Trouble call management that pings the meters before the truck rolls, so crews go only where the power is really off
Crews go only where the power is really off, each outage confirmed in minutes.
See one case, screen by screen ↓When a fuse blows on a lateral, the outage desk gets dozens of no-power calls in minutes, by phone, app and text. Each one has to be sorted by hand: a network outage on a fuse, transformer or feeder, a single home with a failed service cable, an outage the desk already knows about, or a fault inside the customer's own home — a tripped main switch, an RCD or a blown fuse in the consumer unit. The dispatcher decides from the caller's words and a pin on a map, and many single-premise jobs end with the crew finding power at the meter.
The volume hides the calls that matter. “A bang and a flash at the pole” sits in the same list as forty “no power” calls unless someone reads every one, and a life-support customer looks like any other caller. Last gasp alone does not settle it: in a big outage, messages collide on the meter network and some never arrive. And a caller with a tripped RCD inside a network outage stays dark when the fuse is replaced, calls again two hours later, and a second truck goes back to the same street.
Estimated split for the demo utility's Haverford and Kingsmead districts, in crew days a month (8-hour days), by hand and with the solution. The first row uses the demo's counts — 338 jobs a month where the crew found power on at the meter before pings, 14 with them — at about 1.5 hours a job; the other rows are our estimate.
Six agents read, group, ping and trace every no-power call; the dispatcher confirms the outage and sends the crew.
Tuesday 13 October 2026, evening peak, Haverford district of Lumina Utilities. At 19:41:52 fuse F-2217 on pole CR07-118, on Halloway Lane by Larchmere Primary, blows and lateral CR07-L3 drops seven transformers, T-4411 to T-4417. Here is what happens next, screen by screen, in the working solution.
Rhiannon has the evening shift on the outage desk. The first call lands at 19:42:10 and by 19:46:05 there are 40 from Larchmere: 28 phone, 8 app and 4 text. Every call is a ring on the outage map, drawn on feeder CR-07 from Corwen Road substation, its laterals, fuse F-2217 and the nine transformers around the calls. At 19:44:02 the Call grouper opens a cluster on lateral CR07-L3: 3 calls inside 400 m and 10 minutes, and no incident open there in the outage system.
“Larchmere: 40 no-power calls in 4 minutes” · 58 no-power calls tonight since 18:30
Tomasz Brennock reports from the app at 9 Halloway Lane: “Loud bang and a flash at the pole by the school, then everything went off.” The Call reader matches the premise and the meter, finds the hazard words “bang” and “flash”, and flags the call to the dispatcher at once. Whatever his meter says later, this call is never answered with a self-check message.
“The words “bang” and “flash” are on the hazard list. A crew checks pole CR07-118 before the fuse is replaced. A self-check is never sent for a hazard call.”
Before anyone is sent, Rhiannon presses Ping 214 meters: the callers and their neighbours on 9 transformers. The Meter pinger sends on-demand pings through the head-end, with 3 retries per meter at most, and the wave runs out from the calls across the streets of Larchmere.
“Ping the meters to see which homes are really off. Until then every call waits on the cluster, except the hazard.”
At 19:46:31 the answers are in: of 214 homes, 171 are off and 43 are on; of 40 callers, 38 are off and 2 are on. The dark homes trace a single lateral back to fuse F-2217, and the Outage tracer names it the likely device at 92%. From the first call at 19:42:10 to the outage confirmed with its fuse at 19:46:40 took 4 minutes 30 seconds.
“Likely device: fuse F-2217 · 92%” · “First call 19:42:10 → outage confirmed with its fuse 19:46:40 · 4 min 30 s”
The run on the cluster shows each agent's step and its time. All 171 silent meters sent last gasp at 19:41:52–54; the 43 that answer read 236–242 V. Every dark meter is behind F-2217 and no answering meter is. The Reply writer drafts 2 self-check messages, and the Dispatch drafter proposes truck 14 — Joel Varga, 6 minutes away, 100 A fuse kit on board — and waits for the dispatcher.
“Dispatch drafter · truck 14 proposed … Waiting for the dispatcher”
Elsie called at 19:42:35: “Power’s gone. The whole street is dark, I can see it from the window.” Her meter answers at 239.6 V. Her home is on transformer T-4421, fed from lateral CR07-L4, which never lost supply. The load fell to zero at 19:41:55, three seconds after the voltage dip from the fuse on the next lateral, and the premise history shows two earlier RCD resets here, in March and August 2025. Most likely the RCD tripped again.
