Immobiliare.it Deploys Voice AI Agent, Triples Qualified Lead Rate
Italian portal Immobiliare.it has put a voice AI agent on every listing. It answers around the clock, collects contact details, and writes the agent a summary. The share of qualified leads (people genuinely ready to deal) rose 3.3x, and the chance of getting a phone number went up 74%. The takeaway: launch one narrow agent with a clear metric first, then add sub-agents.
AI-processed from ElevenLabs Blog; edited by Hamidun News
Imagine dozens of people calling about every apartment listing, while the agent has no time to figure out who is just curious and who is ready to buy. Italian real estate portal Immobiliare.it tackled this problem with a voice AI agent: it answers questions, collects contact details, and tells the agent how "hot" the client is.
The share of qualified leads, meaning potential clients brought to the point where "it's worth contacting this person," rose 3.3x. The probability that a user leaves a phone number for follow-up went up 74%.
A single agent later grew into a whole network. The story was published by ElevenLabs, whose ElevenAgents platform underpins the solution, so the figures should be read as vendor data, not an independent audit.
How it started
Immobiliare.it was founded in 2005 and is considered the largest real estate platform in Italy. It hosts more than 1.2 million listings in the country, over 50 million visits per month, and more than 31,000 agents and developers. The group also operates in Spain, Greece, Slovenia, Croatia, Serbia, Luxembourg, and Cyprus.
The challenge will be familiar to any marketplace. Buyers want a quick answer, sellers want genuinely interested people, and agents have to serve both sides at once. The company first assigned one agent to each listing. It answers day and night, switches to the required language, collects contact details, and after each conversation writes a short summary so the agent can see at once how hot the lead is.
The team also singled out voice quality. In southern European markets, sounding native with a local accent often decides whether a person stays on the line or hangs up. That is why they chose a platform with a large catalog of voices for the target countries.
What the numbers showed
More than 70% of sellers agreed to connect the agent to their listings. The company reports the following lead results:
- the share of qualified leads rose 3.3x;
- the probability that a user leaves a phone number for follow-up rose by 74%;
- about 80% of users rated the feature positively.
One notable observation: people share personal data more readily by voice than in a text chat, so the contacts they leave are of higher quality. The publication does not disclose the baseline against which "3.3x" was measured, nor the measurement period. The figures are therefore useful as a benchmark, but not as a guarantee for your product.
From one agent to a network
The first agent became the foundation for a whole network. It now includes:
- an agent for inbound requests from private sellers;
- an outbound agent that helps improve listing quality;
- a voice and chat support service;
- conversational search to make finding housing easier.
The key engineering idea here is sub-agents: small, specialized agents, each solving one narrow task. They work behind the scenes and are coordinated with one another, so the user hears a single interlocutor. Instead of one heavy agent that tries to do everything, you get a set of simple executors. Their behavior is easier to predict, and errors are easier to find and fix: if an agent answers rental terms inaccurately, you fix that particular agent without touching the others.
Product director Paolo Sabatinelli says the main work lies in orchestration: deciding how agents interact, how to keep behavior predictable, and how to improve the experience as a whole. Detailed conversation transcripts help here, since the team uses them to see quickly what actually happens during calls. Another benefit of a unified platform is that lessons from one project carry over to the next. The platform is also open via API, so the agents can be connected to a company's own database and other systems.
Safety and what to take from it
The company went through certification under AIUC-1, an emerging safety and reliability standard for conversational agents. According to the authors, the platform's built-in constraints and controls covered most of the standard's requirements. How precisely, is not specified.
The practical takeaways for product teams are these:
- Start with one narrow scenario where the result is easy to measure, such as lead qualification.
- Collect and read transcripts from day one. Without them, optimization becomes guesswork.
- Once you have several scenarios, split the logic into sub-agents with clear responsibilities rather than bloating a single agent.
- Build in quality evaluation and certification in advance, not after launch.
The main limitation of this story is that it is a vendor's client case: there is no independent data, no comparison with other solutions, and no information on cost. Still, the path of "one agent, then a network of narrow sub-agents with continuous evaluation" looks like a workable pattern for any marketplace.
- Nov 21, 2026VibeCBS collects project nicknames in advance: the agent finds 2.8 times more documents
- Nov 21, 2026FDE job postings grew 4.6x in a year: what a forward deployed engineer actually does at the client's site
- Nov 21, 2026Sakana AI Opens Applied Research Lab for Market and Economic Modeling
- Nov 21, 2026Meta Is Reportedly Preparing the $300 Muse Charm AI Tamagotchi With a Year of Verizon 5G
Key Questions Answered
How it started?
Immobiliare.it was founded in 2005 and is considered the largest real estate platform in Italy. It hosts more than 1.2 million listings in the country, over 50 million visits per month, and more than 31,000 agents and developers. The group also operates in Spain, Greece, Slovenia, Croatia, Serbia, Luxembourg, and Cyprus. The challenge will be familiar to any marketplace. Buyers want a quick answer, sellers want genuinely interested people, and agents have to serve both sides at once.
What the numbers showed?
More than 70% of sellers agreed to connect the agent to their listings. The company reports the following lead results: the share of qualified leads rose 3.3x; the probability that a user leaves a phone number for follow-up rose by 74%; about 80% of users rated the feature positively. One notable observation: people share personal data more readily by voice than in a text chat, so the contacts they leave are of higher quality.
What are the key details on From one agent to a network?
The first agent became the foundation for a whole network. It now includes: an agent for inbound requests from private sellers; an outbound agent that helps improve listing quality; a voice and chat support service; conversational search to make finding housing easier. The key engineering idea here is sub-agents: small, specialized agents, each solving one narrow task. They work behind the scenes and are coordinated with one another, so the user hears a single interlocutor.
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.