Bienestar Edificios develops real estate projects. Their problem wasn't getting prospects: Meta campaigns brought inquiries every day. The problem was what happened next.
The problem
Every lead landed on a WhatsApp handled by humans, between meetings, site visits and the rest of the operation. That meant replies that took hours (sometimes a whole day), conversations left halfway, and a constant mix of window shoppers with real buyers.
In real estate, response time is everything: the person asking about an apartment is asking three developers at once. Whoever answers first and best, wins. And the sales team burned hours talking to people who were never going to buy, while the hot leads went cold in the queue.
What I built
I set up an AI agent connected to the business's WhatsApp that does three things:
- Answers instantly, at any hour. The first reply arrives in seconds, whether it's Tuesday at 3 pm or Sunday at 11 pm.
- Filters and qualifies. The agent holds a natural conversation and works out what the person is looking for: unit type, whether they want to live there or invest, timelines, payment method. With that it builds a profile of the prospect.
- Hands off to a human at the right moment. When the lead is qualified, the agent directly coordinates a call with the sales team and passes along the full context of the conversation. The salesperson walks in knowing who they're talking to and what they want.
All fueled by the Meta lead-generation campaigns I also manage: from the moment someone sees the ad to the moment they reach the salesperson qualified, the entire operation connected.
What broke at first
This is the part almost nobody tells, so I will.
The first version of the agent tried to do too much. We had given it instructions to answer anything about the business, and that made it verbose and unpredictable: endless replies, details nobody asked for, and some creative interpretations of the promotions that had to be corrected.
The lesson was to narrow it down. The agent works better with one specific job (qualify and hand off) than trying to be employee of the month. Second lesson: the handoff to a human has to come earlier than you think. When the conversation gets serious, people want to talk to a person, and the agent has to know when to step aside instead of stretching the chat.
We adjusted the instructions, shortened the replies, defined clear limits on what it can and cannot say, and only then did the system really start to perform.
The result
I'm not going to publish the client's numbers, but the operational change is easy to describe: no lead ever goes unanswered. The filter between window shoppers and buyers is done by the machine, not by a tired human at 10 pm. And the sales team stopped chasing ghosts: they only talk to people who already showed real interest, with the context served.
What applies to any business
What I learned here I repeat in every AI implementation:
- The agent doesn't replace the team. It takes off their plate what they shouldn't be doing. A person still closes the sale.
- Narrow beats ambitious. An agent with one clear mission outperforms one that "does everything".
- The complete system is what works. The agent alone is useless without campaigns bringing people in and without a salesperson taking the call. The magic is in the connection, not the piece.
Got an overflowing WhatsApp and a team that can't keep up? Write to me: I've probably already built something close to what you need.