BIM in 2026: The AI Revolution Needs Structured Data, and We've Had It All Along
Julien Benoit | 2026
There is something almost ironic about where the construction industry finds itself in 2026.
For more than a decade, the BIM community has been trying to convince the rest of the world that structured data matters. That modelling buildings with rich, semantic information, not just pretty 3D geometry, would eventually pay off. The pitch was solid and the evidence was there, yet adoption remained painfully slow, confined to the convinced talking to the already convinced, in conferences attended by people who did not need convincing.
Then AI arrived. And suddenly, everyone wants structured data. The same people who said BIM was too expensive, too complicated, not worth the effort now want to plug Large Language Models into their project workflows and expect magic to happen. It would be funny if it were not so predictable.
Everyone is Talking About AI. Almost Nobody is Talking About the Data.
You cannot scroll through LinkedIn for thirty seconds without hitting a post about how artificial intelligence is going to transform construction. Generative AI, predictive analytics, automated everything. The hype machine runs at full speed, and honestly, it is not entirely wrong. These tools are powerful, no question about that.
But here is the thing that nobody seems eager to discuss, probably because it does not make for a good keynote title and it is not particularly exciting to write about: AI is only as smart as the data you feed it.
This deserves to be said twice.
AI is only as smart as the data you feed it.
And in construction, the state of data is, to put it politely, a disaster. We are talking about PDFs of floor plans, scanned drawings that have survived coffee spills and three generations of photocopiers, Excel spreadsheets emailed back and forth until no one can tell which version is the current one, specifications written in prose that even a human struggles to interpret. Meeting minutes lost somewhere in an inbox. Handwritten notes on margins of printed plans that someone eventually forgot to digitize.
Now point a Large Language Model at this kind of data. What happens? Hallucinations. Confidently wrong answers. The AI equivalent of someone who read half a document and decided they understood the entire project. A recent academic review looking at BIM and LLM integration confirmed what anyone working in the field could have guessed: while around 70% of research prototypes show technically promising results, fewer than half have been validated in actual industrial settings. The gap between the laboratory and the construction site remains enormous. And the main reason for this gap is not that the AI models are insufficient. It is that the data is not there, or not in a state that can be exploited.
Genetic Flaw of BIM, and Why It Might Finally Stop Mattering
There is an uncomfortable truth that the BIM community needs to acknowledge, because it explains a great deal about why adoption has been so uneven despite decades of advocacy.
BIM has always been the right answer. But it has been answering the wrong question, or perhaps more accurately, answering the right question in a language that most of the industry could not understand.
For years, the construction world kept asking a very simple thing: “Why should we invest in this? What is the concrete return?” And the BIM community responded with technical demonstrations, interoperability frameworks, maturity matrices, and standards references. Specialists explaining to specialists, using concepts built by specialists, in tools designed for specialists. The echo chamber was comfortable, but it was still an echo chamber.
That is the genetic flaw of BIM. It was created by experts, for experts.
Working with BIM models, creating them, querying them, extracting genuine value from them, required a level of technical fluency that most construction professionals did not possess and, frankly, had no compelling reason to acquire. In that regard, BIM was not so different from learning a new programming language. Extremely powerful once you master it, but the learning curve was steep enough to discourage most people from even starting.
The consequences were predictable. Those who understood BIM valued it immensely. Everyone else tolerated it at best, ignored it entirely at worst. And the industry, collectively, never managed to answer the main question of CFOs, project directors and building owners: “Why should I care about this?”
But now something fundamental has changed.
For the first time, anyone can use natural language to interact with any structured data, including the kind that lives inside BIM models. A Large Language Model does not need to understand IFC schemas or know how to write IfcOpenShell scripts or navigate Revit APIs. It can take a question expressed in plain language, something like “what is the total weight of structural steel on level 3?” or “are all the fire-rated doors on this floor compliant with the current regulation?” and translate that question into a query against a structured data model.
The barrier that kept BIM locked inside a circle of experts is, for the first time, starting to dissolve. BIM itself did not become simpler, but the interface to BIM became radically more accessible. And this distinction matters enormously, because it means that the business value of structured data, which has always been there but was invisible to anyone who could not manipulate the tools directly, is now becoming apparent to people outside the specialist community: to the project managers who need to make decisions, to the finance teams evaluating risk, and to the building owners who want to understand what they actually own.
For the first time, these decision-makers can see BIM's effect on things they respond to: cost savings, risk reduction, automation, regulatory compliance, operational efficiency. All this without needing to become BIM experts themselves.
The BIM community has been struggling to make this case for over a decade. AI might just end up making it for them, almost by accident.
Open Standards: Not Optional, Not Negotiable
There is a condition, however, and it is a critical one.
