AN ADAPTIV SYSTEMS PRODUCT
Physical AI for theBuildings of the World.
Merlin is a new kind of operating layer for physical spaces. Instead of one general-purpose dashboard, it deploys a fleet of specialized AI agents — each one purpose-built for a single operational concern, each one reasoning continuously over the real-world signals coming from your buildings, and each one closing the loop between what's happening and what should be done about it.
Your AI co-worker · Meridian HQ
Ask Merlin about any sensor, room, or SLA…
Merlin acts autonomously. Approve or hold any proposed action.
THE SHIFT WE'RE BUILDING FOR
AI is moving into thephysical world.
AI has spent its first decade living in software. It writes, it summarizes, it answers questions, it generates images. Useful, but abstract — confined to screens, text, and pixels.
The next decade is different. AI is moving into the physical world — into the buildings, the spaces, the operations where things actually happen. Physical AI isn't about smarter chatbots. It's about agents that perceive real-world signals, reason about them in context, and take action that affects the physical environment.
Buildings are where this shift starts. They generate continuous streams of telemetry — temperature, occupancy, cleanliness, security, equipment health — and they require continuous decisions and actions. They're the perfect domain for AI that doesn't just observe but operates.
Merlin is built on the conviction that the right way to build physical AI isn't one giant model trying to do everything, but a coordinated fleet of specialized agents — each one expert in its concern, each one accountable for its outcomes, each one part of a larger intelligence working on the building together.
Generic AI vs Specialized Agents
The loop
Inside an Agent's Loop
Sense
Reads the physical world through authenticated hardware — NFC badge-ins, occupancy, VOC, temperature, vibration. Ground truth, never an estimate or a self-report.
Analyze
Understands space, time, and workflows. Correlates signals across the building, reduces noise, and predicts escalation before it becomes an incident.
Orchestrate
Reasoning runs on deterministic, verifiable data. Merlin prioritizes by operational impact and decides who to dispatch, what to reroute, and when to hold.
Act
Proposes and executes within its delegated authority — files the work order, dispatches the contractor, drafts the report. Every action pauses for human approval at the point of escalation.
Verify & Learn
Watches the outcome and learns each site's patterns over time. A building run for 12 months has a sharper model than one run for 1 — the system improves with every cycle.
THE AI LAYER
Three principles keep Merlin trustworthy.
Physical AI only works in production if operators can trust it. Merlin earns that trust by design — the model is the interface, the hardware is the truth, and the system compounds with every building it runs.
LLMs as interface only
Operators talk to Merlin in plain language. The model translates the question into a structured database call — but all reasoning, dispatch logic, and workflow execution runs on deterministic, verifiable data, not on the language model.
Sensor data as ground truth
Every recommendation Merlin makes is grounded in authenticated hardware data — NFC badge-ins, occupancy counts, VOC readings, temperature, vibration. No estimates. No self-reports.
Compound intelligence
Merlin learns each site's usage patterns over time. A building managed for 12 months has a more accurate model than one managed for 1 — the system improves with every deployment cycle.
The hypervisor
The whole building, on one canvas.
See every floor at once — each agent's work mapped to the space it happened in. Open any item and Merlin shows its full reasoning: what it sensed, what it proposed, the confidence, and the outcome it's tracking. Nothing is hidden in the cloud.
THE CONVERSATION BOX · A 24/7 CO-WORKER
Ask in plain language. Act from any screen.
Instead of hunting through dashboards, operators ask Merlin a question — and it reasons across the physical infrastructure to answer, propose, and execute. Every proposed action requires human approval before it runs.
Merlin acts within its delegated authority and pauses for validation at the point of escalation. Every action is logged.
Answer operational questions
“Which zones are at SLA risk right now?”
Propose and execute actions
“Reroute the cleaning team on floor 4 given the two absences this morning.”
Generate reports
“Produce the monthly contractor performance pack for the GuardWatch Security review.”
Identify financial risk
“What is the projected penalty exposure if current breach rates continue?”
WHAT YOU SEE
One operating picture for the whole estate.
The conversation box is the fastest way in, but Merlin also surfaces the building visually — live, historical, and predictive — across one site or hundreds.
Live maps & heatmaps
Occupancy, environment, and activity across every floor in real time.
Event timelines
Every signal, decision, and action in chronological order.
AI alerts
Merlin flags what needs a human, ranked by operational impact.
Incident replay
Step back through what happened, signal by signal.
Predictive alerts
Escalation surfaced before it becomes an incident.
Resource coordination
Dispatch, reroute, and hand over work across teams and contractors.
Multi-site visibility
One pane of glass across the entire portfolio.
Conversation box
A 24/7 co-worker on every screen, in plain language.
A DAY IN THE LIFE OF THE FLEET
One day across the fleet.
Six specialized agents, each one expert in its concern, working on the building together. Here's a slice of a day across the fleet.
A medication-storage refrigerator drifts toward the upper compliance threshold. The agent recognizes the regulatory regime — pharmaceutical — opens a compliance-grade event, alerts the responsible pharmacist, and stages the audit-grade reading for the next inspection.
