Back to the site
Energy sectorIn production

An AI assistant that calls leads back in 30 seconds, filters them, and hands operators only the ones ready to sign

How we helped a 15-operator call center in northern Italy stop losing hot leads and take 70% of the screening work off their operators.

AI
30s
From request to first call
5–8
Calls handled simultaneously
1,000+
Leads handled in production
−70%
Less screening work for operators
The problem

20 minutes late = a lost lead

The client is a call center in northern Italy with 15 operators, specialized in customer acquisition for major energy providers. Leads come in from Facebook campaigns: a person fills out a form with their details, and from that moment on every second counts.

The problem was simple but devastating: operators could not call every lead back fast enough. Between filling out the form and the first phone call, 20 or 30 minutes went by; sometimes hours. In that time the person loses interest, does not pick up, or has already talked to a competitor.

The result: money wasted on advertising, operators wasting time with people who were no longer interested, and a number of closed contracts that did not reflect the quality of the leads generated. Every day hundreds of requests came in; and a big chunk was lost before ever getting a phone call.

Before
20–60 minutes before the first phone call
One operator handles 1 call at a time
Leads lost to delays and missed calls
Operators spend 70% of their time screening
No organized data from the screening stage
After
30 seconds from form to phone call
The AI handles 5–8 calls simultaneously
Every lead gets called back, always
Operators only work on leads that are already ready
Every conversation produces organized data and a score
The solution

An AI assistant that calls, screens, and routes; without needing an operator

We dropped an AI voice assistant straight into the call center's acquisition process. Every time a person fills out the form on Facebook, the system calls them back automatically within 30 seconds; with a natural voice, in Italian, following a script built for the energy sector.

The assistant runs a real conversation: it checks interest, gathers the important information (type of supply, current usage, timing, contract details) and assigns a score. Only leads with a high enough score get passed to an operator, with all the data already collected and organized.

The detail that changes everything: the assistant handles 5 to 8 calls simultaneously. An operator does one at a time. That means at peak moments; when 10, 15, 20 requests come in within a few minutes; the system loses no leads. Zero queue, zero wait.

1
The lead fills out the form

A person sees the ad on Facebook and fills out the form with their name and number. The system is notified immediately.

Facebook forms
2
The AI assistant calls within 30 seconds

The call starts automatically. The assistant speaks with a natural voice in Italian, following a script built for the energy sector.

~30s from form to call
3
Screening and information gathering

The assistant checks interest and collects current provider, monthly spend, contract expiration, type of supply. Every data point is organized in real time.

Automatic scoring
4
Hand-off to the operator

Only leads with a high enough score get passed along. The operator receives the full record: data, transcript, recording, score. They start from the close, not from screening.

Ready leads only
Human operator
1
call at a time
one after another, with breaks
AI assistant
5–8
calls simultaneously
in parallel, never stopping
0 leads in queueEven during the busiest moments
24/7Always on, day and night
~30sAverage time to the first call
Under the hood

It is not an answering machine with a phone number

The voice assistant is a system built specifically to handle high volumes of outbound calls. It is not a 1990s answering machine, it is not a chat with a voice; it is a system designed specifically to screen leads over the phone in the energy sector.

Tech stackIn production
Voice
Assistant with a natural voice in Italian, dynamic script with objection handling and a conversation that adapts to the answers
Multiple calls
5–8 conversations simultaneously, each handled independently, with an automatic callback if there is no answer
Data capture
Real-time conversation analysis: it extracts and organizes provider, usage, expiration, type of supply; automatic score at the end of the call
Routing
Automatic connection to start calls in under 5 seconds. The client chooses how to connect it: Zapier, Make, or a direct connection
Back-office update
Automatic hand-off to the operator with all the data organized, full transcript, audio recording, and score
The results

Same conversions, lower costs, no lost leads

The share of successfully screened leads stayed at 30%; the same level as human operators. But the point is not to beat humans on a single conversation. The point is that every single lead gets called back immediately, no one falls through the cracks, and operators spend their time only on the people who are ready to sign.

On the opposite front; the customer care requests coming inbound on the switchboard; we built a twin system for the same client: an AI voice agent that answers inbound calls, handles routine requests, and warm-transfers to an operator when a real human is needed.

30%
Screening rate identical to human operators
0 leads lost
Every request called back within 30 seconds, always
−70%
Less screening work for operators
5–8×
Capacity compared to a single operator
"We used to lose leads because we could not call them back in time. Now every person gets a call within half a minute. Operators only work on leads that are already screened; it is a whole different world."
Operations manager · Energy-sector call center, 15 operators, northern Italy
Related cases

Other projects you might find interesting

Do your leads wait too long before getting called back?

We will show you how it works on your specific case. 15 minutes, no commitment.

Book a free demo
15 minutes · No commitment · You will see the system in action