Backup autonomy below minimum.

Trigger, action plan, and escalation criteria.

Related metric

backup_autonomy

Trigger

estimated_autonomy < agreed_minimum_autonomy (at present load)

Value drivers

Operational continuity

Preconditions

The minimum required autonomy agreed per building; the UPS battery-autonomy and generator fuel-level signals mapped and verified; the critical-load basis recorded so the estimate is computed against the loads that matter.

Human role

AI recomputes the estimate at load, attributes the shortfall to the leg responsible (UPS battery vs generator fuel), and routes the referral; operators confirm the estimate against the agreed minimum, notify the customer, and coordinate the vendor.

Action plan

(1) Confirm the estimate is reported against the building's agreed minimum, not a default, and that it is sustained rather than a single interval during a load peak. (2) Attribute the shortfall to the leg responsible — a UPS bank short on the bridge, or the generator leg short on fuel — so the referral is the right one (battery service vs refuelling). (3) Treat the number as a live estimate, not a guarantee — it moves with load and asset health; re-check at the present load before escalating. (4) Keedian surfaces which leg is short and by how much and hands it to the customer and the electrical / power vendor; refuelling, battery replacement, and load management stay with them.

Escalation

(1) Estimated autonomy below the agreed minimum while the primary source is already stressed or flapping → direct customer + vendor alert; the shortfall is about to be tested. (2) A below-minimum estimate on a load the customer has flagged as material → out-of-cycle notification. (3) Buildings trending toward the minimum month over month → carry into the MBR as a battery-replacement or fuel-management program candidate (magnitude expected — pending validation; offices have no named Keedian reference deployment).

Prevention

(1) Agree the minimum required autonomy per building at onboarding as the reference the estimate is reported against. (2) Verify the battery-autonomy and fuel-level signals map correctly so the estimate reflects the building, not a default curve. (3) Recompute the estimate as load changes and trend both legs so a slow erosion is visible before it crosses the minimum.