Harness

Building and civic-facility agent

In buildings and civic facilities, comfort and compliance are hard floors. The harness freezes temperature, illuminance and time-of-use rules as non-negotiable constraints; the model gives itemised strategies and retrofit narrative inside the feasible set, outputting executable setpoint advice or BA template parameters that the system checks before control or work orders.

Feasible solutions under comfort and tariff dual constraintsAdvice–check–approve–dispatch four stages, fully traceableEffect write-back supports continuous iteration and retrofit ranking

Book a demo AI overview

Capabilities

Domain capabilities

Load and carbon-intensity co-optimisation

Under time-of-use tariffs, comfort and indoor-environment standards, optimise HVAC, lighting and power time-of-use strategies; output executable setpoints or control sequences and estimate daily/weekly kWh savings and Scope 2 abatement.

Efficiency benchmarking and retrofit ranking

Normalise energy and carbon intensity by floor area, format and climate zone; find abnormal branches and retrofit candidates (heat-pump substitution, VFDs, group control) and produce a list roughly ranked by abatement cost for budget decisions.

Operating-loop snapshot

01Itemised data hygiene02Load comfort and tariff constraints03Load shape and benchmarking04Control-strategy draft05Hardware and procedure checks06Approve and dispatch07Effect write-back

Operational Loop

Operating loop

From connection alignment to review versions, every step can be checked, trailed and approved.

7 steps

Itemised data hygiene

Pull lighting, HVAC, power items and outdoor weather; rule-detect abnormal readings and drop untrustworthy intervals so later optimisation sits on clean input.

Load comfort and tariff constraints

Inject indoor-environment standards, occupancy windows and time-of-use tariffs. The harness encodes comfort bands as optimisation bounds; the model may not suggest out-of-band setpoints.

Load shape and benchmarking

Normalise by floor area, format and climate zone; find peaks and idle run; call stats and simple clustering to list abnormal branches for the model to explain.

Control-strategy draft

Generate temperature setpoints, lighting groups and schedule advice; each item carries applicable hours and expected savings magnitude (as a range), avoiding false precision.

Hardware and procedure checks

Confirm BA points exist in the ledger and are not maintenance-locked; procedure bans (e.g. cooling in winter) intercept immediately and feed the model a rewrite.

Approve and dispatch

After property/energy-owner approval, strategies dispatch as templates or APIs. Low-confidence or large cross-zone advice requires explicit human tick confirmation.

Effect write-back

Compare post-implementation item curves to baseline and write back to the harness indicator store — a traceable before/after for later retrofit ranking and budget talks.