Dispatch Copilot
The Dispatch Copilot ranks your response units for the operator on every event card — closest and freest first — so the operator's two decisions on each event (do I dispatch? and who?) are assisted, not guessed. The Copilot never rolls a vehicle itself; the operator keeps the phone and the decision.
The Dispatch Copilot report answers one question: is the recommendation actually good, and does it show up fast enough to save time? It proves that with two metrics, both drawn from audit stamps the system already records — nothing new is captured to build this report.
Open it under Performance → Dispatch → Dispatch Copilot.
The recommender only runs on control rooms with Advanced Alarm Handling (AAH) enabled. On any other control room the report shows a note explaining that no recommendations are generated there. If you select a non-AAH control room you will see that notice instead of metrics.
Pre-staging: a "possible dispatch" stand-by
When a control room turns this on, the Copilot gives the nearest idle response unit a quiet heads-up the moment a dispatchable alarm lands — while the operator is still running the keyholder call — so that unit is already oriented if the operator does decide to dispatch.
It fires on the alarms an operator might dispatch on: a camera-verified intrusion, a panel burglar alarm, or a duress. A camera-"verified" event is treated exactly like a panel burglar alarm — "verified" only means the system is confident there is a real person or vehicle in frame (it could be a roaming guard, the owner, or a resident); it is not a confirmed break-in. So a pre-stage is a possible dispatch that is expected to stand down often — never an instruction to roll.
The responder gets:
- A push notification — "Possible dispatch — stand by" — with the site name and rough distance.
- An amber "Possible dispatch — stand by" card at the top of their CleverResponder on-shift screen, with a Standing by button to acknowledge.
The loop closes itself:
- If the operator resolves the event (e.g. the keyholder confirms it's a false alarm), the unit gets an automatic stand-down and the card clears.
- If the operator does dispatch, the pre-stage simply becomes the real dispatch.
Every pre-stage, acknowledgement and stand-down is recorded on the event timeline.
When a unit is pre-staged, the control room can also send the customer a one-tap secure link offering the same three options as the app — All is fine (password-gated), Phone me, Send help — giving the operator a high-confidence signal while they work the keyholder call. A tap is a signal the operator sees, never an automated stand-down or dispatch (and duress alarms are excluded). This is a separate opt-in — see Customer event heads-up. It is an automatic message, so it is not sent while the control room's Automatic messaging switch is off.
Pre-staging is off by default and enabled per control room (Advanced Alarm Handling only). Like the AAH switch itself, only a CleverCam super admin can turn it on — ask your CleverCam contact when your control room is ready for it. It is operator-assist: a heads-up is not a dispatch, and it never rolls a vehicle on its own.
The two metrics
1. Median time-to-dispatch (faster decisions)
The median time from an event being opened to a dispatch being placed for it. It is split into two eras so improvement is visible:
- Baseline — dispatches placed before the recommender was rolled out on the AAH control rooms.
- Copilot — dispatches placed on or after rollout.
The goal is for the Copilot median to be at least 30% lower than the baseline. The report shows both medians side by side with the percent change and a ✓ goal marker when the −30% target is met. Each figure carries its sample size (n=…) so a median built from one or two dispatches is never mistaken for a trend.
Event opened uses the moment the operator opened the event card when that is recorded, and falls back to the event's creation time (when it landed on the board) otherwise. Dispatch placed is the moment a unit was committed to the event — the operator's one-tap confirm in the dispatch picker. This measures the operator's decision latency, which is exactly what the Copilot is meant to shorten.
2. Top-rank hit-rate (best-unit-first)
Of the dispatches whose event had a recommendation, the percentage that went to the recommender's rank-1 unit (its top pick). The goal is ≥80% — a high hit-rate means the recommendation is genuinely good, not decoration. The value is coloured green at or above the goal and amber below it, with the raw hits / dispatches count beside it.
The scorecard cards
The top of the report shows four headline cards for the selected control room:
| Card | What it shows |
|---|---|
| Median time-to-dispatch | The Copilot-era median (event opened → dispatch placed), with the era's sample size. |
| Top-rank hit-rate | Percentage of recommended dispatches that went to the top pick, against the ≥80% goal. |
| Recommendations | How many recommendations the Copilot has produced, how many were acted on (the event went on to a dispatch), and when the last one was. |
| Dispatches | Total dispatches recorded, and how many of them had a recommendation to compare against. |
Below the cards, a Time-to-dispatch: baseline vs Copilot panel lays the two medians out as baseline → Copilot with the percent change, next to a short What this proves reminder of the two goals.
Dispatch decision audit (glass box)
The audit table lists every dispatch whose event had a recommendation — the evidence behind the hit-rate. Newest first:
| Column | What it shows |
|---|---|
| Event | The event ID (click to open the event details) and the time the dispatch was placed. |
| Time to dispatch | Event opened → dispatch placed for this event. |
| Recommended (rank 1) | The unit the Copilot put first, its estimated ETA, the ETA source, and how many units were ranked. |
| Dispatched | The unit the operator actually dispatched. On a miss it also shows that unit's rank in the recommendation (e.g. "rank 3"), so you can see how far from the top pick the choice landed. |
| Result | Top pick ✓ (green) when the dispatched unit matched rank 1, or Other unit (amber) when it didn't. |
| Era | copilot or baseline, per the rollout split above. |
The ETA source tells you how the proximity was estimated: road ETA (green) is a real driving-time estimate; straight-line, patrol-area estimate, and GPS distance are approximations. Right now the recommender runs in straight-line mode (road ETAs are switched off to control mapping costs), so proximity is approximate — good enough to rank, but not a precise drive time.
Reading the numbers honestly
The Copilot's dispatch and recommendation rails are new, so at the moment volumes are thin — only a handful of recommendations and very few confirmed dispatches across the AAH control rooms. The report is built to fill in as structured dispatch volume accrues; until then, a single-digit n beside a median or hit-rate is directional, not conclusive. The report deliberately shows every sample size so nothing is over-read. A low-volume banner appears while samples are small.
This report is read-only. It reflects what operators did; it never changes routing, arms autonomous dispatch, or touches the recommender's behaviour.