Lead processing · Production

Immobiliare.it builds conversational real estate voice agent in days using ElevenLabs

The problem

Property marketplaces face a structural tension: sellers want to avoid spam calls while buyers are often frustrated by a lack of response to listing queries, leaving inquiries unanswered and leads unqualified.

First attempt

The team initially evaluated OpenAI APIs but found it lacked the flexibility and Italian language support required for the use case.

Workflow diagram · grounded in source
1
JSON listing data input
Trigger
Listing data in JSON format is taken as input to initiate a natural conversation with users.
▾ source quote
“create an assistant that could take listing data in JSON format, hold a natural conversation with users, and return a structured summary of the interaction”
2
Voice conversation in Italian
Ai action
The voice agent answers listing-specific questions on behalf of sellers, running 24/7 and responding in natural Italian.
▾ source quote
“answer listing-specific questions, gather lead information, and summarize the interaction. It runs 24/7, responds in natural Italian”
3
Lead qualification questioning
Ai action
The agent asks questions to determine how interested the user is, filtering for serious inquiries.
▾ source quote
“The agent asks questions to determine how interested the user is - filtering for serious inquiries”
4
Structured interaction summary
Output
The agent returns a structured summary of the interaction.
▾ source quote
“return a structured summary of the interaction”
5
Transcript-driven prompt tuning
Feedback loop
Transcripts of real conversations are used to tweak prompts in staging, test multiple variations, and push changes live to production.
▾ source quote
“Thanks to the transcripts, we could take a real conversation, tweak the prompt in staging, test multiple variations, and then push the change live to production. It felt like fine chiseling: every small adjustment brought the agent closer to behaving …”
Reported outcome

The ElevenLabs-powered voice agent achieved more than 70% seller opt-in, raised lead qualification from 19% to 63%, increased the share of users providing a phone number from 42% to 73%, and earned positive ratings from roughly 80% of users.

Reported metrics
Seller opt-in ratemore than 70%
Lead qualification raterisen from 19% to 63%
Users providing phone number for follow-upincreased from 42% to 73%
User positive ratingRoughly 80%
Show all 6 reported metrics
seller opt-in ratemore than 70%
lead qualification raterisen from 19% to 63%
users providing phone number for follow-upincreased from 42% to 73%
user positive ratingRoughly 80%
feedback loop timefrom weeks to hours
build timebuilt in just days
Reported stack
ElevenLabsn8nOpenAI APIsLovable
◆ Does this fit your context?

Compare to your context

Tell us your scale, team, and constraints. We'll show what changes at your size, what fails at your scale, and whether this case is a fit, needs adaptation, or won't scale to you. Free demo, no signup.

Compare to your context →
~30 seconds · free
Source
https://elevenlabs.io/blog/immobiliare
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

The ElevenLabs-powered voice agent achieved more than 70% seller opt-in, raised lead qualification from 19% to 63%, increased the share of users providing a phone number from 42% to 73%, and earned positive ratings fr…

What tools did this team use?

ElevenLabs, n8n, OpenAI APIs, Lovable.

What results were reported?

Seller opt-in rate: more than 70%; Lead qualification rate: risen from 19% to 63%; Users providing phone number for follow-up: increased from 42% to 73%; User positive rating: Roughly 80% (source-reported, not independently verified).

What failed first in this deployment?

The team initially evaluated OpenAI APIs but found it lacked the flexibility and Italian language support required for the use case.

How is this lead processing AI workflow structured?

JSON listing data input → Voice conversation in Italian → Lead qualification questioning → Structured interaction summary → Transcript-driven prompt tuning.

Related lead processing cases

WHAT TO DO WITH THIS

Now compare it to your context

This case is one data point. Whether its pattern fits you depends on your volumes, your stack, and your exception load — that comparison is the step no case study can do for you.

Browse the Lead processing cases →