Morning alarms, tenant calls, evening wind-down

The day starts at 06:30 with overnight alarms scattered across three or four dashboards, and the morning peak is half an hour away without a clear picture of whether the lift lobby will hold the inbound crowd. By 09:30 the first tenant complaint comes in, by 12:00 the cafeteria is full and someone has to dispatch cleaning, by 16:00 you are trying to write a weekly report from data that lives in five different systems, and at 18:00 you walk the building because the schedule does not match how people actually leave.

This is the day a building manager actually has, the buildings get more complex, the systems get more varied, and the team stays the same size or shrinks. There are not enough hours, and the gap between what the building demands and what the team can deliver keeps growing.

Now imagine adding one more person to the team, not a consultant, not a service contract, not another dashboard, but a teammate who handles the routine, sees the building in real time, and knows when to ask a human for help. That is what the Building AI Operator is for.

Hire one for your building

A role, not a product

The Building AI Operator joins your team the way a human operator would: you give it a job description covering what it should handle on its own, what it should recommend, and what it should always escalate. You configure how careful it should be at first, and you decide what happens automatically versus what requires a human in the loop.

The operator lives where your team already works: the chat tools your team uses become the front door for most buildings, with conversations happening in plain language, alerts arriving with context rather than raw data, and reports showing up in the channel where the team already meets. There is also the Digital Twin, the spatial view of the building, for the moments when somebody needs to look at a floor and see what is happening.

Most of the time, you do not log in because the operator comes to you.

What it does

Three modes, seeing, acting, reporting

It sees

The Vision Sensors give the operator real-time perception of people: where crowds form, where queues build, which floors are filling and which are emptying. The lift system tells it what cars are moving where and what the queues at each lobby look like, access control adds a thin badge layer on top, and where building systems are connected, ventilation status, lighting state, and equipment alarms, the operator picks those signals up too.

The result is a single, coherent picture of what is happening in the building, who is affected, and what is changing, rather than a stack of disconnected dashboards.

It acts

This is where the word operator matters, the Building AI Operator does work rather than just summarising events for a human to act on.

Elevator control and optimisation. When the lobby starts to fill, the operator pre-positions cars from upstream flow data through the elevator integration before queues form, and when an escalator faults it redirects flow to adjacent banks and increases lift service to compensate. Vision Sensors see the people and the operator turns that into smoother vertical transport: this is the integration we know best, and the one most customers feel first.
Maintenance tickets. Equipment anomalies become work orders with crowd context attached, filed in whatever maintenance system the building already runs, so the technician arrives knowing "escalator E3 stopped at 08:12, 340 people expected through this corridor in the next 15 minutes".
Security dispatch. When the system detects unusual presence, such as a restricted zone occupied after hours or a crowd forming in an unexpected area, security gets a notification with location, context, and a recommended response, turning reactive patrols into targeted dispatch.
Alerts with context. When density rises on Platform 2, the alert lands in the safety channel with the trend, the time of last similar event, and a recommended response rather than raw numbers without meaning.
Wider building systems, where connected. Ventilation, lighting, access control, and similar systems benefit from the same people-aware signal: for example, restricted-zone access flagged with crowd context, or ventilation behaviour reviewed against actual occupancy. We expand these integrations as customers ask for them, and vertical transport is where we lead today.

The elevator example is the clearest expression of what the operator is for: Vision Sensors see people, and the operator translates what it sees into commands that change how the building behaves.

It reports and recommends

The operator drafts the routine reports the team used to assemble by hand, a daily operations digest summarising zone-by-zone occupancy and elevator performance, a weekly equipment reliability summary listing faults, mean-time-to-respond, and upcoming maintenance, a monthly utilisation report with floor-by-floor heatmaps and trend arrows, and a quarterly sustainability pack that pulls everything together for an ESG auditor.

The operator also recommends changes backed by simulation: "Closing gate 3 during the morning peak would reduce queueing by an estimated N%, based on last quarter's flow." The team sees the proposal, the simulated outcome, and decides. See Simulation → for how that works.

Where it plugs in

Integrations that match the building you already run

The Building AI Operator works with the systems your team already uses:

Team chat: the primary front door, where alerts, reports, approvals, and conversational queries arrive in the channels the team is already in.
Building management systems: the operator reads equipment status and writes commands at a safe priority level, with full audit and reversibility, covering the modern installed base and most legacy controllers.
Maintenance and workplace systems: work-order creation, asset lookup, and status sync, with the operator opening tickets if the building runs on tickets or handing off through a configurable webhook if it runs on a different workflow.
Vertical transport: lifts and escalators, integrated for both visibility and control.
Email: the fallback for tenants without a chat tool, and the channel for asset-manager PDF reports.
Digital Twin: the spatial view, used when a human needs to see a floor and walk through what happened.

A building does not need every integration to start, a pilot can begin with Vision Sensors, the team's chat tool, and vertical transport, and grow from there.

You set the rules

Configurable, not coded

Every team draws the line between "system handles it" and "human decides" in a different place, a metro safety team will keep crowd thresholds advisory at first, an office facilities team may automate routine queue and meeting-room workflows from day one, and a hospital will keep nearly everything human. All of those are correct.

The AI Operator supports the full spectrum from fully manual through recommend-only and auto-approve-with-notification to fully automatic, with the setting configurable per use case, per zone, and per time of day. Defaults are cautious, and every automated action, every override, and every rule change is logged.

For writes into building systems, there is a hard architectural rule the operator never breaks: it writes at the manual operator priority level, beneath fire, smoke, emergency, and any safety interlocks. A higher-priority command from the building system or the life-safety layer always wins, and the operator backs off and logs the override.

Why this matters

The productivity story

The point of hiring the Building AI Operator is not to replace your team but to give them the leverage they need to run a more complex building with the same headcount.

The routine work of alert triage, queue monitoring, security dispatch, and manual report assembly gets handled, while the consequential decisions on evacuation routing, capital planning, tenant relationships, and security escalation stay with the people whose names go on the regulator's letter. The operator gives those people more time, better information, and a clearer audit trail when they need to defend a decision.

10xProductivity gain compared to manual monitoring and decision-making.

The 10× target is what the platform is designed around, and customers see the value differently: some report fewer tenant complaints, some report smoother vertical transport, some report faster incident response, but all of them describe the same underlying shift: the team spends more time on judgement and less on transcription.

What it does not do

Machines think, people decide

The AI Operator does not make safety-critical decisions on its own, evacuation routing, security escalation, and restricted-zone breaches with people present all route to a human by default, and the operator may accelerate the human's decision by pre-assembling context but the call is always human.

The AI Operator does not override the priority hierarchy of the building's safety systems, because fire, smoke, emergency, and safety interlocks always win and the operator's writes sit at manual operator priority where they belong.

The operator does not see images or biometrics, it inherits the Vision Sensor's privacy posture, and since Vision Sensors are physically incapable of capturing identifying data the operator simply cannot work from it.

And the operator does not replace the team; it is a hire that frees the team's hands, and the buildings that get the most out of it are the ones where the human operators are already good at their jobs and want more time to do them well.

What's next

The memory and the foresight

The AI Operator is most useful when it has memory and foresight, the Digital Twin is the memory, which is the connected, queryable model of the building that lets the operator answer "is this normal for a Tuesday?", and Simulation is the foresight, which is the way the operator tests a recommendation before it lands in front of you.