In the fourth installment of Monday Live!’s March series on governance, the conversation took a sharp turn. After weeks of exploring data ownership, semantics, and open platforms, the panel turned to a more foundational question: what happens when the systems we built to govern our buildings become obsolete?
The answer came in two words: biodegradable BMS. The idea is simple. The building management system, as we know it, is a layer of software and hardware placed on top of fundamental controls to provide oversight, alarming, and scheduling, which may slowly dissolve. What remains will be the core control systems themselves, now governed not by a legacy BMS but by distributed AI agents operating at the edge.
This session was not about whether AI will replace existing systems. It was about how and what that transition means for an industry still grappling with fragmented data, degraded guardrails, and the painful realities of legacy infrastructure.
Key Insights from the Conversation
Controls Came First, Governance Came Second
Decades ago, a VAV box had one job: maintain 72 degrees. Pneumatic tubes, electric signals, and simple sequences handled the work. Then came networking, and suddenly you could see the temperature, change the setpoint remotely, and overlay schedules and alarms. That overlay, what we now call the BMS, was the first governance layer. It was never the control system itself. It was the thing that managed the control system.
The BMS Is Already a Governance System
What we call a building management system is already a governance system. It sets the rules. It enforces limits. It decides when to alarm and when to stay quiet. The control system keeps the space at 72 degrees. The BMS decides whether the occupant can adjust it by two degrees or five. That decision is governance.
Guardrails Degrade Over Time
Guardrails are often specified at the beginning of a project and implemented with good intentions. But over time, they degrade. Operators defeat them to keep things running. Turnover means no one remembers the original rules. The result is a BMS that still exists but no longer governs. The guardrails are there in name only.
AI Does Not Need Guardrails. People Do.
A spirited debate emerged around whether AI systems require guardrails. One panelist, who recently removed the safety constraints from his AI-driven chiller optimization system, argued that guardrails were originally there to protect vendors, not customers. They allowed companies to sell AI without proving it was smart enough to operate without limits. Now that the technology has matured, the guardrails can come off. Another panelist countered that the real issue is not whether AI needs guardrails, but whether the humans responsible for buildings will trust systems that lack them. Trust, not capability, is the bottleneck.
Standards Provide a Foundation for AI
While AI can develop its own optimized sequences, the panel agreed that industry standards like ASHRAE Guideline 36 provide a critical starting point. They define what good looks like. They create a common language. For AI developers, they offer a known baseline from which to innovate. Having a standard rulebook makes it far easier to build an AI agent that can check compliance, optimize performance, and explain its decisions.
The Biodegradable BMS Is an Iterative Process
No one will walk into a building tomorrow and rip out the BMS. Instead, the panel described a gradual erosion. AI will first tackle the biggest problems, chiller optimization being the classic example. As trust builds and results appear, it will take on more. Scheduling, alarming, fault detection. Piece by piece, the functions that once belonged to the BMS will migrate to AI agents running in the cloud, at the edge, or embedded in controllers. The BMS will biodegrade back to its original purpose: fundamental control.
Products Must Be Made AI-Friendly
A key question emerged for manufacturers and vendors. What features in today’s control products lend themselves to AI? The panel suggested that the industry needs to think carefully about how to expose data, enable override capabilities, and support external optimization without creating chaos. This is not about replacing products but about making them ready for a world where intelligence is distributed rather than centralized.
Governance at the System-of-Systems Level
The PAE Living Building in Portland was cited as a real-world example. The building has multiple siloed systems, HVAC, lighting, battery storage, and solar, all running their own standalone AI. The challenge now is not controlling any single system but providing governance across the system of systems. This is governance at a new scale, one that demands interoperability, semantics, and a clear framework for how AI agents interact.
Self-Governance Is Already Happening
The panel noted that the most successful AI deployments today rely on self-governance. The developers of these systems set their own boundaries, test their own limits, and learn from their own mistakes. This organic approach, driven by real-world results rather than committee-defined rules, may ultimately be more effective than top-down mandates.
The Impact
If the BMS biodegrades, the industry loses a familiar anchor. For decades, building owners have known what a BMS is, even if they did not fully understand how it worked. Replacing that mental model with a distributed, AI-driven ecosystem will require new skills, new contracts, and new ways of thinking about accountability.
For control manufacturers, the shift is equally profound. If AI can optimize a chiller better than a legacy BMS, the value proposition changes. The product is no longer the software running on the basement server. The product is the AI agent, the data it consumes, and the outcomes it delivers. This opens the door to new business models, subscription-based intelligence, performance guarantees, and continuous improvement rather than static installation.
For building owners, the promise is finally realizing the efficiency gains that have been talked about for decades. One panelist’s example, showing a move from 48 percent energy savings to over 60 percent after removing guardrails, illustrates the potential. But it also illustrates the risk. Without guardrails, the AI must be trusted. And trust is built on experience, transparency, and results.
The biodegradable BMS is not a prediction of obsolescence. It is a description of evolution. The governance layer built in the 1990s and 2000s served its purpose. It gave us visibility, remote control, and basic rules. Now, a new generation of intelligence is emerging, one that can do more with less, learn continuously, and adapt to changing conditions without human intervention. The systems we install today will not look the same in a decade. That is not a problem to be solved. It is progress to be embraced.
Industry interest is increasing as we cross over into several industries.
Our messages are providing thought leadership
https://www.automatedbuildings.com/2026/03/we-now-have-a-thought-leadership-liaison/
Follow Ken. He has gained over 300 followers in the last few weeks
