AI assistant for phone customer care with warm transfer
How we helped a 15-operator call center in northern Italy take repetitive customer care off their operators' shoulders, and stop making anyone wait in a queue.
| 24/7 | |||
| Wait queue for callers | Always on, even after hours | Calls resolved without human intervention | Repetitive customer-care load on operators |
Seven calls out of ten do not need a human; but humans answer them anyway
The client is a call center in northern Italy with 15 operators, specialized in the energy sector. Inbound calls are a mix: a small share are hot leads who want to sign a contract, but most are customer care requests; where is my case, when is the technician coming, how much will I pay this month, I got a strange email.
The problem was not complexity: it was volume. Most of these requests are resolved in thirty seconds by looking at the back-office system. But every call tied up an operator for two or three minutes between opening systems, searching, answering, and wrapping up.
At peak hours the wait queue grew, hot leads ended up stuck behind trivial requests, and after hours the calls were lost entirely.
Every call ends up with a human operator
Wait times at peak: 3–10 minutes
Most of the time spent on repetitive requests
Hot leads queued behind routine customer care
After hours: calls lost
The AI agent answers every call immediately
0 seconds of waiting, always
The AI handles routine customer care on its own
Operators only talk to people with a real problem or who want to sign
After hours: answers and data keep coming in
A voice agent that does customer care; and when needed, passes the hot call to an operator
We put an AI voice assistant in to answer on the call center's main number. When a customer calls, the agent answers immediately with a natural voice in Italian, identifies the caller (from the number or a customer code), and understands what they need.
For routine requests it does everything on its own: it reads in real time from the back-office system the status of the case, the billing, the deadlines, the contract details. It answers, updates where authorized, and wraps up. A natural conversation; no voice menu "press 1 for…", no 1995 robocall.
The technical point that changes everything: when a human is needed, the agent does not say "we will call you back" and does not hang up. It transfers the call live. Same phone call, same customer on the line. The operator picks up and already has in front of them a summary of what was asked, what the AI answered, and where their input is needed. They do not start from scratch: they step in mid-conversation with the full context.
The customer calls the number
The call comes in normally on the call center's switchboard. No change of number, no app.
Existing switchboardThe AI agent answers and identifies
Natural voice in Italian. The agent recognizes the customer from the calling number or, if needed, asks for a code.
0s of waitingHandling or routing
For routine requests: lookup in the back-office system, answer, an update where applicable, wrap-up. For out-of-scope requests: transfer to an operator.
Back-office lookupWarm transfer to the operator
The call is passed live to an available operator, with the conversation summary already visible on their record. No callback, no waiting, no explaining from the top.
With summary3–10
minutes of waiting
at peak hours, after hours calls lost
0
seconds of waiting
always, day and night
It is not an IVR on steroids
The voice agent is a system built to handle complex inbound calls: not a disguised voice menu, not a chatbot with a microphone. It is designed to have a real conversation in energy-sector customer care.
| Voice | Assistant with a natural voice in Italian, a conversation that adapts to the answers and handles interruptions and off-script requests |
| Customer recognition | Automatic identification from the calling number; alternatively, a customer code collected in the first seconds of the call |
| Back-office integration | Real-time lookup: case status, billing, deadlines, contract details. Written updates where authorized |
| Warm SIP transfer | The call is routed live to an operator without dropping it; technology that most of the "AI voice agents" on the market do not have: they hang up the call and ask you to call back |
| Operator summary + transcript | When an operator picks up a transferred call, they have in front of them a summary of the conversation, the full transcript, and the customer data already open |
Operators freed up, customers served right away, 24/7 coverage
The system is in production and handles most inbound requests on its own. Operators spend their time only on the conversations that need a real human; people with an actual problem or who want to sign. No customer is left waiting, not even after hours.
On the opposite front; the new leads coming in from advertising campaigns; we built a twin system for the same client: an AI voice agent that calls every lead back in 30 seconds, qualifies them over the phone, and passes operators only the leads ready to sign.
| Calls handled end-to-end without a human | |
| Wait time for callers | |
| 24/7 | Continuous availability, even after hours |
| Repetitive customer-care load on operators |
"Before, 7 calls out of 10 were people who wanted to know the status of their case. Now the agent handles those, and my operators only talk to people with a real problem or who need to sign. It is a whole different job."
Other projects you might find interesting
Are your inbound calls clogging up your operators?
We will show you how it works on your specific case. 15 minutes, no commitment.
Book a free demo15 minutes · No commitment · You will see the system in action