Monday Live! June 8, 2026
Readiness for AI-Native Smarter Buildings continuing the June theme of Readiness
The Monday Live crew gathered June 8th to road-test a term they’ve been circulating: “AI native.” The question on the table was whether it could do real work as both a technical standard and an industry rallying point, much like BACnet did a generation ago.
Three Tiers, One Trajectory
The group mapped a progression the industry is living through right now. AI-enabled systems simply bolt a chatbot onto an existing dashboard. AI-ready systems place a language model on top of existing infrastructure without rebuilding it. AI-native systems are architected from the ground up for agents, not browsers, with semantic structure and live data baked in from the start.
The Remove-AI Test
The cleanest definition came from feeding a Vinton Cerf article into an AI and asking it to draw the line. The answer: remove the AI entirely. If the system still functions, it is AI-assisted at best. If the operating model collapses and agents are central to decision-making and execution, it qualifies as AI native.
Why the Data Has to Change
AI does not use browsers. Agents do not navigate dashboards. The entire technology stack built for humans navigating web interfaces is largely useless to an agent reasoning about a building. AI-native demands machine-readable, semantically structured, live data. Knowledge graphs, digital twins, and agent-accessible ontologies built on standards like ASHRAE 223P, Brick, and RDF are not aspirational features. They are the minimum required infrastructure.
The Readiness Gap
A recent poll of 70 to 100 controls professionals found only a handful had any hands-on experience with AI. The term’s near-term value, like “BACnet” before it, is as a shorthand that consultants and specifiers can put in documents before they fully understand what it means. The validation comes through use, not theory.
Testing and Certification
The group debated whether AI-native claims could be independently tested. A single stamp seems unlikely. A scoring system across multiple sub-standards, 223P compliance, RDF accessibility, semantic interoperability, is more realistic and already partially possible given existing validation tools in those standards.
Humans Stay in the Loop
The ability to override or disable AI systems is not a limitation of AI-native architecture. It is a design requirement. Insight, optimization, and automation can all be AI-driven. The kill switch is non-negotiable.
Trust is the Foundation
The thread running through everything is trust: in the data, in vendors making AI-native claims, and in the technology itself as it takes on more operational responsibility in buildings.
