Monday Live | May 2026
Monday Live is a weekly open session where practitioners, technologists, architects, and systems thinkers work through the future of smarter buildings. The views are personal, not institutional, which is what makes the discussions worth paying attention to.
This week opened a new monthly theme: getting back to basics. It turned out to be a more layered question than it sounds.
“Basic” Is Not the Same for Everyone
Ask a room of experienced practitioners what “basic” means, and the answers diverge quickly and productively.
For one voice, the basics have always been the same: connecting two systems that were never designed to talk to each other, with enough rigor to serve a specific use case. Simple in concept, difficult in practice, and still the foundational challenge after decades of evolution.
For another, the basics are the ones and zeros. The discussion traced a line from X10 power-line modules purchased with paper-route earnings in the 1980s to the question of where those bits live today. The Coalition for Smarter Buildings (C4SB) is doing deep working-group-level thinking on exactly this, connecting the fundamentals to the frontier.
For the architect, the basis is context. A use case without a location, a purpose, and a relationship to the people inside the building is not a use case it is a demo. Technology applied to buildings without that context drifts toward abstraction and away from value.
From a business perspective, the basic value proposition is the key: whether the industry’s standard arguments still hold as people’s use of buildings continues to shift. Energy savings and predictive maintenance are well-understood. What buildings need to become hybrid work and changing real estate patterns reshape demand is less clear.
Cutting across all of these was the fourth and most-discussed type of basic: AI itself.
The 10% Trap and Why Most AI Projects Stall
A Gartner figure cited in the session: 72% of AI projects are failing to move the P&L. The room was not surprised, but had precise language for why.
It takes 10 minutes to build a convincing AI demo. It takes 10 months to productize it, add appropriate guardrails, understand the true cost of inference at scale, and confirm it is solving something the business will pay for. The gap between those two timelines is where projects go quiet.
The failure mode is specific: activity gets mistaken for progress. An AI model produces outputs that look meaningful. But without a bounded, measurable use case connected to a real outcome, the project stays a science experiment.
The correction is not to abandon AI. It is to stop treating it as a general-purpose answer and to deploy it where the problem is well-defined and the data is well-structured.
Why AI Cannot See Most of Your Building
AI cannot see X10. It cannot see most existing building systems. The protocols, siloed controllers, and proprietary data structures built up over 30 to 40 years of automation deployment are not natively visible to the AI tools entering the industry today.
That is not a criticism of AI. It is a gap with a known solution: the knowledge graph.
A knowledge graph is a category of database. Data pulled from existing systems -BACnet, Modbus, and older serial protocols is stored in a format that carries meaning alongside values. A shared vocabulary makes that data interpretable not just to one application but to any application that understands the vocabulary. With data stored this way, AI can query with precision rather than inference.
The practical question the session raised was direct: a building owner with 200,000 square feet of 15-year-old native BACnet wants a knowledge graph. Who do they call? The answer demonstrated at the PAE Living Building is that an Intelligent Building Backbone (IBB), a device that connects to existing network infrastructure and ingests data into a knowledge graph database, is the path. The controls contractor or systems integrator already on-site becomes the person who deploys it.
The minimum viable starting point is lower than most assume. A spreadsheet matching room IDs to BACnet devices is, technically, the beginning of a knowledge graph. What matters is not completeness at the outset but maintaining consistent unique identifiers across systems so data from different applications can be joined and reasoned about over time.
A 6-day collaborative exercise illustrated why this matters. Participants received an RDF file with global unique IDs for every element in a building: spaces, sensors, systems. The task was to add data and return the file with IDs intact. Most applications strip identifiers they do not recognize. Maintaining the ID is the handshake that enables cross-system collaboration. One team went further and built a complete emergency response application on top of it.
From DDC to AI-Native
The 1980s were about DDC. The 1990s were about networking. The 2000s were about moving from proprietary to open protocols and getting data to the internet. Each transition felt disruptive at the time and became obvious in retrospect.
The current transition is toward AI readiness. The fundamentals of control have not changed; buildings still need programmable logic, configurable sequences, and reliable actuation, but those systems now need to expose data in formats AI can interpret. “AI-native” is the term gaining traction.
ASHRAE is actively working to define what control algorithms should look like for consistency across the market. That interoperability work, at its foundation, makes it possible for AI to read what a building is doing and respond usefully. This is an extension of what came before, not a break from it.
The Technician on the Ground Is Already Moving
The expectation within a large field organization was that AI adoption would require top-down tooling and a structured rollout. What was actually happening was different.
Technicians, particularly younger ones, were already using AI tools on their own. They were taking CSV exports and points lists, dropping them into AI tools, asking questions, getting answers, and teaching each other informally at a pace no formal program could match. They were not doing anything dramatic just doing familiar work faster, with less friction.
A parallel example: a daughter, asked to collect floor plans for a grocery store, returned with a fully digitized 3D scan produced on an iPhone using Matterport, exported as a DXF, with no prior training. Within two days, that scan had been combined with room IDs and sensor data to produce a working first version of a digital twin.
The consistent pattern: when the task is focused, the tools are accessible, and the outcome is concrete, people figure it out. Complexity arrives when the scope expands before the foundation is solid.
What Buildings Need to Be
The session closed on a question that gets lost when the conversation stays technical: why do buildings exist, and what is that purpose becoming?
Buildings designed for dense five-day occupancy are operating at a fraction of that. The case for earning the commute by creating enough value that people choose to show up is real, but has limits. That means owners face two simultaneous challenges: making the building worth coming to, and rethinking what to do with it if that case cannot be made.
Both challenges require information. A building with no machine-readable signal about its condition, occupancy, systems, age, or structure cannot be intelligently managed, repurposed, or safely assessed in an emergency. The illustration offered was unambiguous: when firefighters approach a burning structure, the critical question is not energy performance. It is whether the floor will hold.
At its most basic, a smarter building is one that can speak. Simple signals are used to indicate whether it is occupied, what its structure is, and what it is used for, creating the foundation for more sophisticated application that follows.
The industry has spent years talking about what AI will do for buildings. The more useful question is what buildings need to become before AI can do anything with them at all.
Further Reading from AutomatedBuildings.com
- Knowledge Graphs in the Modern Building: How BACnet scans become RDF-based knowledge graphs using current tools, and where AI fits in the tagging process.
- Why AI Needs to Stop Guessing and Start Reading the Knowledge Graphs — What 122,000 relationship triples in the PAE Living Building reveal about the gap between design intent and operational reality.
- When the Building Knows More Than the People Running It: How AI fits into the purpose and operations layers of the building intelligence stack, and why connected data is not the same as understood data.
- A Coalition in Motion: Building the Bridges at Machine Speed How BACnet and cloud-native approaches converge, with persistent IDs as the connective tissue between geometry, graphs, and live data.
Monday Live meets every Monday afternoon. Sessions are open and interactive. Details and past recordings at mondaylive.org.
