How does structured data from open standards like IFC improve AI outcomes in construction?
AI systems are only as good as the data they are trained on. When construction information environments are fragmented, inconsistent, and ungoverned, AI does not fix the problem — it scales it, making poor data harder to recover from. Conversely, when the industry uses structured data and consistent classification frameworks like IFC, AI has a reliable foundation to deliver genuinely useful results: it can correctly identify that a model element is a door or a window because the schema defines it explicitly.
This is not a theoretical concern. As Eric Bugeja explains in the Inside buildingSMART — The Future of IFC episode of the Digital Construction Podcast, bad or unstructured data fed into an AI model produces bad outputs — and the more automated the process, the faster those errors compound. The construction industry's longstanding investment in governed workflows and structured classification actually gives it an advantage over many other sectors when it comes to putting AI to meaningful use.
IFC files, for instance, are designed around a schema that makes every element's type, properties, and relationships machine-readable. That explicitness is precisely what AI needs to reason about a building model reliably. buildingSMART International — the body stewardring these open standards across its 40 chapters worldwide — has been building this data governance infrastructure since the late 1990s, initially to enable information sharing across software platforms. That foundation is now paying dividends as AI enters the picture.
What is IFC?IFC (Industry Foundation Classes) is an open, ISO-certified data schema developed and maintained by buildingSMART International. It defines construction model elements — walls, doors, windows, spaces — in a machine-readable format that any conforming software can read and write. Because the schema is explicit and vendor-neutral, IFC files are designed to remain accessible for 30 years or more without requiring a specific software version, making them a durable basis for long-term data governance and, increasingly, AI-driven workflows.
"AI systems are only as good as the data that they are trained on. If our information environments are fragmented and inconsistent and ungoverned, then AI doesn't solve that problem. It actually scales it."
— Eric Bugeja, Chairperson, buildingSMART Australasia
About Eric Bugeja
Eric Bugeja
Chairperson, buildingSMART Australasia & Director, buildingSMART International · Founder, Kefron
Eric Bugeja began his career as a mechanical engineer in an unusually formative setting: a two-man consultancy where a highly experienced Dutch engineer gave him complete freedom to explore emerging design technologies. That early autonomy shaped a career-long orientation toward digital innovation — he completed a 3D design task in two days that had been allocated two weeks, an early signal of what structured, technology-driven workflows could achieve.
He subsequently worked across manufacturing automation and food processing consultancy before moving into BIM leadership roles at large design and construction firms, where he developed digital strategy and drove structured data adoption at scale. This operational background — spanning engineering practice, software implementation, and organisational change — is what underpins his authority on the relationship between data quality and AI performance.
Bugeja now runs Kefron, his own consulting and training business focused on helping organisations adopt open standards and digital engineering practices. In parallel, he serves voluntarily as Chairperson of buildingSMART Australasia and as a Director of buildingSMART International, the global body responsible for IFC, IDS, BCF, and related open standards across 40 chapters worldwide. His work sits at the intersection of standards governance, technology adoption, and industry advocacy — making him one of the most informed voices on what the construction sector's data infrastructure can and cannot yet deliver to AI systems.
See also
buildingSMART is actively working on the next generation of open standards that moves away from file-based, fixed packets of information toward cloud-based, continuously accessible data environments.
A common mistake is that designers simply export an IFC file with no specific instructions, resulting in massive, unmanageable files containing every piece of data in the model rather than only what the recipient actually needs.
If one or two software vendors become dominant and an organisation's data is locked into their proprietary format, that organisation faces serious risk if the vendor changes pricing, discontinues a product, or shifts strategy.