Paying a bill or have a question about your account? Make a paymentConsumer help
Paying a bill or have a question about your account? Make a paymentConsumer help
You can't hire for a storm that's 6 hours out. You can decide now who answers, in what order, and what sets each layer off.
Staff an outage spike in layers set up before storm season: automated restoration updates for the "when's my power back" calls, your own agents plus cross-trained back-office staff, then trained standby overflow agents switched on by a written trigger. Size each layer from your past storms' calls per half hour, and send safety and medical calls straight to people.
A dated surge, such as a renewal deadline, gives you months to hire and train. Staffing a dated surge covers that plan.
A storm gives you hours. Calls climb as the lights go out, peak while restoration times are unknown and fall off as crews report in. Then a second, slower wave of bill and hardship calls arrives weeks later. Our post Every storm hits a utility twice describes both waves and what customers need in each.
This page is about the math and the triggers.
Here's what the standard queueing math (Erlang C) says about one half-hour, with a target of 80% of calls answered within 30 seconds. The inputs are illustrations; use your own interval history.
| Half-hour scenario | Calls offered | Average handle time | Agents on the phones |
|---|---|---|---|
| A normal weekday peak | 300 | 5 minutes | 56 |
| Outage peak, no automation | 2,400 | 3 minutes | 247 |
| Outage peak, automation answers half | 1,200 | 3 minutes | 126 |
| Bill wave weeks later | 600 | 7 minutes | 149 |
TSI calculations, Erlang C, 80% answered in 30 seconds. Agents on the phones means people logged in and taking calls; scheduled headcount runs higher once breaks and absence are counted.
Three things jump out. An 8-fold spike in calls needs more than 4 times the agents, even with shorter calls. Automation that answers half the restoration calls cuts that need roughly in half. And the bill wave needs almost as many agents as the outage peak, for weeks instead of hours, because hardship and payment-arrangement calls run long.
Erlang C assumes steady arrivals and patient callers. Outage callers redial and hang up, so count unique callers when you can and treat the model as a floor.
Outage maps, texts and IVR messages with estimated restoration times answer the most common question without an agent. Update them as estimates change, or callers come back angrier.
Billing and back-office staff who've had outage-call training move to the phones when the trigger fires. Pause outbound work and non-urgent queues.
A partner team already trained on your scripts, tariff rules and transfer paths logs in on your signal. With trained overflow staff in place, TSI's standby agents can be taking calls within about 30 minutes of the go signal.
Downed wires, gas smells, medical-equipment customers and life-support accounts route to a trained person every time, with the address confirmed before transfer.
Decide in advance what switches each layer on. Tie triggers to things you can see early, for example: a severe weather watch for your territory, customers out above a set count, or calls in queue above a set number for 2 intervals running. Name who gives the go signal at night and on weekends.
Write the step-down rule too. Release layers in reverse as queue time falls, and keep enough agents for the callback and estimate-change calls that follow restoration.
The bill wave is predictable, so staff it like a dated surge. Usage spikes show up on the next bills, hardship requests rise, and winter disconnection rules change what agents can offer.
Hardship is already widespread. The U.S. Energy Information Administration counted 13.4 million residential electricity disconnections for nonpayment in 2024 (EIA, April 2026). Payment-arrangement and assistance calls need agents trained on your tariff as well as extra seats. Collections and customer care for utilities covers how both sides of that work fit together.
If you report service metrics, keep the overflow queue on the same definitions. Pennsylvania's electric distribution companies, for example, report the percent of calls answered within 30 seconds, busy-out rates and call abandonment rates to the Public Utility Commission each year (52 Pa. Code 54.153). After each event, review abandonment by interval and how safety-call transfers went. What to ask before outsourcing customer contact lists the vendor questions.
Results vary. They depend on each client's accounts, program and the rules that apply, and past results aren't a promise of future ones.
It depends on your peak calls per half hour and handle time. In the illustrative example above, a jump from 300 to 2,400 calls in a half hour took the need from 56 agents to 247, or 126 if automation handled half the calls. Run the same math on your own storm history.
Both. Routine restoration questions suit automated updates. Safety calls, medical-equipment customers and hardship conversations should reach a trained person every time.
Only as fast as its training allows. A team trained on your scripts in advance can start within minutes to hours of your signal. A team trained after the storm hits can't help with that storm.
Keep standby agents current on your scripts with refresher sessions and a few live shifts each season, and test the trigger and transfer paths before the first big weather of the year.
General information, current as of Nov 5, 2026. It isn't legal advice. Rules change and differ by state, so check the sources linked here and talk to your own counsel before you act on it.
The CX Surge Readiness Review is a 30 to 45 minute working session on your next surge, followed by a written readout with a ramp outline and the gaps we found.
