00:11:53: revealed.