In the Age of AI, the winner won't be the best clicker - it'll be whoever understands information
For the last twenty years, the BIM market rewarded one skill above all others: fluency in software. Whoever modelled faster, knew more features and mastered the next plugin was the valuable one. It was a world where you built your edge with clicks.
That world is ending. And not because the tools are disappearing because they are ceasing to be scarce.
Artificial intelligence radically lowers the cost of "operating a tool". It understands natural language, writes code, and uses external applications and data. Things that once required a specialist clicking inside a specific program can increasingly be described in a sentence. And if that's the case, value moves elsewhere. Not to whoever knows how to use the tool, but to whoever understands the information the tool works on.
In infrastructure, that information is openBIM. And that is exactly why skills around IFC 4.3 don't lose relevance in the age of AI, they gain it.
What actually changes
It's worth separating the facts from the hype. AI does not "do BIM for us". AI changes one very specific thing: the way we work with information.
Three breakthroughs matter here:
- Natural language: you can talk to the model and the data instead of learning an interface.
- Code generation: the analyses, queries and reports that a programmer used to write now appear on demand.
- Using tools and data: AI can reach into a model, a property database or a standard.
It sounds like the end of the need for expert knowledge. It's exactly the opposite. To ask a model anything meaningful, you have to know what a road, an alignment, a spatial structure, a property set, a classification and a coordinate reference system actually are. AI can write a query but it doesn't know what's worth asking, or whether the answer makes sense. That is still human knowledge.
Why software isn't the real challenge
Every program you learn today will look different in five years, or disappear. Interfaces, shortcuts, features, that's the most fleeting layer. Investing your whole professional identity in one application is like building a house on someone else's land.
What does not age so quickly is understanding how infrastructure information should be designed, structured and exchanged. Because that isn't a feature of a program. It's the foundation on which all programs, formats and, more recently, all AI-driven automation stand.
The challenge is not learning yet another application. The challenge is understanding how to design, structure and exchange information.
AI won't fix bad data. It exposes it
The most common misconception goes: "we'll roll out AI and it will tidy up our data for us." It won't. AI works with whatever it's given. If the model is poorly structured, you'll get an answer confident and wrong.
In project practice, the problem isn't a lack of AI, but the information itself:
- properties recorded inconsistently, everyone their own way,
- wrong or missing object classes,
- no semantics and no relationships between elements,
- no defined model purpose (what is it actually for),
- georeferencing problems,
- models prepared as geometry only.
Point the best AI at a model like that and you'll get an impressive demo, not a reliable process. In that sense AI is a mirror, it mercilessly reveals how much the structure of your data is really worth.
The layer AI actually runs on
Think of information in infrastructure as a stack of layers:
AI is only the last layer here. It makes an impression because it's visible. But its reliability depends entirely on the layers beneath it. Without well-structured IFC, correct classifications and a clear model purpose, the top layer has nothing to stand on.
That's why openBIM and IFC 4.3 are ceasing to be "a format to hand over at the end of the project". They are becoming a data layer on which automation is built. And whoever understands that layer controls what happens above it.
The quiet shift: from exchanging files to designing information
A generational shift in how we approach openBIM is happening right now - quiet enough that many companies miss it.
| Old-school openBIM | New-school openBIM |
|---|---|
| IFC exported at the end of the process | IFC as a structured source of information |
| Opening a file in a viewer | A model built for a specific use |
| Visual model checking | Automated queries and checks |
| Requirements written in documents | Machine-readable requirements |
| IFC as a by-product | IFC as a data layer |
| Manual reporting | Data-driven reporting |
| The viewer as the interface | AI as the interface to information |
Old-school openBIM focused on exchanging files. New-school openBIM focuses on designing information that humans, software and AI can understand.
IFC Infrastructure Masterclass
Want to truly understand IFC in infrastructure?
If you want to go deeper into IFC 4.3, openBIM, infrastructure models and the practical use of data in real projects, take a look at the IFC Infrastructure Masterclass. It's a course for people who don't just want to export models, but to understand how to make data actually work.
Explore the masterclassThe skills that are gaining value right now
If the edge is moving from "operating a tool" to "understanding information", it's worth knowing which skills exactly are gaining value. In infrastructure, these are:
- Understanding infrastructure and its ecosystem - how it differs from vertical construction, who the stakeholders are, what the project lifecycle looks like.
- Model purpose - what data is needed at each stage and for which use: coordination, quantity take-off, construction, asset management.
- IFC 4.3 - objects, classes, infrastructure entities (IfcRoad, IfcBridge, IfcRailway, IfcAlignment), spatial structure and relationships.
- Spatial structure and model organization - so the data can be navigated and queried.
- Properties and classifications - consistent property sets and classifications (including bSDD) that make automation possible.
- Georeferencing and coordinate systems - the foundation you only remember once it fails.
- Information quality - the ability to define and verify whether a model is truly ready to use.
- Moving from standard to implementation - because a standard nobody implements changes nothing.
Notice that none of these skills is tied to a specific piece of software. They are all tied to information. And that is precisely why they will survive the next wave of tools, including AI.
Two types of professionals in the age of AI
The market is splitting into two groups, and AI only deepens the divide.
The first group consumes applications. It waits for the vendor to add a feature, clicks on what's already there, and is only as good as the tool it happens to have. In a world where tools become a commodity, this group's value falls.
The second group designs workflows. It understands the information layer well enough to assemble a process from building blocks IFC, a processing library, a viewer, a report, AI for a specific goal. This group doesn't wait for a feature. It defines it.
AI doesn't replace the second group. AI is a tool in its hands. The whole difference comes down to which side of that line you want to be on three years from now.
Where to start, without big licences
The good news: to get into modern openBIM you don't need expensive software to begin with. The path is simpler than it seems, and it fits into three steps:
- View - open and inspect IFC (free viewers are more than enough).
- Check - verify properties, classes, georeferencing and structure.
- Automate - generate reports and queries; this is where AI comes in.
Getting into openBIM isn't a question of a licence budget. It's a question of understanding what to ask.
Information is the new advantage
AI hasn't devalued knowledge of openBIM and IFC. It has done the opposite, it has shown how valuable well-structured information is, and how helpless even the best tool is without it. The people who truly understand IFC 4.3 and openBIM will design the workflows of the future, instead of merely using ready-made applications.
That's not a skill you pick up by accident. It's a structured path.