None of this works if the data remains locked inside proprietary formats.
An AI system that can only query models from one vendor, in one vendor's format, using one vendor's ecosystem is not progress. It simply replaces one type of silo with another, slightly more sophisticated one. The entire promise of AI-powered BIM depends on data being interoperable, portable, and vendor-neutral. Without this, we are just automating within the same walls that were already limiting us.
This is why open standards matter more today than they ever have. IFC (Industry Foundation Classes) is the open international standard for BIM data, formalized as ISO 16739. It is not merely a 3D file format. It is a structured data model that codifies identity, semantics, attributes, and relationships between building elements. Every wall knows it is a wall. Every door knows which wall it belongs to and what its fire rating is. Every space knows its boundaries, its function, its classification. The format is vendor-neutral, platform-agnostic, and was designed from the ground up with interoperability in mind.
Around this format, the buildingSMART ecosystem provides a set of complementary standards: IDS for information requirements, BCF for issue management, bSDD for data dictionaries. Together, they form the plumbing that allows data to flow between tools, between disciplines, and across the different phases of a building's lifecycle. More than 30 countries now mandate BIM on public infrastructure projects, and a growing number specifically require open formats for regulatory submissions.
In a world where AI tools will need to consume data from multiple sources, across multiple disciplines and throughout an asset's entire life, proprietary formats are not only inconvenient, but a strategic liability. Open standards have stopped being a nice-to-have. They are the non-negotiable foundation.
The BIM Comeback That Nobody Expected
For years, the complaints about BIM were consistent and, in some cases, legitimate. Too expensive to implement. Too complicated for most teams. Slows down the early project phases without providing visible enough returns. Why invest in all this data richness when we just need to build the thing?
And now, in 2026, everyone wants AI, automation, predictive analytics, automated regulatory compliance checking and intelligent maintenance scheduling. Every single one of these applications requires the same foundation: structured data.
The companies that invested in BIM early, that did the difficult, unglamorous, often thankless work of building clean, rich, well-structured models, are the ones who find themselves ready for this moment.
The companies that bypassed BIM, that stayed with 2D documentation and disconnected spreadsheets, are now discovering that they are not merely behind on one technology. They lack the foundational layer that makes the next generation of technologies possible.
The market numbers reflect this realization. Global BIM spending is projected to grow from roughly $4.7 billion in 2025 to over $5.4 billion in 2026. Around 65% of construction projects worldwide now incorporate BIM workflows in some form. But these adoption numbers, while encouraging, tell only part of the story. The question that actually matters is not whether a company uses BIM. It is whether the data inside their BIM models is rich enough, structured enough, and open enough to be genuinely useful for downstream applications such as AI, automation and lifecycle management.
Vertical Transportation: Where Structured Data Gets Concrete
The elevator industry offers a particularly clear illustration of why these ideas matter in practice, and not only in theory.
An elevator is, in many ways, one of the most complex products that exists inside a building. It involves mechanical, electrical, structural, and safety components that must all function together with precision. The configuration space is vast: shaft dimensions, car sizes, door types, speed and load requirements, compliance with national codes that vary from country to country. A single incorrect parameter can lead to consequences ranging from expensive rework to serious safety concerns.
But the complexity does not stop at the product itself. An elevator does not exist in isolation. It lives within a building that was designed by architects, engineered by structural consultants, coordinated with MEP systems. The elevator data needs to flow seamlessly into and out of the building model.
When the elevator configurator operates disconnected from the rest of the project data, every handoff between stakeholders becomes a manual re-entry exercise. And with every manual re-entry, as anyone who has worked on a construction project knows, comes an opportunity for error.
This is precisely where native BIM tools, where data is the foundation and not an afterthought, make a fundamental difference. Not tools that can export to IFC when asked, but tools that think in structured data from the very first interaction. Where the elevator model is not a drawing that gets converted at the end, but a rich, parametric, standards-compliant object from the moment it is created. A model that participates in the building's data ecosystem, rather than sitting apart and waiting to be translated.
Integration into the building's data ecosystem is the first dimension of what a native BIM tool has to deliver. The second is maturity evolution. A BIM model is not a static snapshot. It is, or it should be, a living thing that grows with the project. In the early stages, the elevator might be nothing more than a rough concept, a placeholder that says, “we need a lift here, approximately this capacity, approximately these dimensions.” As the project advances, decisions are made, specifications tighten, and the model matures progressively. What started as a conceptual placeholder becomes, step by step, a fully detailed product configuration with precise components, exact dimensions, and complete performance data.
A good native BIM tool supports this progression. It provides control over the maturity of the model at each stage, from early design through to final production specification, without requiring the user to restart the process when the level of detail increases. The model evolves. The data becomes richer. And at every point, it remains structured, queryable, and connected to the broader project context.