Without the ping, she would stay dark after the fuse is replaced and call again.
INC-26-30871 groups 40 calls into one incident: 171 customers off, 3 on the priority register. The Outage tracer lists its reasons — the last gasps from T-4411 to T-4417, the seven transformers sharing only lateral CR07-L3, the 43 meters on CR07-L4 answering, the bang at pole CR07-118, and no incident open on CR07-L3 before the calls. It also shows what it ruled out: transformer T-4411 alone, at 3%, because six other transformers are dark too.
Below 80% confidence the Outage tracer names the area, not a device. The dispatcher confirms the device.
Rhiannon sends the self-check to the two homes with power, confirms the incident and approves truck 14, with a restoration estimate of 21:30. 169 customers get the restoration update by text and app; Amara Quist in the contact centre phones 2 of the 3 priority customers; the third asked for text only. At 19:52 Elsie Hale replies: “FIXED, the trip switch was down.” Marcus Webb at 7 Orchard Close replies FIXED at 19:55.
“Elsie Hale, 14 Linden Row: “FIXED, the trip switch was down” · 19:52”
The message tells the customer their smart meter shows power reaching the home, asks them to check the fuse box — main switch and trip switch first, and not to touch it if it is wet, damaged or hot — and to reply FIXED or STILL OFF. Twenty minutes later the meter is pinged again: load back closes the call with no visit; still off with voltage present brings electrician advice and a call-back; a meter that has gone silent gets a crew job drafted.
Sent only when the meter answers with voltage, the call has no hazard words and the customer is not on life support.
Joel is on the way to pole CR07-118 to inspect it and replace F-2217, with the hazard note “Bang reported · check before re-fusing”. Every job on the crew board carries its proof: a ping with 171 off and last gasp, hazard words with the meter on, a meter off alone, a silent meter with its neighbours on. Seven homes whose meter answered with voltage got a self-check message instead of a crew.
“The meter answered with voltage, so the home has power: a self-check message went instead of a crew.”
The same blown fuse, seen by the people who take the calls, decide, drive out and switch the lights back on — what their night looked like, and what it looks like now.
Six agents read, group, ping and trace every no-power call; the dispatcher confirms the outage and sends the crew.
Reads every phone summary, app report and text; matches the premise and its meter from the caller ID, account or registered mobile, notes the symptoms in plain words and finds any hazard words.
Places each call on its transformer, lateral and feeder; opens a cluster when 3 calls share a device within 10 minutes and 400 m, and attaches calls inside an open incident or planned work so the customer is told its estimate.
Pings the callers' meters and every neighbour on the same transformers; records the answer, voltage, load and last gasp. A meter with no answer and no last gasp is silent, and its neighbours are pinged.
Walks the network model upstream from the dark meters to the deepest device they all share and no answering meter shares, weighs the callers' words and the last-gasp times, and gives the likely device with its confidence and evidence.
Drafts the self-check message for a home whose meter answers at 207 V or more, the restoration update with the estimate the dispatcher set, and call-back notes for priority customers; books the re-ping 20 minutes after a self-check.
Drafts the crew job for a confirmed incident or a single home with no answer — device, pole, hazard notes and evidence — and proposes the nearest free crew with the right skill and kit.
Confirms every incident and crew. The agents propose; a named person decides.
Ask about any call, home or incident in plain words, or say what to change. Every answer points to the ping, the meter events or the network model behind it.
Why are 14 Linden Row and 7 Orchard Close not part of the outage?
Because their meters answer. At 19:46 the ping found 239.6 V at 14 Linden Row and 241.2 V at 7 Orchard Close. Both homes are on transformer T-4421, fed from lateral CR07-L4, which never lost supply; the homes across the road are on CR07-L3 behind fuse F-2217. At Linden Row the load fell to zero three seconds after the voltage dip, so the RCD most likely tripped; it has happened there twice before.
Which calls tonight could still waste a truck roll?
Three to watch: 4 Mill Pond Lane — the meter is silent with no last gasp and its 6 neighbours are on; truck 3 is going, but it may be the meter's radio, not the supply. 3 Cooper's Yard — meter on at 238.8 V, self-check sent 19:38, no reply yet; re-ping at 19:58. 21 Tanners Hill — still off with voltage present after the self-check: an electrician is advised, not a crew.
