Relentless. It waits. It watches. It learns. Then it wraps—and acts. Ai that optimizes your building from the inside out.
In 2002, the world watched Minority Report envision a future of predictive policing. But in real life, Keith Gipson was building it a couple of years later.
As one of the originators of the building automation space and inventor of EEM as Founder at Silicon Energy in 1997 and just before Co-Founding Phoenix Energy Technologies, in 2004 at his third startup company, ShieldOPS, Keith’s AI “Predictive Policing” tech was literally featured on CSI: Las Vegas —the popular crime procedural TV series that dramatized predictive forensics. It was so powerful that producers reportedly asked him to tone it down. It worked too well.
He didn’t stop there. As CTO of Phoenix, he carried those same predictive AI principles into the built environment—replacing suspects with setpoints and turning energy waste into actionable intelligence.
AI isn’t catching criminals.
It’s catching wasted kWs, broken chillers, and drifting setpoints
Today, that same level of relentless, prescriptive AI isn’t catching criminals. It’s catching wasted kWs, broken chillers, and drifting setpoints in a one-year pilot at a California State University project. Quietly. Efficiently. Without fanfare.
And the foundation for you catching these savings is already in your buildings. This article points to how you can do just that.
facil.ai Interview at The Huntington
Chilling in the Garden
No whiteboards. No dashboards. Just three people in a Zen garden, tracing patterns in gravel and talking about AI, chillers, and spiders. We were chilling at The Huntington in Pasadena, California, watching shadows shift across carefully raked stone. Keith Gipson, Danielle Radden—both with facil.ai—and I were surrounded by centuries-old precision and simplicity, talking about inputs and outputs, the two gates of the walled garden, and how systems—natural or built—need flow to stay alive.
All the while, facil.ai’s AI agents were hard at work—deep inside dusty mechanical rooms, relentlessly optimizing performance and quietly saving thousands of dollars in energy. While we discussed flow, the machines were creating it.
That was the moment the metaphor clicked: this wasn’t just about AI. It was about relentless AI—a system that learns, layer by layer, until it knows exactly how and when to act. Later, Keith would describe it perfectly: like a spider wrapping its prey, slowly and methodically, until the system is fully understood and controlled.
The physical pattern is beautiful, but it’s the behavioral pattern—the data, the rhythm, the feedback—that facil.ai captures and acts on. It’s the same duality we see in digital twins: the physical and the informational coexisting.
A Crime Scene for Energy
We discussed how using AI to enhance building performance is no longer science fiction. Danielle brings a particularly fascinating background to this space—combining investigative journalism, communications strategy, background in neuropsychology, and tech innovation. It gives her a unique lens on how to frame complex technical systems in ways owners can actually understand and act on. It’s science in operation. facil.ai’s platform doesn’t just model buildings—it controls them. It listens to the data, learns behavior patterns, and acts. Not just predict, not just alerts—acts.
Keith describes it as “relentless layering” – like a spider slowly wrapping its dinner- training itself based on the data stream, gaining confidence, and then taking action. In their Cal State pilot, facil.ai controlled a chiller system with 40%+ efficiency improvement. At a retail portfolio, it delivered $200,000 in verified energy savings per month—measured and validated.
It does this by deploying autonomous AI agents that connect to existing smart HVAC assets to monitor, learn, and prescribe control actions in real time. Think of it as a digital twin of real-time control—an invisible overlay of logic and learning that reacts instantly to data flowing from physical systems. It doesn’t need a high-gloss model to operate. It needs live connections to your building’s heartbeat.
This Isn’t a Dashboard. It’s a Direct Line to the Machine.
facil.ai doesn’t sell dashboards. It doesn’t ask you to babysit alarms. It doesn’t even need a screen.
As Keith told me: this kind of AI talks like a machine—straight to the data stream, in real time.
No human interface required. Just constant feedback, learning, and action.
And if you’re wondering where the dashboard is? It’s in the owner’s bank account—measurable, relentless savings, showing up month after month.
That’s prescriptive AI: not telling you what might happen, but executing what should happen—based on actual system behavior.
Owners often think they need to build a perfect
3D model before they can have a digital twin.
Owners often think they need to build a perfect 3D model before they can have a digital twin. Keith flips the script: “You don’t need to model anything. You just need the data.” As I said during the ALN@3 session, it’s like trying to model a tornado down to the blades of grass—you’ll never catch up. Instead, facil.ai captures the pattern and behavior and lets the AI respond in real time.
This is a recurring problem in the BIM and Digital Twin world—what we often call “Hollywood BIM”: overly detailed models that look impressive but fail to support the actual use case. I’ve been calling this out for years. There’s a time and place for detail, but it must be driven by purpose. facil.ai is the most extreme—and brilliant—example I’ve seen of skipping the overmodeling and delivering immediate, actionable value.
This is the reality-check moment for our industry: You are wasting time modeling the crime scene while the suspect is walking out the back door.
And in a fitting twist, we’ve come full circle. Just as we opened with a nod to Minority Report, we close with one too—those spiderbots scanning every room, relentless in their search. Keith’s metaphorical spider doesn’t crawl walls, but it wraps buildings in insight the same way—strand by strand, with quiet persistence.
Only this time, the mission isn’t surveillance. It’s transformation.
Join the discussion on this post on LinkedIn here.
Governance Meets Intelligence
For the Asset Leadership Network and facility owners who are serious about long-term stewardship, this is a breakthrough. AI becomes not just a cost-saver, but a governance enabler. It enforces the rules you set, measures its performance, and feeds results back into your strategy.
As Keith said: “Energy efficiency is no longer optional. You either do this or you go out of business.”
Owners, It’s Time
If you’re an owner still waiting for the perfect BIM,
perfect integration, or perfect vendor suite
before getting started—stop waiting.
The AI is already in the building.
It’s just been quiet. Watching.
Learning. Ready to act.
And it doesn’t need your permission
to be relentless
Move to Open
As part of this movement toward more open, interoperable building intelligence, I also encourage readers to follow and engage with the work of the Coalition for Smarter Buildings (C4SB). Their mission to promote open standards and open source aligns with what FACIL is demonstrating: that true digital transformation in the built environment doesn’t come from locking down data—it comes from opening it up.
Referenced in this article:
- Interview with Keith Gipson and Danielle Radden of facil.ai (April 2025)
- Asset Leadership Network ALN@3 live Session with Onuma + FACIL.ai (April 3, 2025)
