Honeywell launches AI agents to manage building alarms, maintenance and energy

Facility operators are getting a way to let AI move beyond flagging equipment problems and help coordinate responses across existing systems. Honeywell Technologies introduced Forge Cognition on October 14, 2026, combining AI agents with live facility simulations. Developed with NVIDIA and Dell Technologies technology, the platform lets operators choose how much control to hand over.

From equipment data to operating decisions

The platform targets buildings, campuses and industrial operations, where teams must sort through alarms, plan maintenance and manage energy use. Honeywell says it connects to existing customer deployments, adding AI capabilities rather than requiring an entirely new facility system. Its installed base spans more than 300,000 sites, giving the company a large potential market for the platform.

Forge Cognition puts equipment and systems data into a structured knowledge base, alongside Honeywell's engineering expertise. Agents use that context to interpret equipment behavior and relationships between systems. Real-time digital twins—simulations of the physical facility—let operators test operating changes before choosing a response. Honeywell describes three initial applications:

• Alarm management: Agents classify and verify building-management alarms so operators can prioritize events needing attention and reduce false positives.
• Equipment health: Agents examine real-time heating, ventilation and air-conditioning data to identify emerging problems and guide maintenance before equipment issues disrupt operations.
• Energy use: Digital-twin simulations help teams evaluate operating changes and look for energy savings while maintaining occupant comfort.

Automation is adjustable. Customers can run Forge Cognition in manual, semi-autonomous or autonomous modes. Operators set the level of oversight and can override the platform at any time. That puts the decision about delegating action with the facility team, rather than making full autonomy a requirement for using the system.

NVIDIA supplies models; Dell supplies infrastructure

NVIDIA's contribution reaches across several jobs. Its Nemotron open models power agents that interpret operational data, including a True Alarm agent that improves through operator feedback. Omniverse libraries support physically accurate simulations. Cosmos models, paired with NVIDIA's Metropolis video-search-and-summarization blueprint, support vision agents that understand, search and summarize live or recorded sensor data.

The platform uses NVIDIA RTX PRO graphics processors. Honeywell says support is expanding toward far-edge deployments—computing closer to physical operations—with NVIDIA DGX Spark. Dell provides infrastructure through its AI Factory with NVIDIA, including PowerEdge XE servers, data and storage engines, cyber-resiliency tools and services supporting model development, training and operation.

Up to 90% reduction, on averageUnnecessary alerts

Honeywell reports this reduction at NVIDIA's Voyager campus, saying fewer unnecessary alerts let operations teams focus on events that matter.

Up to 10% potential reductionEnergy and utility consumption

The pilot identified potential savings. Honeywell does not present this figure as an achieved reduction in consumption.

Above 95%Comfort score

Honeywell says the pilot maintained this score, calculated as a weighted average of temperature, humidity and carbon-dioxide measurements across zones.

Pilot sites span campuses, a hotel and a mall

Voyager, at NVIDIA's headquarters in Santa Clara, California, is one of several pilot locations. Honeywell also names the Fairmont Hotel in Dubai, NEX Mall in Singapore, Vanderbilt University in the United States, and its own campuses in Bangalore and Richardson, Texas. The Voyager deployment also monitors real-time equipment data to provide health predictions for maintenance planning.

For prospective customers, the immediate step is deployment planning. Honeywell says customers can engage it now to identify and scope Forge Cognition deployments, with capabilities expanding over time. That is the availability offered at launch: a path to defining a facility deployment, alongside early pilots showing how the platform works in operating buildings.