Immobiliare.it 如何从单个对话智能体扩展到智能体网络

Founded in 2005, Immobiliare.it is Italy’s largest real estate platform and part of a group operating across Italy, Spain, Greece, Slovenia, Croatia, Serbia, Luxembourg and Cyprus. In Italy, the platform currently features more than 1.2 million listings, attracting over 50 million monthly visits across its website and app, and serves more than 31,000 real estate agents and developers.

Immobiliare.it saw an opportunity to use conversational AI to improve interactions across its marketplace. Buyers need their questions answered quickly, sellers need to reach genuinely interested parties, and real estate agents have to manage both sides at once. The team set out to build and orchestrate interconnected AI agents and subagents that could support natural conversations, while giving them the tools to understand agent behavior, evaluate performance, and continuously improve the experience across every interaction.

Building the first agent

Immobiliare.it began with a conversational AI agent to support each property listing. Natural voice quality was essential, especially in Italian and across the other languages in its southern European markets, where a native-sounding voice is often the difference between a caller staying on the line and hanging up. The team also needed to shape agent behavior, experiment quickly, and understand what actually happened in real conversations.

ElevenAgents delivered on everything. It offered a deep catalog of natural, native-sounding voices with authentic local accents across their core markets, alongside tools for agent development, conversation analysis, and workflow orchestration. 

“Building an agent is simple, but optimizing continuously is way more complex. ElevenLabs gives you all the tools possible for both phases,” said Paolo Sabatinelli, Chief Product Officer at Immobiliare.it. “We prefer to invest in creating a culture and know-how about building agents on the best platform. And with ElevenAgents, you don’t need to be an experienced developer.”

Every call answered, every lead qualified

The first agent went live on listings, responding day or night in the right language, answering questions, gathering lead information, and summarizing each interaction so real estate agents could see at a glance how hot a lead was. 

More than 70% of sellers opted in, and the early results validated the bet. Buyers were also more comfortable sharing personal details by voice than by text, which produced higher-quality leads.

Metric

Improvement in lead qualification rate

3.3x

Likelihood a user will share their phone number for follow-up

+74%

Users rate the feature positively

~80%

The lessons learned from the deployment of this initial agent laid the groundwork for extensibility across more agentic use-cases. 

From one agent to an agentic network

ElevenAgents became a platform Immobiliare.it could keep building on. The first inbound and lead-qualification agent became the starting point for an interconnected network, with each new agent taking on a different part of the marketplace, which grew to include:

• Inbound agent for private sellers
• Outbound agent to support listing quality
• Customer support across voice and chat.
• Conversational search to explore listings easily

“ElevenLabs is API driven, so every capability can be integrated externally with another platform, our database, or anything we want,” Sabatinelli said. “This translates into heavy usage and adoption for our team.”

As more agents became part of the product experience, how they work together became central. Subagents are smaller, specialized agents that each handle one narrow task, coordinated behind the scenes so the user still gets a single, seamless conversation. Instead of one heavy agent trying to do everything, Immobiliare.it runs a set of these subagents that stay invisible to the user while keeping behavior predictable and easy to improve. That architecture, combined with rich conversation transcripts, is what let the team iterate at a pace they never could before. 

Working within a single platform also meant the team learned together, sharing lessons from individual projects and building shared expertise. This reinforced the ambition to continue adding conversational AI experiences.

“A lot of the innovation is in the orchestration: deciding how agents should work together, how to keep their behavior predictable, and how to improve the experience as a whole,” said Sabatinelli. “For the user, the interaction should feel simple. Behind it, we are coordinating agents and subagents, with clear responsibilities and an ongoing process of evaluation and refinement.”

Building safely, always

As these systems grew, responsible deployment mattered as much as capability. Immobiliare.it pursued AIUC-1, an emerging certification standard for the security, safety, and reliability of conversational AI agents. Out-of-the-box, ElevenAgents offered built in guardrails and platform controls which covered the majority of the standard. This gave the team the confidence to keep building quickly, and doing so responsibly.

With ElevenLabs, Immobiliare.it is advancing from individual conversational AI assistants to a coordinated system of interconnected agents, while building the expertise to easily orchestrate, evaluate, and continuously improve agents across its marketplace.