From Design to Production: Where Parametric Data Closes the Gap
When a design tool is natively parametric, the relationship between the model and the manufactured product changes fundamentally. The model ceases to be documentation and becomes the driver. A parameter changes, and the downstream consequences propagate: the shaft layout adjusts, the structural requirements update, the bill of materials regenerates, the production specifications follow.
For elevator manufacturers, this represents the bridge between what the architect specified and what the factory will actually produce. And this bridge depends entirely on structured data. Without it, information passes through a series of translations, from model to drawing to specification and further to production order, and at each translation step, there is a risk of information loss or corruption. With a native BIM approach, the data flows continuously. Each stage adds information; none destroys it. And because the data is parametric, when the architect modifies the shaft dimensions (as architects will inevitably do), the downstream impact becomes visible immediately, not three weeks later when someone spots the discrepancy on the construction site.
Maintenance: Where the Long-Term Value Actually Lives
There is something that the construction industry, broadly speaking, tends to underestimate but that the elevator industry understands very well: the real economic value is not in the initial installation. It is in the decades of maintenance that follow.
Elevator manufacturers and service companies generate the most significant portion of their revenue from maintenance contracts. Keeping an elevator running safely and efficiently for 25 or 30 years is the actual business. And this is where well-structured, well-maintained models become extraordinarily valuable, in ways that are only beginning to be fully appreciated.
Consider what a model can become over the operational life of an elevator. At installation, it contains the exact as-built configuration: every component, every specification, every parameter that was selected. Then, progressively over the years, it starts accumulating operational data, from sensor readings and service intervention logs to component replacements, performance trends and inspection records. Each piece of information attaches to the structured model, because the model provides the context that gives the data meaning.
This is where the Internet of Things stops being an abstract concept and becomes useful. IoT generates enormous volumes of data, from temperature readings and vibration patterns to usage cycles and energy consumption. But raw data without structure is noise. A temperature reading is meaningless unless you know which component it refers to, what the normal operating range is for that component in that specific configuration, and how that reading relates to the maintenance history of the system. The structured model, the one that has been evolving and accumulating information since the design phase, is what transforms raw sensor data into intelligence that can actually be acted upon.
You might think that predicting maintenance needs, diagnosing errors remotely or even anticipating failures before they happen is a delusional, futuristic vision. It is not. These are practical applications of how structured data can connect the physical asset to its digital representation and maintain that connection across decades of operation. The model that began its life as a design tool becomes, progressively, the portal through which the entire operational existence of the elevator is monitored and managed.
But this only works if the data was structured properly from the start, that is, if the model was built natively, in open formats and with the right level of information at each phase of the project. Attempting to retrofit structure onto years of unstructured data is not impossible, but it is exponentially more difficult and costly than getting it right from the beginning.
A Necessary Reality Check
What Comes Next
The gap between data-mature and data-immature companies will widen quickly, and AI acts as an amplifier: it multiplies whatever advantage structured data already provides. Companies with solid data foundations will become faster, more efficient and more competitive. Companies without them will discover that they cannot effectively use the same tools their competitors rely on daily.
Open standards will become increasingly critical, not less. In a world where AI agents need to query building data across disciplines, tools and lifecycle phases, proprietary lock-in ceases to be merely an annoyance. It becomes a genuine competitive disadvantage.
Specialized verticals such as elevator design, MEP engineering and structural analysis are likely to lead the adoption curve. They do not command larger budgets than the rest of the industry, but their domains are well-defined, their data requirements are precise, and the return on structured, parametric, lifecycle-oriented BIM is immediate and measurable.
And perhaps most significantly, the question that the industry has been struggling with for so many years, “BIM, yes, but why exactly? Convince me.”, might finally receive an answer that resonates beyond the community of specialists. The fundamental argument has not changed, but AI has, for the first time, made the value of structured data visible and accessible to the people who needed to see it most.
BIM's moment is arriving. It simply took a different technology to make it obvious.
ABOUT THE AUTHOR
Julien Benoit is VP Customer Success at DigiPara, which develops native BIM software for the Vertical Transportation industry. He writes occasionally, when he has something to say.
ABOUT DIGIPARA
DigiPara GmbH is the global leader in elevator CAD and BIM software solutions, headquartered in Germany and deployed in more than 130 countries. Its flagship products, DigiPara Liftdesigner and DigiPara Elevatorarchitect, enable automated 2D drawing production, parametric 3D BIM generation, and ERP-integrated order processing for elevator manufacturers, vertical transportation consultants, and architects worldwide. For over two decades, DigiPara has been the technical infrastructure behind BIM quality in vertical transportation.