Conventional software AI
- Produces text, summaries, or recommendations on a screen
- Depends on a person to interpret the output and coordinate the next step
- Reasons mainly over documents and software context
- Stops before the physical outcome is known
Physical AI for building operations
Sense the building.Reason. Act. Verify.
Physical AI connects intelligence to the physical world. Merlin reads authenticated signals from real spaces, turns them into operational decisions, acts within the authority you define, and verifies what happened next.
The category shift
Most AI stops at the recommendation. Physical AI closes the loop.
A building does not improve because a dashboard found a pattern. Improvement happens when the right operational decision reaches the right person or system, within the right controls, and the result is measured.
Physical AI with Merlin
- Perceives live conditions through connected devices
- Reasons over telemetry, operating rules, SLAs, and context
- Coordinates action within explicit human-defined boundaries
- Measures the result and records a continuous reasoning trail
How Physical AI works
Five steps. Running continuously.
Merlin connects perception, reasoning, action, and verification in one operating loop. Specialized agents apply that same architecture to different building concerns.
Sense
Adaptiv devices capture conditions such as occupancy, temperature, air quality, leaks, equipment vibration, and verified service activity.
Understand
Specialized agents interpret each signal against the rules, tolerances, schedules, and operating context of that space.
Decide
Merlin ranks what matters, explains the evidence, reports confidence, and selects an approved playbook.
Act
The platform proposes or coordinates the next step: dispatch work, adjust a setpoint, escalate an incident, or close a compliance gap.
Verify and learn
The same operational system checks the result, logs the outcome, and carries that evidence into the next decision.
The operating foundation
Physical AI is only as trustworthy as its data and controls.
Merlin is built on devices Adaptiv designs and operates, years of authenticated telemetry, explicit autonomy policies, and a continuous record of every decision.
Authenticated data
Signals originate in real environments and remain traceable from the device through the decision and its outcome.
Bounded autonomy
Operators set the autonomy level by agent, site, and concern. Merlin pauses whenever human judgment or approval is required.
Continuous accountability
Inputs, rationale, confidence, actions, timestamps, and outcomes form an audit trail designed for operational review.
8,000+
Buildings deployed
5+
Years of operational telemetry
SOC 2
Type 2 controls
Global
Coverage across Europe and the US
Physical AI in practice
Start with one operational problem.
The category becomes concrete when the loop is pointed at a measurable workflow. These are three places to start.
Physical AI, operating today
See how Merlin closes the loop in your buildings.
Start with a real workflow, the signals you already have, and the operating boundaries your team needs.