Comfort-band drift on a rooftop unit ahead of morning occupancy. The agent distinguishes seasonal variation from equipment degradation, schedules a predictive-maintenance visit, and adjusts setpoints so the morning arrives at the right temperature anyway.
Missed-check pattern at the early shift in a corridor restroom. Telemetry says the area is busier than the SLA assumes. The agent dispatches a human, logs the gap, and adjusts the recurring schedule for the room.
A storage zone briefly crosses its threshold during a delivery cycle. The agent recognizes a transient versus an actual failure, records the event end-to-end, and confirms compliance is maintained.
Underutilized zone identified on the lower level; a reserved spot misuse on the executive level is flagged with evidence. The agent coordinates enforcement, surfaces live availability to the Smart Display at the entrance, and drivers stop circling.
An access pattern at a side entrance falls outside the routine. The agent distinguishes routine from suspicious, escalates to the on-call responder, and coordinates the response — with the full event trail attached.
Tomorrow, the fleet does it again. The roster grows as new operational concerns become learnable. The loop is always the same.
THE MERLIN AGENT LIBRARY
Specialists. Coordinating.
Cleaning agent
Reads cleaning telemetry, missed-check sensors, occupancy patterns, and SLA terms. Decides when human dispatch is warranted, when the situation can wait, and when a recurring pattern suggests a deeper accountability problem.
HVAC agent
Watches comfort-band drift, equipment behavior, and energy patterns. Flags anomalies before they become failures. Distinguishes seasonal variation from equipment degradation. Schedules predictive maintenance.
Cold-chain agent
Monitors temperature-controlled storage end-to-end. Catches deviations the moment they cross compliance thresholds. Knows the regulatory regime — pharmaceutical, food, biological — for whatever's being stored.
Pharmacy temperature agent
Purpose-built for clinical refrigeration. Continuous compliance with the regulatory standards governing medication storage. Audit-grade traceability for every reading and every action.
Security agent
Watches access patterns, anomalies, and incident signals. Distinguishes routine from suspicious. Coordinates response when something requires it.
Parking agent
Tracks occupancy across lots and garages, identifies underutilized zones, and detects violations and reserved-spot misuse. Surfaces capacity intelligence to operators in real time and feeds live availability back to the Smart Displays at entrances — so drivers know where to go before they circle.
Why we are unique
Grounded in hardware we own.
Every agent is grounded in five-plus years of telemetry from devices Adaptiv designs, manufactures, and operates across 8,000+ buildings — credible from day one, not learning your building from a cold start.
WHY A FLEET, NOT ONE MODEL
Specialization is the whole point.
As the data foundation deepens and new operational concerns become learnable, new agents are trained and added to the library — each grounded in the same real-world telemetry that makes the existing roster credible from day one.
Stakes are real — judgment has to be specific
A pharmacy refrigeration deviation is a regulatory event, not a tip. An HVAC anomaly in a hospital is a patient-safety question, not a maintenance ticket. Specialized agents bring vertical-specific judgment by design.
Improvable one agent at a time
When the cleaning agent gets better, cleaning operations get better — without disturbing HVAC, cold-chain, or anything else. The system gets smarter agent by agent, rather than depending on one monolithic model that has to be right about everything at once.
Mirrors how the real world organizes expertise
Nobody hires one person to do plumbing, refrigeration, fire safety, and access control. You hire specialists, and they coordinate. Merlin's agent fleet mirrors that structure — specialists, coordinating.
Grounded in the data foundation
Every agent is grounded in five-plus years of telemetry from devices Adaptiv designs, manufactures, and operates across 8,000+ buildings. Agents are credible from day one — not learning a customer's building from a cold start.
THE FOUNDATION
The Foundation: Our Devices, Our Data, Our Buildings
We design, manufacture, and operate our own devices.
Smart Displays, sensors, and connected hardware purpose-built for the physical environments they serve. Most companies trying to build AI for physical spaces start from scratch — software layered on whatever third-party sensors happen to be installed. Adaptiv is different. The hardware came first.
Five-plus years. Eight thousand-plus buildings. Global coverage.
A continuous, real-world data foundation. Corporate headquarters, warehouses, healthcare facilities, banking branches, government and institutional spaces. Climates, building stocks, regulatory regimes, occupancy patterns, cultural norms of how spaces are actually used. The data came from the hardware.
Every specialized agent is grounded in that foundation.
Merlin's agents are credible from day one in a customer's environment. They're not learning your building from a cold start — they're applying years of physical-world judgment, refined across thousands of buildings before yours, to the specific operational picture you give them. The AI is built on the data.
The hardware came first. The data came from the hardware. The AI is built on the data. Software-only competitors can build agents. They can't ground those agents in years of physical-world telemetry from hardware they own and operate.
Our vision
Buildings are the first place we proved it. Buildings are not the limit.
The companies that figure out physical AI will be the companies that built the operating layer for the physical world — the equivalent of what the cloud-computing platforms became for software in the last era. Merlin is one of the early reference implementations of that layer.