But first.. Monday Live is Six Years Old!
MondayLive! launched on April 27, 2020, in the earliest days of the pandemic, when offices sat empty, and the industry suddenly had time to think.
What started as a weekly conversation among a small group of practitioners has grown into a persistent institution, six years of live, unscripted dialogue on the hardest problems in building automation. Every session is preserved in the AutomatedBuildings.com archive, free and open to anyone.
April’s discussions on AI Across the Stack, moving through the smarter stack layer by layer and asking the same question at every level: where does AI actually fit, and what has to be true before it can work? Previous sessions covered the bottom of the stack, the physical sensors and controllers that form a building’s nervous system, and the top, where executives ask portfolio-level questions about performance and mission. This month landed squarely in the middle.
The middle of the smarter stack sits between raw physical data and executive decision-making. It covers the integration layer, where disparate systems are connected and normalized across protocols and vendors; the context layer, where semantic models, knowledge graphs, and business rules give that data meaning; and the analytics and applications layer, where AI models, fault detection, and dashboards interpret data to surface actionable insight. This is where data becomes understanding — and where the session asked whether AI is filling these layers in or beginning to collapse and redefine them entirely.
The real problem is meaning, not data
A knowledge graph built for a real, living building mapped 3,000 assets into 122,000 specific relationships, each expressed as an RDF triple. That is not a dashboard. It is a forensic map of how the building is actually wired and why things connect the way they do. Without that structural context, even the most capable AI is reduced to probabilistic guesswork, even though deterministic accuracy is what operations actually require.
“Connecting data is not the same as understanding it. APIs don’t create meaning. Dashboards don’t explain relationships.”
There is also a business model problem layered on top of the technical one. Building data has real value, but the path from data existing to someone actually unlocking that value runs directly through the middle stack. Without the integration and context layers in order, that value stays locked. Edge devices and AI-ready infrastructure at the building level are part of what makes the extraction possible, but they depend on the semantic groundwork being laid first.
Two failures that were not technology problems
A building with solar panels pushed excess energy to the Portland grid at the wrong moment and incurred a utility penalty. The sensor data was flowing. What was missing was the business rule in the context layer connecting PV output to local utility tariff conditions. In another case, east-facing and west-facing PV arrays were specified to report separately so morning and afternoon performance could be tracked. For months, the data showed two identical parallel lines. An investigation eventually found that the installer had split the circuits in the wrong orientation. The design intent existed. The data existed. The semantic connection between them had been lost somewhere in the handoff from design to installation, and nothing in the analytics layer could surface what the context layer had never captured.
Neither was a hardware failure. Neither was an AI limitation. Both were middle-stack failures, and both were costing money every day until someone went looking.
AI is accelerating the reckoning, not solving it
When someone prompts a building system for a live answer, and nothing comes back, the failure is no longer invisible. Buildings have been opaque to their owners for decades. What AI changes is not the opacity itself but the speed at which it becomes undeniable. The decisions being made now about data infrastructure, open standards, and semantic structure are setting the conditions for tools that do not yet exist. A building whose middle stack is incomplete today will not be ready for what arrives next.
All sessions from AI Across the Stack are archived at AutomatedBuildings.com and on the MondayLive YouTube channel.
Is AI About to Eat Your Stack?
Stop Guessing, Start Reading Blueprints
When the Building Knows More Than You
The Road to AHR Chicago
AHR Chicago is coming. The question asked of vendors and integrators is simple: What are you bringing? If the answer does not involve AI at the edge or an open gateway that costs less than a restaurant dinner, you are already behind.
