A city may contain thousands of buildings, but some of its most important decisions depend on a single facade panel. Its insulation, glazing, and materials influence building performance. Repeated across a neighborhood, those choices shape a larger story about energy, carbon, and the future of the built environment.
Yet the connection between those scales is often missing. Urban models rely on estimates, while detailed design information rarely reaches the broader planning picture.
In an interview about building data, automation, and construction technology, Aurimas Sabulis, founder and CEO of Dextall, described how the experience behind Kora points toward a more connected approach that’s grounded in accurate information, creative freedom, and an understanding of how digital decisions translate into physical buildings.
“A city model is only as good as the building data underneath it,” Sabulis said.
Where the City Meets the Facade
At the city scale, a building envelope often becomes a collection of assumptions: estimated insulation, an approximate window ratio, or a generalized carbon figure. At the manufacturing scale, there is less room for ambiguity. A panel has to be made, so its materials, dimensions, and performance requirements must be defined.
“Today most city models treat the facade as an assumption,” Sabulis said. “Assumed insulation, assumed window ratio, assumed carbon.”
Kora approaches the problem from the building scale, where facade designs must account for actual panels and their specifications. For Sabulis, that level of detail is precisely what makes the information valuable to cities. “If cities pulled that data up from real designs instead of estimating it, energy and carbon planning would rest on facts.” The opportunity is not simply to collect more information after a building is finished but to make reliable information available while design decisions can still change the outcome.
Pointing to New York City’s Local Law 97 as an example of building-level attention to carbon performance, Sabulis identified an earlier information gap: “What’s missing is accurate building data at the moment of design, not years after construction.”
Automate the Repetition, Not the Architecture
For developers building tools for architects, Sabulis offered a straightforward starting point to understand where their time actually goes.
Much of architectural practice involves detailing, documentation, and coordination rather than design. Water-management details for a recurring wall assembly, for example, can require similar work across multiple projects. Those standardized tasks are strong candidates for automation, but creative control is not. “Build tools that leave design freedom and design control with the architect and automate the engineering and standardized tasks around it,” Sabulis said.
The distinction matters because making a building easier to manufacture does not necessarily make it a better place to live. Software that forces architects into a narrow framework risks prioritizing production efficiency at the expense of architectural quality.
“If we restrict the artist, we end up surrounded by the most boring buildings imaginable.”
Kora’s approach, Sabulis explained, is to preserve the architect’s role while introducing buildability checks earlier in the process. “The architect designs. Kora makes sure what they draw can actually be built, and it checks that when the panel is created, not three revisions later.” A useful tool must also fit the economics of practice. Many architects work under lump-sum fees, so saving time on one task means little if the software creates unpaid responsibilities elsewhere. “Learn how your user gets paid,” Sabulis advised.
Being transparent about a tool’s limits can be just as important as demonstrating its benefits. When sharing project results, the team identified steps that still required the same amount of time. “That earned more trust than any savings number.”
Beyond the Prompt
AI makes the distinction between speed and reliability especially important. A convincing digital model is not proof that a facade can be manufactured, shipped, or installed. “In facades we’re in the millimeter business,” Sabulis said. “A small inaccuracy repeated over tens of stories adds up fast.”
His proposed workflow begins with clean data and ends with human verification. AI performs the task between those points; it does not remove responsibility for the result. “A human initiates the process using the available clean data. AI does the job. Then a human has to verify the result. Every time.” For early-career BIM and spatial data specialists, that means learning the structure beneath the interface. Sabulis recommended understanding how Revit parameters and IFC organize building information, alongside learning Python, the Revit API or Dynamo to automate work with greater control.
Just as important is knowing what information a tool used and what it did not have access to. Manufacturing constraints, tolerances, and project requirements cannot inform an answer unless they are available to the system. “Understanding what data the tool used and what’s behind the answer is paramount.”
Technical skills also need a physical foundation. Fabrication limits, structural loads, tolerances, and code requirements remain essential knowledge, regardless of how quickly a model can be generated. “AI can generate a facade in seconds,” Sabulis said. “Someone still has to know whether it can be manufactured, shipped, and installed. That person will always have a job.”
From a Neighborhood Map to an Existing Wall
A retrofit-first strategy begins with a wide view. Benchmarking data, building age, construction type, and carbon exposure can help analysts identify promising candidates across a neighborhood. Then the work changes scale. “Use city data to pick the building. Then go physical,” Sabulis said.
Scanning can capture an existing exterior and provide the foundation for a detailed BIM model. But a point cloud should not automatically replace field measurements. Older buildings move, walls bow, and windows shift. Existing conditions need to be checked before they become the basis for fabrication.
“We’ve had projects where we found differences between the point cloud and physical measurements,” Sabulis said. “Better to find that before panels are made than after.” Once those conditions are verified, the team can build a dependable model and design the new envelope around it, an approach Sabulis described in relation to Dextall’s overclad retrofit work. The larger opportunity lies in connecting urban analysis with construction knowledge.
“Few people can read a city dataset in the morning and understand a panel connection in the afternoon,” Sabulis said. “Young analysts who can do both will be rare and valuable.”
Learn the System Before Changing It
That combination of knowledge develops through experience. For Sabulis, the progression from Intus, a window company he scaled nationally, to Dextall’s prefabricated exteriors and then Kora was cumulative. Each business revealed problems and supplied the knowledge needed for the next.
“Working on exteriors every day, I kept seeing the same inefficiencies in how facades were designed and coordinated,” Sabulis said. “That’s how Kora was born.”
It is also how he interprets the idea that “discipline equals freedom”: understanding the fundamentals creates room to innovate responsibly. “When you have both, you can start to play and innovate, because you know the boundaries and you know the risks.”
That perspective comes with a warning for newcomers to construction technology. Complexity is not always evidence of a broken system. Some practices exist to manage financial, contractual or safety risks that are easy to underestimate from the outside. “People coming from outside the industry often think everything in our workflows is broken,” Sabulis said. “Many don’t go deep enough to understand the nuances. Those nuances exist for a reason.”
A new venture needs to understand retainage, pay applications, and long cash-flow cycles, alongside liability and responsibility for stamped drawings. It must also recognize that every project brings a new team and contract.
Review processes such as submittals, mockups, and inspections may appear slow, but they help catch decisions that could become expensive or dangerous. Meaningful innovation distinguishes unnecessary repetition from necessary care. “Learn the system first. Then change it.”
Connecting Digital Ambition With Physical Reality
Across the interview, a consistent principle emerged: better tools depend on stronger connections between information and expertise. City data can identify where action is needed. Building models can reveal what that action involves. Automation can remove repetitive work without taking creative control away from architects. AI can accelerate tasks, provided people understand the inputs and verify the results.
A city-wide strategy eventually reaches an existing wall, a manufacturing tolerance, or a panel connection. Connecting those details to the larger plan is what turns digital ambition into better buildings and better cities.
