The building is a process too.
Energy is a top-three line item on most industrial sites and the building systems that drive it are usually the least instrumented thing on the property. This is the same stack — admitted agents, granted reach, retained evidence — pointed at the plant’s own building.
What it catches
- Energy use intensity by area, line and shift rather than one site meter
- HVAC running against itself, or conditioning space nobody is using
- Compressed air leaks and off-shift base load that should not exist
- Equipment left running through non-production hours
- Setpoint drift away from what the schedule actually calls for
Where it reports
- BMS and building automation through governed interfaces
- Energy and sustainability reporting, including EUI by period
- CMMS work orders where a fault rather than a schedule is implied
- Existing dashboards via outbound webhooks
- A retained record of every reading and every change proposed
What actually runs, and where
- Task
- Signal models over meter, sensor and BMS data, with CV where only a gauge exists
- Typical target
- x86_64 gateway or Raspberry Pi 5, alongside the existing BMS
- Mode
- Continuous sampling with scheduled reporting
- Reach
- Whatever you grant. Read-only, or setpoints you name explicitly
Every one of these passes the Oddy test against your named device before it is admitted. A candidate that does not fit the device envelope never receives an edge directive.
Common questions
Does it change setpoints on its own?
Only where you grant it that reach, and the grant is specific: named setpoints, named equipment, bounded ranges. Plenty of sites run it read-only for a season first and act on the recommendations by hand. Whichever you choose, every change it proposes or makes is recorded against the agent that made it.
How is this different from our existing BMS?
A BMS executes a schedule and reports what it did. It rarely tells you that the schedule itself is wrong, that two systems are fighting, or that a zone has been conditioning empty space since a line moved. This sits alongside the BMS, reads what it reports, and reasons about whether it should be doing that at all.
Do we need to replace our building controls?
No. The agent reads from the BMS and the meters you already have, and writes back only where you grant it. The controls, the interlocks and the equipment safeties stay exactly as they are.
Can it produce the numbers we report on?
That is a large part of why sites run it. Energy use intensity by period, by area and by line, from retained readings with the model identity attached — which is a materially different position than assembling the figure from a spreadsheet at year end.
Start with one building.
Send a year of meter data and your BMS point list. The free audit comes back with where the load actually goes and what a first agent would watch.