Explore capability
Automation for AI building control.
The ranked queue proposes the change. You grant write access one scope at a time, and choose whether each change waits for a person or runs on its own.
Book a time directly- Writes only where you allow
- Shows its work before anything moves
- Included in Professional
Lower supply-air setpoint on unoccupied weekends
20.5 °C→19.5 °CSat–Sun, 00:00–24:00
- Evidence
- Return air 2.1 °C above target for 3 consecutive weekends. Occupancy sensors: 0.
- Reasoning
- VAV boxes fully closed, so the AHU is heating air nobody uses. The physics model confirms no comfort penalty.
- Expected effect
- -4.2 % heating kWh on this AHU, about 3 400 kr per month. Reversible.
- Scope
- Floor 4 HVAC setpoints · granted · human review
From finding to write-back
Detection ranks the problem. Automation is where you let Explore act.
Automation is Explore's permissioned BMS write-back for AI building control. The BMS analytics layer detects, forecasts and ranks. Automation is the action layer on top: it proposes a change, shows the reasoning and the expected effect, then writes only inside a scope you granted, through the FrostLogic Edge Agent.
Detect
BMS analytics
Reads your BMS, meters and sensors. Finds drift, waste and faults. Ranks them by the cost of waiting.
Reads only
Propose
Automation
Turns a ranked finding into a concrete change: a setpoint, a schedule or a control tweak, with evidence and expected effect.
You decide: review or automatic
Write
Edge Agent
Installed on the BMS machine. Writes only to points inside a granted scope. Everything else stays untouched.
Stays off until you grant a scope
Scoped write access
You set the leash, one scope at a time.
A scope is a set of BMS points you choose to open. Inside each scope you pick the mode. Outside it, Explore can read but never write.
That is supervisory control with a leash: Explore proposes, and you set how much runs unattended. New installs ship with no scopes granted, so nothing is written until you turn a scope on.
What Automation does
Propose, schedule, grant, write
Proposals with reasoning
Every proposed change lands with the evidence trail, the physics reasoning, and the expected effect. Nothing invented at inference, and nothing invented at action either.
Time and event triggers
Schedules can fire on a clock or on an event from the ranked queue. Useful for HVAC optimization where a setpoint or schedule only makes sense under a stated condition.
Human review or automatic
Inside a granted scope you choose the leash: wait for a person, or let the change run on its own. Points outside that scope stay untouched.
Scoped write via Edge Agent
The FrostLogic Edge Agent installs on the BMS machine and reads and writes points. Write access starts off. You turn it on one scope at a time.
How it works
Propose, decide, write
Propose
Change with reasoning
Explore proposes a setpoint, schedule or control tweak with the evidence, the reasoning and the expected effect.
Decide
Review or auto
Inside each granted scope you choose: wait for a person, or let matching proposals run automatically.
Write
Scoped Edge Agent write
The FrostLogic Edge Agent writes only to the points inside that scope. Everything else stays untouched.
How the three approaches compare
Static schedules, black-box autonomy, or permissioned write-back
| Raw BMS schedules | Fully autonomous write-back | Explore Automation | |
|---|---|---|---|
| Starting state | Hardcoded timers and setpoints | Writes continuously once connected | Read-only until you grant a scope |
| What you see | Schedule table, little context | Opaque control loop | Proposal with reasoning and expected effect |
| Who decides | Whoever edited the schedule last | The vendor model | You set human review or auto per scope |
| Scope of write | Whatever the schedule touches | Broad HVAC write-back by default | Only points inside a granted scope |
| Trigger | Clock only | Continuous closed loop | Time or event from the ranked queue |
| Honesty posture | Transparent but dumb | Fast, hard to audit | Shows its work; you set the leash |
At a glance
The mechanics
Automation sits on top of the same sensor-intelligence layer as detection and forecasting. It does not replace the BMS, the integrator's logic, or a CMMS work order.
Explore reads. Nothing is written.
A person approves the change before it is written.
Only inside the granted scope.
Questions
Automation, answered
Yes, once you turn it on. Write access starts off. You grant it one scope at a time, and Explore proposes the change with its reasoning and the expected effect before anything moves. You choose whether that proposal waits for someone to approve it, or runs automatically inside the scope you set. Points outside that scope stay untouched.
Explore starts read-only. No write scopes are granted out of the box, so the Edge Agent does not write anything until you turn a scope on. Read-only until you grant a scope.
No. Explore is Sensor Intelligence. It reads your existing BMS and can write back inside scopes you grant. It does not replace the control system, the integrator logic, or a BAS. Building automation systems remain the control layer; Automation is the permissioned action layer on top.
No. Explore can propose a change or surface a finding for a technician, but work-order scheduling, spare parts and crew dispatch stay in your CMMS. Automation writes to BMS points you grant; it does not become your maintenance management system.
Automation is included in the Professional package. It is not sold as an add-on and it is not locked behind a higher tier.
Some products push continuous closed-loop HVAC changes with little operator visibility. Explore shows its work: every proposal carries reasoning and expected effect, and you set how much runs unattended inside each granted scope. We do not match blanket autonomy. You keep the leash.
See Automation against your own BMS
Connect one building, keep writes off until you are ready, and walk through a proposed change with its reasoning before anything moves.