Autotorino Uses a Voice AI Agent to Automatically Handle More Than Twice as Many Calls
Italian car dealer Autotorino receives about 100,000 calls a month and can't answer all of them. Previously, callers left a voice message. Now an AI agent built on ElevenAgents picks up the phone, answers around the clock and creates a lead. The volume of automatically handled calls has more than doubled. So when you launch an agent like this, track leads, abandoned calls and sentiment right away.
AI-processed from ElevenLabs Blog; edited by Hamidun News
Picture a car dealership that gets called about 100,000 times a month, yet not every call gets answered. At night, at lunchtime or during rush hour, the caller hears an answering machine and most likely just goes to a competitor. That is the problem Italian dealer group Autotorino faced.
It replaced the "leave a message" recording with a voice AI agent built on ElevenLabs' ElevenAgents platform. The agent answers the call, works out what the person needs and enters a request (a lead) into the CRM. According to the company, the volume of automatically handled calls has more than doubled compared with the previous system.
A caveat up front: this is a case study published by the vendor itself, so it contains few figures, and the customer's assessments should be read with marketing in mind.
Before: voicemail and a chain of services
Autotorino has been operating since the 1960s. The group was the first in Italy to enter the ICDP top 50 European dealer groups, the first in the country to import Japanese SUVs, and the first to show competing brands in one showroom. Today its dealer network handles roughly 100,000 calls a month, and the team physically cannot answer them all.
The old setup was a compromise. The caller heard a recorded message and left a voice request, and then a chain did the rest: Zapier (a service for connecting applications), a separate speech transcription service and a separate language model turned the recording into a lead in Salesforce. This approach covered only a fraction of incoming calls, and outside business hours many inquiries went unanswered.
Now the agent answers at any time of day. The goal is unchanged: qualify every request. But the caller speaks with the agent instead of dictating a message into a void. According to Michele Spalla, head of projects and innovation, round-the-clock operation gives customers continuous service.
How the platform was chosen and what is measured
Autotorino compared several providers on three criteria: response latency (the pause before the agent's reply), how natural the conversation feels, and how well the system understands what is said without moving data between different tools. That last point is worth remembering: in the old setup, data bounced between Zapier, transcription and the language model. ElevenAgents was chosen because recognition, dialogue and analysis sit on one platform.
The company judges the rollout's success by three metrics:
- the number of leads created;
- the share of callers who hang up after the agent's reply and before a lead is created;
- sentiment analysis of conversations, built into ElevenAgents.
According to the company, all three metrics improved, and the share of callers who hung up stayed below the set threshold. The priority is simple: the number of leads should match the number of calls, so that no inquiry is lost. The material gives no absolute lead figures and does not state the threshold itself.
What the analytics showed and where Eleven v4 fits in
When all calls are recorded and reviewed, it becomes clear what customers actually ask for. About 70% of inquiries concern after-sales service: booking scheduled maintenance and arranging for a car to be dropped off at the workshop. In addition, people often ask what stage their repair is at. This prompted the company to start work on a way to report that information automatically.
Autotorino also tested the new speech synthesis model Eleven v4. With the same voice, settings and system prompt, the agent, in the company's assessment, sounded more natural and less like a machine. Three changes were noted: a less artificial voice, a steadier and more engaging tone that adapts to the stage of the conversation, and smoother dialogue. The agent is deliberately kept direct and deterministic so that customers get clear answers.
Next, the company plans to replace the classic IVR voice menu ("press 1, press 2") with ElevenLabs voice agents at its first branch in November. In parallel, it is building an AI culture: it trains "ambassadors" in each department and holds Innovation Fridays twice a month, with topics ranging from 3D printing to the European AI Act.
What to take away
For anyone building voice agents, the case offers three practical lessons. First, choose a platform not only for voice quality but also for how many handoffs between services you will have to maintain. Second, define your metrics in advance: leads, abandoned calls, sentiment. Without the abandonment share, it is easy to end up with an agent that "talks" but frustrates customers. Third, recorded and reviewed calls can themselves become a source of product ideas, as happened with the repair-status feature.
The limitations are also clear. There is no data on the share of calls the agent closes without a human, on the cost per conversation, or on how well Italian speech is recognized in noisy conditions. For now, this is more a careful report on a launch than proof that operators can be fully replaced.
- Nov 21, 2026Perplexity Releases Decider v1.3: 70 Points on the Decision Index and $68 per Million Tasks
- Nov 21, 2026Alibaba Open-Sources AI Code Reviewer That Tens of Thousands of Developers Used for Two Years
- Nov 21, 2026Gemini 4 Hydrogen: Google's Junior Model Beats the Senior Argon and Runs 2.5x Faster Than Flash
- Nov 21, 2026Claude Code Taught to Hunt Vulnerabilities: 83 Skills Built from 681 HackerOne Reports
Key Questions Answered
What are the key details on Before: voicemail and a chain of services?
Autotorino has been operating since the 1960s. The group was the first in Italy to enter the ICDP top 50 European dealer groups, the first in the country to import Japanese SUVs, and the first to show competing brands in one showroom. Today its dealer network handles roughly 100,000 calls a month, and the team physically cannot answer them all. The old setup was a compromise.
How the platform was chosen and what is measured?
Autotorino compared several providers on three criteria: response latency (the pause before the agent's reply), how natural the conversation feels, and how well the system understands what is said without moving data between different tools. That last point is worth remembering: in the old setup, data bounced between Zapier, transcription and the language model. ElevenAgents was chosen because recognition, dialogue and analysis sit on one platform.
What the analytics showed and where Eleven v4 fits in?
When all calls are recorded and reviewed, it becomes clear what customers actually ask for. About 70% of inquiries concern after-sales service: booking scheduled maintenance and arranging for a car to be dropped off at the workshop. In addition, people often ask what stage their repair is at. This prompted the company to start work on a way to report that information automatically. Autotorino also tested the new speech synthesis model Eleven v4.
Ship AI Faster with Founders, Engineers & Weekly Architecture Reviews
Inside Vibe Club (hamidun.com/club) — weekly live masterminds with Zhemal Khamidun, Tech Radar, full LMS access, and 370+ production Claude Code skills.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.