The construction and real estate industries have spent a decade trying to digitalize existing buildings, with poor results. Several startups worldwide have attempted to automate the creation of digital twins for the built environment, yet the vast majority of the world’s 2.4 billion buildings remain what the industry calls “dark assets,” structures with no usable digital data attached to them.
The urgency is structural. Those 2.4 billion buildings represent roughly 80% of the structures that will still be standing in 2050, and most of them will need to be retrofitted rather than replaced. Yet approximately 90% currently have no digital record at all. Creating that data today is a fully manual process, slow, expensive, and inconsistent in quality.
Scan-to-Data
The traditional approach to digitalizing an existing building follows a familiar path: a surveyor captures a laser scan, producing a point cloud, a three-dimensional map of surfaces created by millions of individual laser measurements. That point cloud is then manually converted into a Building Information Model, a detailed 3D representation enriched with data about materials, systems, and dimensions. This process, known as Scan-to-BIM, produces heavy models packed with geometry that are difficult to share across teams.
An approach known as “scan-to-data” takes a lighter path. Rather than modeling an entire building in geometric detail, this method, developed by UK-based startup SnapTwin, creates an accurate geometric shell for each room, with AI-extracted data embedded within it, without recreating the building’s full geometry. SnapTwin co-founder and CEO Pierre Saunal describes the goal this way: “Instead of modeling the entire building, which is continuously changing, we create a universal shell.” That shell can absorb data from multiple sources: point clouds, manufacturer specifications, asset managers’ spreadsheets, centralizing what has historically been scattered across disconnected systems.
By industry estimates Saunal cites, $2 trillion is wasted annually across the architecture, engineering, construction, and operations sector because stakeholders work with incompatible or missing data sets. The bet behind approaches like this is that a shared, lightweight digital foundation – one compatible with existing software rather than requiring new platforms – can reduce that waste.
The Compatibility Problem
Adoption failures in construction technology are frequently tied to workflow disruption rather than technical shortcomings. A tool that requires teams to abandon familiar software faces resistance regardless of its capabilities, which is why the strongest reactions tend to come not from what a new tool adds, but from what it doesn’t require firms to change.
“You are not asking me to use a new software, to learn a new thing,” Saunal says of his own product’s design logic. “We embed ourselves in your software.” Outputs built on IFC, an open international standard, can be imported into most asset management and design platforms without conversion, a compatibility-first approach that stands in contrast to tools requiring teams to learn new systems entirely. The underlying need extends beyond any single solution: as more digitalization tools enter the market, compatibility with existing workflows is emerging as a bigger adoption driver than raw technical sophistication.
Point Clouds Are Going Mainstream
The most significant shift underway isn’t about any single tool but about the raw material itself: point clouds are moving from specialist territory into common professional use. Architects, engineers, contractors, operators, and real estate managers all increasingly want point cloud data as a starting point for their work.
“We consider that in 2026, give or take, 50% of our industry is using point clouds on a daily basis,” Saunal says, citing conversations with surveyors across multiple countries rather than internal research. “The projection is that in 2030, so in four years, 87.5% of the industry will use point clouds.”
That trajectory means real estate professionals will increasingly work with digital assets alongside physical ones, not a virtual world layered on top of reality, but building-level digital records becoming standard infrastructure.
The Scaling Challenge
Digitalizing “standard” buildings – flats, houses, small multifamily or office buildings of three to four stories – is one thing. Larger and more complex structures present a distinct computational challenge: their point clouds are heavier, and the AI required to interpret spatial relationships differs fundamentally from language-based models.
“LLMs, all the AIs we use every day, are not built to see in space,” Saunal says. “When you start to look in space, you need a totally different approach in terms of understanding the space, the relationship between the elements.”
That gap has pushed some companies to build spatial AI specifically for point cloud preprocessing and segmentation, the step that prepares raw scan data for interpretation by identifying distinct structural elements within it. In SnapTwin’s case, 15 months of pilot testing validated its algorithm on standard buildings across multiple countries. When tested on complex structures, such as a U.S. power plant filled with dense piping rather than rooms, the system reached the processing limits typical of early spatial algorithms. It’s a normal milestone in algorithmic and spatial AI development: models built on standard geometry continuously improve as datasets expand, allowing them to progressively adapt to more complex environments.”
The U.S. market illustrates the scale of the challenge ahead. What American developers consider a normal-scale building, a 50 to 100-story tower in Manhattan, or massive commercial developments, would be exceptional in most European cities. These larger assets, along with expanding megaprojects in the GCC, for example, produce significantly heavier point clouds that demand proportionally more computing power to process.
None of this changes the underlying math: 2.4 billion buildings, most of them undocumented, and a 2050 deadline for retrofitting the majority of them. Whether through scan-to-data shells, spatial AI, or approaches not yet built, the industry’s dark-asset problem remains far larger than any single tool’s ability to solve it, and how quickly compatible, scalable data standards emerge may determine how much of the built environment stays legible to the people who have to manage it.
About the Expert: Pierre Saunal is co-founder and CEO of SnapTwin, a UK-based startup developing lightweight digital twin technology for existing buildings using a scan-to-data approach.
This article is based on information provided by the expert source cited above. It is intended for general informational purposes only and does not constitute legal, financial, or real estate advice. Readers should conduct their own research and consult qualified professionals before making any real estate or financial decisions.
