← All case studies

01Energy · Power and gas·Inbound · Email

An energy retailer’s customer service, handed to an AI agent.

A power and gas retailer with a network of local agencies put an AI voice agent on the front line of customer care: it identifies the customer, reads invoices and payments from the billing system, sends documents and opens tickets, around the clock.

70%+of calls closed by the agent, with no transfer to an operator
5,500+inbound calls handledMay – Nov 2026
24%of calls arrive out of hours: all answered
~1,000tickets opened and routed automaticallyJul – Nov 2026

The challenge

Bills, payments and disconnections: lots of calls, always the same ones.

In the deregulated energy market, customer care is under constant pressure: invoice copies, payment details, forms, reminders and instalment plans. Every spike in wholesale prices turns into a wave of “why did my bill go up” calls.

The client ran support through an external call center and a network of local agencies. Simple requests kept operators busy, out-of-hours calls went unanswered, and tickets had to reach the right agency rather than a generic inbox.

The solution

A first-level agent wired into the billing system.

We designed a voice agent that answers on a dedicated number and works directly on the client’s systems. It identifies the caller by customer code (never by name or phone, as a security choice), reads supplies, invoices and payments from the billing platform, and checks for tickets already open.

It can email the invoice PDF to the registered address, send the right forms (direct debit, change of ownership, change of address, power upgrade, meter relocation), explain bill increases and open typed tickets, routed to the responsible agency or to central customer care. It remembers recent interactions from the same number, so customers don’t have to start over.

When a person is needed (urgent cases such as a suspended supply, angry customers, sales enquiries) the agent transfers the call to the second-level call center, during office hours only. A second instance works the payment-receipts mailbox: it verifies the customer and their account status and files the case.

How it works

One call, step by step.

  1. 01

    Instant answer

    The agent picks up immediately, day and night, and discloses that it is an AI assistant.

  2. 02

    Identification

    Customer code, billing-system lookup, supplies, invoices, payments and open tickets.

  3. 03

    Resolution

    Invoice or forms by email, bill explanation, payment details, instalment requests.

  4. 04

    Ticket or transfer

    If action is needed: a typed ticket to the right agency. If urgent: a handoff to an operator.

  5. 05

    Traceability

    Every ticket is logged and handed to the CRM with a link to the call recording and transcript.

What customers ask

Payments and bills come first.

Call reasons from a sample of 200 classified conversations.

  • Payments, arrears and instalments22%
  • Bills: copies and disputes20%
  • Suspensions, disconnections, reactivations18%
  • New supplies, transfers and switches14%
  • Other: social bonus, reports, operator13%
  • Contract changes8%
  • App access and customer code5%

Results

What actually changed.

An automated front line

More than 7 calls in 10 end with the agent: information given, documents sent or tickets opened, without tying up an operator.

Always reachable

A quarter of calls come in the evening, early morning or at weekends: the agent handles them even when nobody is on shift.

Peaks absorbed

In October traffic doubled for three days (up to 92 calls a day): the agent held the same quality.

A back office that moves itself

About 1,000 tickets opened and routed and 500+ invoices emailed by the agent in four months.

Traceable tickets

Tickets handed to the CRM three times a day, each linked to its call, with operational alerts on every anomaly.

Privacy by design

Full DPIA with the client’s DPO, limited recording retention and AI disclosure at the start of every call.

Running a similar customer service? Let’s talk.

Book a call→
Next caseAn AI front desk for a network of credit agencies.→