I didn’t start 2025 expecting the architecture, engineering, construction and operation (AECO) industry to suddenly change direction.
It didn’t.
What became clear instead was a widening gap: between those continuing to rely on familiar patterns and those beginning to ask harder, more consequential questions, even when the answers challenged long-standing assumptions.
The year opened at the Design Intelligence Conference with an explicit focus on “consequential questions.” Not tools. Not roadmaps. Questions that force accountability.
Why does so much information disappear between design and operations?
Why do we treat handover as an endpoint instead of a beginning?
Why are models optimized for delivery rather than use?
That framing mattered. When questions stop, inertia takes over. The status quo persists not because it works, but because it’s easier to defend than to change.
Across conferences, workshops, podcasts, and closed-door conversations, the same dynamics surfaced again and again. Automation, systems, and intelligence are no longer evolving in parallel. They are colliding, and the friction is now visible to everyone involved.
In conversation after conversation, especially with owners, a similar pattern emerged. Many had begun asking AI tools to analyze their facilities and assets, hoping for insight or efficiency. What they got back instead was a blunt response:
Your data is fragmented.
Your systems don’t talk.
Your digital assets don’t align.
AI wasn’t hallucinating. It was exposing something uncomfortable: humans had been hallucinating for years, filling gaps between silos with assumptions, spreadsheets, and manual workarounds.
That realization shifted the conversation. What had lived comfortably in IT discussions suddenly reached the C-suite. When AI can’t answer basic questions about assets, energy, space, or risk, it stops being a technical inconvenience and becomes a leadership issue.
Still, much of the industry resisted that conclusion. Fear played a central role, particularly fear of sharing data, exposing inefficiencies, or losing perceived IP control.
In practice, fear centralizes dependency.
At the same time, there was no shortage of hype. Digital twins, Hollywood BIMs, and 3D visuals were everywhere, impressive to look at, but often hollow beneath the surface: models disconnected from operations, twins reduced to dashboards, intelligence trapped inside vendor ecosystems.
The language sounded advanced.
The outcomes often weren’t.
One small but telling moment came during a Monday Live session I was invited to participate in as a member, where LEGO bricks became a simple metaphor.
When systems are built as monoliths, you can’t see how they come apart, and you can’t recombine them without breaking something. When they’re modular, you can take them apart, understand the pieces, and put them back together in new ways. That’s exactly the shift owners need to demand — decomposing systems, not to destroy them, but to make them adaptable, interoperable, and resilient over time.
Greenbuild brought one of the most striking contradictions into focus.
Sustainability and energy performance were front and center, as expected. But in the same rooms, AI was being enthusiastically discussed, often without acknowledging its massive energy demands. At the same time, architects publicly declared they do not share data because of IP concerns, even while positioning themselves as sustainability leaders.
In some cases, AI was being used to reverse-engineer information from static documents and PDFs, extracting data that had already been created, flattened, and locked away. Enormous computational effort spent reconstructing what should never have been lost in the first place.
The irony was hard to miss: using energy-intensive AI to recover data from energy-inefficient processes, while claiming progress toward sustainability.
Standards were another place where the pressure became visible. I’ve been involved with buildingSMART for years, and in 2025 the conversations shifted toward confronting friction.
After the conference, I agreed to rejoin buildingSMART USA as co-chair of the Buildings Committee with a practical goal: helping connect standards to operational reality and accelerate the convergence already underway across AECO.
That means reducing abstraction, aligning IFC and COBie with real-world workflows, and working alongside efforts like C4SB, ASHRAE 223P, NIST, and owners to close the gap between standards development and how buildings actually operate.
A similar signal appeared in AI research communities. At large conferences like NeurIPS and in Urban AI workshops, buildings and cities were largely absent, not because they lack relevance, but because usable, real-world data is difficult to access.
The workaround became normalized: synthetic data.
Simulated buildings. Generated sensor streams. Fabricated urban patterns. Not as a preference, but as a necessity. When researchers are forced to invent reality because real systems are silent, the problem isn’t AI. It’s governance, access, and ownership of data.
All of this unfolded against a backdrop of broader anxiety. Economic uncertainty. Geopolitical instability. And growing unease about AI itself. Some tried to dismiss it. Others tried to defend against it, hoarding data, reinforcing silos, or longing for a slower past that felt safer.
Fear, however, doesn’t stop change.
It only delays learning.
And while much of the industry hesitated, a smaller group began to move the needle.
In parallel with the conferences, work through efforts like the Coalition for Smarter Buildings and the Linux Foundation ecosystem shifted from talking about what might be possible to building what works, in the open.
Using real buildings.
Using real data.
Accepting imperfection and iterating anyway.
The PAE Living Building became a proving ground for a different approach. Systems were connected without ripping them out. Owners stayed in control of their data. Semantics and ontology were treated as infrastructure, not academic exercises. Progress came not from eliminating complexity, but from navigating it methodically.
That acceleration didn’t come from new technology. It came from willingness, to connect, to share, and to learn in public.
Open source changed the pace. Instead of waiting for perfect alignment or multi-year roadmaps, teams documented what worked, exposed what didn’t, and adjusted.
AutomatedBuildings played a role here as well. Having a place to publish in real time mattered. Ideas didn’t need to wait for approval or polish. Failures, contradictions, and opportunities could be surfaced while they were still useful. In a year where formal processes often lagged reality, that openness helped ideas move faster.
Across conferences, the contrast kept repeating.
At AHR and ASHRAE, physical systems collided with digital consequences. At NFMT, operational fatigue replaced optimism. SCaLE showed what happens when permission is removed. Realcomm revealed executive impatience. Greenbuild exposed unresolved contradictions between sustainability, data, and AI. Tradeline connected capital planning with live data. Asset Leadership sessions traced execution failures back to governance, not tools.
As we move into 2026, the opportunity belongs to the doers, especially those willing to share, connect, and move beyond fear.
For owners, the message is direct.
You are paying for the buildings, the systems, the data, and the long-term consequences. Do not accept monolithic, expensive solutions that trap your data and limit your future. Do not accept the idea that interoperability is risky while dependency is safe.
Ask more questions.
Demand access to your data.
Demand systems that can connect, adapt, and evolve.
The cost of inaction is no longer abstract. It shows up as duplicated effort, inaccessible information, stalled decisions, unmanaged risk, and AI trained on synthetic stand-ins instead of reality.
2026 will not be about whether change is coming.
It already has.
The real question is whether you participate, or continue to watch from the sidelines.
I’ll be watching closely how these same rooms respond as 2026 unfolds.
