Mark Elster
Mark Elster

Architecture is entering a strange new phase. Software can generate increasingly sophisticated images, automate documentation, and turn spoken prompts into design possibilities within moments. Construction, meanwhile, is dealing with a shortage of people capable of carrying those ideas into physical reality. A September 2026 survey from the Associated General Contractors of America and NCCER found that 88% of firms with craft openings said those positions were as hard or harder to fill than a year earlier.

This pressure is making automation less of a futuristic proposition and more of an operational necessity. Every hour spent manually counting components, revising drawings, or transferring information between disconnected systems represents labor that an industry with a constrained workforce can ill afford to waste. The question facing architecture is increasingly how intelligently that labor can be redirected.

Residential architecture adds another complication. A house can be rendered, modeled, and optimized with extraordinary precision while still missing the human circumstances that make one home feel entirely different from another. A floor plan cannot reveal who dominates a conversation between spouses, which preferences remain unspoken, or how a family hopes its home will shape everyday relationships. Those judgments emerge through observation and interaction.

Digital modeling is already changing how architects communicate those decisions. Graphisoft's Archicad, for example, can generate plans, sections, elevations, details and schedules from a shared 3D model, allowing changes to the embodied virtual building to flow into its documentation. Graphisoft's current BIMx visualiser platform also allows project models to be explored and interrogated with associated documentation and project-level data.

Mark Elster, founder of Seattle-based AOME Architects, has spent four decades working through successive shifts in residential architecture, from hand drafting to increasingly sophisticated digital environments. His argument about AI is less about resisting technological change than directing it toward the parts of architecture where it can create genuine leverage. "Freeing up more time for listening to our clients to gain an understanding of what's important to them," Elster explains. "Tidbits about their background, personality nuances, which give us some cues about how we'd approach the delivery of service to them."

Those cues become particularly important in custom residential work, where clients rarely arrive with perfectly aligned priorities. Elster points to the dynamics he observes between couples, where one person may focus heavily on cost while another concentrates on how the home should function or feel. Such tensions can become useful design material.

"Oftentimes the best design comes out of what on the surface appears to be conflicting requirements between two people," he says. "The most interesting designs come in the seam between those two conflicts."

AOME's approach starts from that human complexity. The firm's website notes that every luxury home is tailored to its client's needs, with its design process examining lifestyle, requirements, and the realities of the site before the design develops. Its projects typically range from $3.5 million to $25 million or more, making the consequences of those early decisions substantial.

Technology enters that process through AOME's virtual twin, a single-source-of-truth building model that faithfully represents what will actually be built. The model contains architectural, structural, and mechanical information alongside finishes, furnishings, and other project data. More importantly, Elster treats it as a shared source of information rather than an elaborate presentation tool.

The approach requires more work upfront, but the model can generate much of the project's two-dimensional documentation from the same underlying information. A later change to a room can therefore propagate through connected drawings instead of requiring each document to be revised separately. AOME also requires consultants to fully integrate their information into the same model, with Betwixt Integrated Modeling supporting consultants who work in other systems.

For contractors, the value becomes especially tangible. Traditional drawing sets still require contractors to measure and count information manually. All the quantities embedded in AOME's virtual twin model are extracted and provided for the contractors to use for estimating, reducing a category of labor that Elster considers low-level work. "We give them the quantities directly out of the model, and so they're able to generate more accurate, speedier cost estimates than they can with any other architect they work with," he says.

Elster sees AI extending this logic. AOME has developed a visual interface that lets contractors inspect model elements and verify quantities without learning Archicad. The technology makes a previously unattainable tool feasible because AI can absorb much of the development labor.

The larger shift, then, is less about whether machines can participate in architecture and more about where their participation creates value. "We need to lean into AI as a tool to relieve us of low-level tasks and some high-level tasks we find tedious," Elster says, "so we can devote ourselves to the high-level, human tasks that AI can't do as well or at all in many cases." For residential architecture, that means technology can carry more of the information burden while architects remain responsible for understanding the life that information is ultimately meant to serve.