Simulations predict buildings. Buildings perform differently. The gap is where intelligence dies.
For decades, design intelligence has been stranded after occupancy. The model becomes a report. The report becomes a PDF. The PDF becomes an archive. And the operating building starts telling a different story.
After attending SimBuild 2026 in Minneapolis, my strongest observation was not that simulation is becoming more sophisticated. That was obvious.
The more important shift is this:
Simulation may be moving from a design-phase prediction tool to an operational feedback system.
To clarify what this means: In design, simulation is predictive; architects and engineers use BIM and energy models to forecast how a building should perform before construction. In operation, simulation becomes calibrative: using actual building data, sensors, controls, and measured performance to understand what is really happening. The opportunity is bringing that actual building data back into the simulation loop. When design-phase predictions meet operational reality, simulation becomes a continuous conversation between intended and actual performance.
What if the simulation never stopped?
What if it stayed alive with the building? What happens when it remains connected to the real building, its controls, sensors, semantic model, AI layer, and the owner’s operational decisions?
Not as a static report. Not as a compliance artifact. But as an ongoing operational conversation between predicted behavior, measured performance, semantic building knowledge, control systems, AI, and the living building itself. Also known as a digital twin.
That exact statement was rarely made directly during the conference. But the signals were everywhere.
SimBuild Expands Its Orbit
SimBuild reflected a larger shift: simulation is no longer only a design-phase exercise. As buildings become more connected, automated, and AI-enabled, simulation becomes more valuable because it provides something AI desperately needs: grounded physics, measurable constraints, and a way to compare what should happen with what is actually happening.
That may explain why the conference felt broader this year. The U.S. Department of Energy was a major sponsor. National labs, research universities, major engineering and architecture firms, controls vendors, simulation software developers, AI startups, and operational technology groups were all represented. Even NASA appeared in the mix, appropriately enough for a conference increasingly dealing with systems, climate, resilience, and larger operational scales.
There were a lot of very smart people in the room. But what made the conference especially interesting was not just the expertise itself. It was the sense that many groups were approaching the same larger problem from different directions, using different languages, tools, and boundaries.
This was no longer just an energy modeling audience. It was a broader ecosystem beginning to converge around simulation, AI, operations, digital twins, and lifecycle intelligence.
The New Silos of the AI Era
One observation: many teams are trying to solve the right problem, but risk recreating a familiar one.
For decades, the building industry has struggled with fragmented BIM systems, disconnected platforms, and siloed intelligence. In the AI era, fragmentation can reappear in more sophisticated forms. Phrases like “bring everything into this platform” or “use this data lake” sound appealing.
None of those approaches is inherently wrong. Centralized repositories can reduce chaos. The issue is not whether these systems exist, but whether they become the boundary of intelligence.
The deeper question: does building intelligence remain portable, continuously connected, and operationally accessible outside any one platform, application, or specialist workflow? That matters because buildings outlive software ecosystems by decades.
The Missing PhD in the Loop
Mary Ann Piette of LBNL joked in her keynote that today’s workflows can still require a “PhD in the loop” just to access or interpret building intelligence.
That landed because the room was full of experts who understand the complexity, and also understand that expertise alone is not a scalable operating model.
As buildings become more connected through simulation, AI, digital twins, controls, and operational analytics, sophistication is not enough. Some BIMs and digital twins become so detailed and impressive that they are difficult to maintain after handover.
That is the risk of “Hollywood BIMs” or “Hollywood Digital Twins”: visually compelling, technically impressive, but fragile in daily operations.
Buildings last for decades. Their intelligence must remain understandable, portable, and accessible across changing teams, vendors, and systems.
The Pattern Emerging Across SimBuild
Many presentations at SimBuild described pieces of building performance practice, calibration workflows, AI optimization, fault detection, semantic interoperability, and digital twins. But these conversations often remained separated by familiar domain boundaries. Simulation in one context, controls in another, AI in another.
