What would AI Native look like, compared to BACnet Native, that changed the industry?

Looking at the evolution of network architecture through the lens of Vint Cerf and Mallik Tatipamula’s Continuum of Network Architecture, we can see a beautiful, repeating rhythm in technology.

Systems always start as rigid, centralized, vendor-locked monoliths. Then, a disruptive architectural movement comes along, completely dissolves those silos, and distributes the intelligence.

To understand what AI-Native will look like in building automation, we have to look at how BACnet-Native first broke the mold, how Cloud/Semantic-Native (Haystack/RDF) shifted the landscape, and how AI-Native is about to change the game entirely.

Here is the continuum of how building technology is evolving from Syntax to Orchestration to Semantics and Cognition:

1. The BACnet-Native Era: Standardizing the Syntax (Packets & Protocols)

• The Network Architecture Parallel: IP-Native Networking (Moving from Circuit-Switching to Packet-Switching).
• The Building Automation Reality: Before BACnet, buildings ran on proprietary “circuits.” Vendor A’s thermostat could not talk to Vendor B’s chiller because they didn’t speak the same language. BACnet arrived and standardized the syntax. It turned control commands into standardized “packets” of data.
• The Industry Shift: It broke the proprietary lockdowns. Suddenly, upstart companies didn’t have to build an entire closed empire; they just had to build a better BACnet controller. It forced the major manufacturers to open up their communication protocols or lose market share entirely.

2. The Cloud/Semantic-Native Era: Standardizing the Meaning (Workloads & Ontologies)

• The Network Architecture Parallel: Cloud-Native Networking (Moving from Hardware Appliances to Programmable Workloads and Service Meshes).
• The Building Automation Reality: While BACnet lets devices talk, it doesn’t tell you what the data means. You just got a packet that said Analog-Value 1 = 72. You still needed a human to manually map that point and explain that it belonged to a specific VAV box on the 4th floor.
• The Industry Shift: This is where Project Haystack, Brick Schema, and RDF/TTL graphs stepped in. They decoupled the data from the physical hardware box. Just like cloud-native technology abstracts infrastructure into portable workloads, semantic data abstracts a building’s architecture into a digital graph. It allows us to scale from managing single People (manually mapping points) to managing massive autonomous Portfolios.

3. The AI-Native Era: The Cognitive Fabric (Agents & Intent)

• The Network Architecture Parallel: AI-Native Networking (Moving from Reactive Workloads to Autonomous, Cooperating Agents).
• The Building Automation Reality: Today’s smart buildings are still largely reactive. They respond to a threshold after it’s broken (e.g., the room gets too hot, so the VAV box opens). If an automated sequence breaks, a human still has to log in, read the charts, and troubleshoot the logic.
• What AI-Native Looks Like: In an AI-Native building ecosystem, intelligence is no longer an application “overlay” or a software tool you buy; it is an intrinsic property of the building infrastructure itself. The basic atomic unit shifts from a static control loop to an Agent.

What does an AI-Native Building Ecosystem actually do?

According to the continuum framework, an AI-Native infrastructure introduces a semantic control plane in which independent AI agents continuously perceive context, reason about intent, and coordinate with peer devices in real time.

Instead of rigid lines of code, an AI-Native building relies on these core dimensions of cognition:

1. Semantic & Intent Routing: Instead of a programmer hard-coding a BACnet command like “If Outside Air Temp > 22C, close damper to 20%,” an asset manager issues a high-level intent: “Optimize this portfolio for a 15% carbon reduction this week without letting tenant comfort scores drop below 90%.” The AI-native fabric propagates this intent down to individual equipment agents, which negotiate how to achieve it.
2. Predictive Autonomy: Equipment doesn’t wait for a sensor to trip. AI-native agents actively ingest weather forecasts, occupancy patterns, and utility grid pricing to dynamically model and adjust energy loads before an energy surge or peak-pricing event occurs.
3. Cooperating Agentic Networks: The chiller, the air handlers, and the zone thermostats operate as a localized “do-ocratic” community of agents. If an AHU detects a failing bearing, its agent communicates with the zone agents to gracefully shift the cooling load to an adjacent unit, notifies the maintenance agent to order the part, and updates the portfolio’s digital twin—all without human intervention.
4. Reflective Telemetry & Modelling: The system continuously runs simulations against its own RDF/TTL graph definitions. Before the AI alters a major plant sequence, it validates the change in a sandboxed digital model to eliminate operational risk.

The Ultimatum: Open Up or Get Replaced

Just as the majors in the 1990s resisted BACnet to protect their proprietary hardware, legacy companies today are trying to restrict data access to protect their proprietary clouds and API fees.

But as Cerf and Tatipamula point out, history shows that monolithic boundaries always dissolve in favor of distributed, open, and collaborative fabrics.

An AI cannot think, predict, or act if a legacy vendor has locked its data inside a silo. Therefore, the industry is forcing the same ultimatum we saw thirty years ago: Vendors must provide completely open, standardized semantic data access (RDF/TTL), or they will be entirely box-specified out of the next generation of AI-Native portfolios.

The upstarts of tomorrow aren’t just building cheaper controllers; they are building autonomous agents that will make closed systems look like mechanical dinosaurs.

Resources:

https://cacm.acm.org/blogcacm/the-continuum-of-network-architecture/

The Case for AI Native Registries, Operational Context, and Live Building Intelligence