Ping the cluster automatically at the third call
Done. New rule, version 8: when 3 calls cluster on a shared transformer or lateral within 10 minutes, the Meter pinger pings every meter on it without waiting. You still confirm the incident and approve the crew. Hazard calls still go to a crew first. Tonight that would have pinged Larchmere at 19:44:02, two minutes sooner.
13 screens from the working solution, on its sample data. Pick one to see it large.
How calls are grouped, when meters are pinged, who gets a self-check and who decides are settings. Every change is saved as a new version.
| Setting | Default | Choose from |
|---|---|---|
| Calls that make a cluster | 3 calls | Calls on a shared transformer or lateral inside the window and radius — set with − and + |
| Window | 10 min | Time between the first and the last call — set with − and + |
| Radius | 400 m | Distance between calls in a cluster — set with − and + |
| Ping a cluster | Off | Off: the dispatcher presses Ping · On: the Meter pinger pings at the third call |
| Ping every single-home call before a job is drafted | On | On · Off — a crew is drafted only if the meter does not answer |
| Voltage that counts as power at the home | 207 V | 230 V minus 10% — set with − and + |
| Re-ping after | 20 min | Minutes after the self-check message — set with − and + |
| Never to a hazard call | Always on | Fixed rule, cannot be switched off; the hazard word list is versioned (version 7) |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample outage-desk data · “estimated” = our arithmetic, inputs shown ($1.1M = 3,700 truck rolls a year to homes whose meter shows power × $300 a roll; the crew-days split is our estimate on the demo's counts) · “typical” = published figures (S&C Electric, The Real Cost of a Truck Roll — $250–$500 a dispatch; E Source, 15 utilities — $9.54 a call; US EIA, Electric Power Annual Table 11.1 — 366 minutes in 2023 including major events). Practice described as context: the smartgrid.gov OMS ping and on-demand poll use cases. Lumina Utilities and the people, streets, substations, crews and records named on this page are fictional — characters and sample data in the working solution.
Trouble call management is how an outage desk handles no-power calls: deciding, call by call, whether it is a network outage, a single home, an outage already known or a fault inside the customer's home, and sending a crew only where one is needed. Trouble Call Management does it with six agents that read, group, ping and trace every call, while the dispatcher confirms every incident and approves every crew.
When calls arrive, the Meter pinger sends an on-demand ping to the callers' smart meters and their neighbours on the same transformers, and reads the answer, voltage, load and last gasp. A meter that sent last gasp and does not answer is off; one that answers with voltage has power at the home; one with no answer and no last gasp is silent, and its neighbours are pinged. In the demo, the ping of 214 meters took 11 seconds: 171 off and 43 on.
Every single-home call is pinged before a job is drafted, and a crew is drafted only if the meter does not answer. A home whose meter answers at 207 V or more, with no hazard words and not on life support, gets a self-check message instead. In the demo's last 30 days, 312 truck rolls were held and jobs where the crew found power on at the meter fell from 338 a month to 14.
No. A call with any word on the hazard list — wires down, sparking, bang, flash, smoke, burning smell, tree on the line and more — goes to a crew. It is a fixed rule that cannot be switched off; in the demo's last 30 days, 41 hazard calls went to a crew and none to a self-check.
The Outage tracer walks the network model upstream from the dark meters to the deepest device they all share and no answering meter shares, and weighs the callers' words and the last-gasp times. It gives the likely device with its confidence and evidence, and below 80% it names the area, not a device.
They are never handled by a self-check message. A person in the contact centre phones them, and the Reply writer drafts a call-back note for each. On the demo incident, 2 of the 3 priority customers were phoned — the third asked for text only — while 169 customers got the restoration update by text and app.
Yes. The agents propose and draft; they never dispatch, never close or merge an outage-system incident and never connect or disconnect a meter. The dispatcher confirms every incident, approves every crew and sets the restoration estimate, and the supervisor approves a second crew or an estimate change over 2 hours.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your call scripts, hazard words, self-check rules and a sample of past no-power calls, meter events and incidents — and it goes live once every quality gate has passed. We will walk you through it on your own calls first.
We'll run Trouble Call Management on a sample of your own no-power calls, meter events and outage incidents.