Less common was a clear description of how these pieces could become a connected ecosystem. The larger opportunity is not smarter dashboards alone, but continuously connected operational intelligence, where simulations remain alive after design, operational systems remain semantically connected, and the real building continuously informs the simulation just as the simulation informs the building.
The Living Building Enters the Conversation
Our session, “You Don’t Design Net Zero: You Operate It,” illustrated the problem. Presented with PAE, NIST, ONUMA, and the Coalition for Smarter Buildings, it addressed a core challenge: even one of the most advanced Living Buildings in the country struggled with fragmented operational information after occupancy.
The building’s sophisticated design intelligence, microgrids, sensors, controls, and radiant systems became distributed across disconnected operational dashboards. The challenge was not a lack of data, but preserving meaning across the lifecycle.
The session showed how semantic relationships, shared identifiers, and interoperable APIs can reconnect simulation logic, operational systems, controls, and the real building, not as a one-time export, but as a living operational environment. This creates feedback loops between predicted behavior, measured performance, and future optimization.
When Simulations and Buildings Learn From Each Other
This may become one of the next major evolutions of the simulation industry, not simulation as isolated prediction, but as an ongoing operational participant.
The building continuously teaches the simulation, and the simulation continuously informs the building. Measured performance reveals drift, unexpected behavior, or control issues. The simulation helps test responses and anticipate consequences.
That feedback loop requires more than raw data. It requires a semantic layer: machine-readable relationships, stable identifiers, and continuously connected meaning. Without these, simulation, AI, and operational systems remain disconnected. With them, buildings become operationally intelligible and simulations become continuously calibratable.
A useful digital twin is not one application, one dashboard, or one data lake. It is a system of systems that connects simulation models, semantic relationships, operational data, controls, sensors, and AI reasoning across time. The danger is when one system becomes the boundary of the twin rather than a participant in it. The opportunity is for multiple systems to remain connected through shared identifiers, semantic relationships, and open standards, so that the building, simulation, and digital twin keep learning from each other.
From Reactive Buildings to Predictive Buildings
This is especially important for owners.
Today, many building problems are discovered too late: equipment fails, energy drifts, a storm or outage reveals weaknesses already present but not visible. By the time these appear, owners are reacting to consequences.
A continuously connected digital twin changes that pattern. If the simulation, sensors, controls, and the real building remain connected, the system can detect drift and expose risks earlier. The question shifts from “What went wrong?” to “What’s changing, what might happen next, and what can we do before it becomes expensive?”
For owners, the measurable benefits include fewer surprises, better decisions, earlier warnings, reduced operational risk, and greater confidence that the building continues to perform as intended. As buildings become more complex and dependent on AI, connected simulation becomes critical; if those systems only react to alarms, they are already late.
Simulation may become part of how owners continuously understand, operate, and reduce risk in buildings after occupancy, one of the most practical lessons from SimBuild.
Building Intelligence Stays Alive
SimBuild 2026 felt like a conference beginning to ask larger lifecycle questions.
How does simulation connect to operations? How does AI connect to trustworthy building context? How does operational intelligence survive after handover, when teams change, software changes, and the building continues to operate?
The technical pieces are rapidly emerging: AI agents, semantic graphs, APIs, real-time operational systems, and interoperable digital twins.
The larger, unresolved challenge is whether the industry will connect these pieces openly or create larger, more sophisticated intelligence silos around them. This may become one of the defining questions of the next decade.
The future of building intelligence may depend less on which platform wins and more on whether the intelligence itself remains portable, interoperable, continuously connected, and operationally accessible over time.
More to Come
We are seeing similar patterns emerge across conferences, owner discussions, standards groups, and real projects worldwide. This is not a one-conference observation. It is part of a larger industry trajectory, and we will continue reporting on it as we connect the dots between simulation, operations, AI, digital twins, and the real buildings where all of this has to work.
