On-site Deployment

Local deployment when site constraints require it

Local deployment is selected only for approved BMS integration, supervised writes, constrained network conditions, low-latency needs, or resilience requirements. Work still moves from read-only monitoring to advisory operation, and only later to supervised writes after connectivity, mapping, fallback behavior, and operator workflows are verified.

This route assumes data intake and model training are far enough along to define the measurement boundary, candidate control points, and commissioning plan. If those inputs are still unclear, site work should wait rather than force a live cutover.

Responsibility

Who does what on site

The exact staffing model varies by project, but the site visit or remote commissioning window should make handoffs explicit.

Customer

Owns site access, business approval, operating constraints, and the decision to expand from monitoring to advisory or supervised control.

Approve network paths, local gateway placement, maintenance windows, and the people who can pause or override optimization.
Confirm comfort-critical periods, tenant constraints, and when the building should stay on native BMS control only.
Sign off on the approved write list, limits, and escalation contacts before supervised writes begin.

ClimaMind

Leads deployment engineering review, local runtime and telemetry validation, staged activation, and performance evidence inside the agreed scope.

Verify point mapping, telemetry quality, alarms, and operating records against the intake package and model readiness summary.
Commission read-only monitoring first, then advisory mode when the site needs staged confidence building.
Document open items, fallback events, and the criteria for moving to supervised writes or acceptance.

SI / System Integrator

Owns BMS-side integration, local operator workflow, graphics, trends, alarms, overrides, and write-path configuration.

Confirm protocol access, point names, histories, command paths, and BMS-native fallback behavior.
Configure approved read and write permissions, alarm interactions, and operator-visible status where needed.
Support issue response when the BMS rejects a command, a point goes stale, or local sequence logic needs adjustment.

Sequence

Typical on-site work sequence

The order can compress when the site already has approved connectivity and a known BMS integration path, but the review gates should stay explicit and local deployment should remain conditional.

Pre-site readiness check

Confirm that intake, model training, security review, and SI coordination are far enough along to justify field work instead of staying cloud-only.

01Review the optimization boundary, candidate points, metering path, and unresolved readiness blockers.
02Confirm network approvals, local gateway power and placement constraints, and maintenance access windows.
03Agree who will be on site or on call for BACnet, override, and alarm questions during commissioning.

Connectivity and local runtime placement

Establish the approved production path between the site-local runtime and ClimaMind services.

01Install or validate the local runtime location, power, cabling, and outbound-only communication path.
02Verify BMS read access, time sync, certificate or credential handling, and remote support boundaries.
03Confirm that customer BMS control paths remain inside the site boundary and are not replaced by the local runtime.

Point mapping and commissioning

Turn the approved point list into verified live telemetry and operator-visible behavior.

01Map intake point names to live BMS objects, units, schedules, and command semantics.
02Check freshness, ranges, stale-point handling, alarms, and trend availability for required signals.
03Run read-only monitoring long enough to catch naming mismatches, missing histories, or equipment-mode surprises.

Staged activation

Increase authority only when the site has reviewed the evidence for the next operating mode and approved local deployment involvement.

01Start with read-only monitoring when the site needs more confidence in telemetry and boundaries.
02Move to advisory mode when recommendations should be reviewed by operators or the SI before implementation.
03Enable supervised writes only after the approved write list, rate limits, fallback rules, and operator training are in place.

Handoff to acceptance

Close the on-site phase with evidence that the deployment is operating inside the agreed boundary.

01Review comfort feedback, equipment constraints, overrides, rejected writes, and issue logs from early operation.
02Document open items, excluded periods, and whether M&V reporting can start under the agreed method.
03Define the cadence for ongoing optimization review after the initial commissioning window.

Cautions

Cautions before and during site work

Most deployment delays and operating surprises come from skipped review gates, not from missing AI capability.

Do not enable writes before the boundary is approved

Supervised BMS writes should wait until the write list, numeric limits, change rates, fallback behavior, and operator override path are reviewed with the customer and SI.

Treat network and security approvals as lead-time items

Firewall rules, certificate issuance, service accounts, and remote support paths often take longer than hardware installation. Start them before the commissioning date is committed.

Plan around comfort-critical operating windows

Peak cooling seasons, occupancy events, maintenance outages, and tenant-sensitive periods may require read-only monitoring or advisory mode even when the technical path is ready.

Keep operator authority explicit

Operators should know how to pause optimization, where manual override takes precedence, and who to contact when telemetry or BMS behavior looks abnormal.

Use fallback events as operating signals

Connectivity loss, stale points, BMS rejections, and operator overrides are not just IT issues. They should trigger a deliberate review before authority is expanded again.

Do not assume every signal from intake will map cleanly live

Point renames, undocumented overrides, seasonal modes, and local sequence changes often appear only after read-only monitoring starts.

Evidence

What we verify before expanding authority

The site should be able to show that the live system matches the approved deployment package.

Required points read with credible units, timestamps, and equipment-state alignment.
Local runtime health, outbound connectivity, and agreed remote support boundaries.
BMS alarms, override behavior, and native fallback paths that remain understandable to operators.
A short commissioning record listing open items, excluded periods, and the current operating mode.

Deliverables

What the on-site phase should produce

The output is operational evidence, not a generic installation checklist.

Validated point map and telemetry quality notes for the approved boundary.
Current operating mode: read-only monitoring, advisory, or supervised writes.
Issue log covering overrides, rejected writes, stale telemetry, and unresolved BMS questions.
Handoff package for acceptance, M&V planning, and post-go-live optimization review.