Real estate

Every New Follower, a Real Conversation

A Los Angeles real estate team wanted each new Instagram follower greeted, qualified, and booked in, without it ever feeling automated. We built an AI agent that talks like a person over coffee and does exactly that, no buttons, no scripts.

AI AutomationConversational AIInstagram DMsLead Generation

Start a project

Instagram AI Lead Agent shown across a laptop and two browser windows

About the product

Real Estate AI Instagram Lead Generation

For a real estate team, a new Instagram follower is a warm lead for about a day. After that the moment passes. Following up by hand with every follower does not scale, and the usual chatbot answer, a menu of quick-reply buttons, feels like a form. People can tell, and they stop replying.

The brief was clear: an Instagram DM assistant that reads as indistinguishable from a real person. It opens the conversation the moment someone follows, keeps it casual, asks only the few questions that matter, handles the light objections that come up, and steers the right people toward a consultation call.

Behind the conversation sits a full lead pipeline. ManyChat catches the follow and carries the messages, Make.com orchestrates every step, the OpenAI API writes each reply, Calendly holds the real availability, and Follow Up Boss receives every qualified lead with their contact details, tags, and a summary of what was said.

Industry: Residential real estate, Los Angeles
Platform: Instagram Direct Messages
Timeline: About three weeks
Team: Dezy automation team

What we did

Conversation design
Prompt engineering
Automation build
CRM integration
Testing and launch

Stack

ManyChat (Instagram)Make.comOpenAI APICalendlyFollow Up Boss

Goals

What the build had to achieve

Reach every follower first

A new follower gets a casual opener in their DMs straight away, such as a quick thanks for the follow and whether they are local to Los Angeles, while their interest is still fresh.

Sound like a person, not a bot

Free-text replies only, in a friendly, relaxed tone, like talking to a friend over coffee. No quick-reply buttons, no rigid flows, nothing that gives the automation away.

Qualify without interrogating

Only the questions that matter, woven into the chat: the area they want, condo or house, whether they rent today, and a gentle sense of where their credit stands.

Turn the chat into a booked call

Guide interested leads to a consultation inside the conversation itself, and hand the team a complete, tagged lead record instead of a DM thread to dig through.

Challenges

The parts that were genuinely hard

Free text without losing the thread

The problem

Button-based chatbots are predictable because they control every answer. Letting people type anything they like means replies arrive out of order, off topic, or answering two questions at once, and a scripted flow breaks the moment that happens.

What we did

Each reply is written by the OpenAI API with the whole conversation so far in view and a clear picture of which qualifying details are still missing, so the agent answers what was actually said and then steers back naturally.

The outcome

Leads can talk the way they normally would, and the agent still gathers everything the team needs to know.

Credit is a sensitive subject

The problem

Whether a buyer can get financing matters early, but asking about credit too bluntly ends conversations.

What we did

The agent raises it softly and in context, with questions like how they think their credit looks or whether there is anything like collections or repos to be aware of, and treats hesitation as something to talk through rather than a dead end.

The outcome

The team learns where a buyer stands on financing before the call, without the chat ever feeling like an application form.

Booking without a link

The problem

Dropping a scheduling link into a DM is the clearest sign of a bot, and it hands the lead a task instead of a conversation.

What we did

The agent offers a few specific time windows within the next two days, as a person would, while Calendly stays behind the scenes holding the real availability. Once a time is agreed it asks for a phone number first and an email second.

The outcome

Booking feels guided and personal, and the call lands in the team's calendar with the details needed to reach the lead.

Built for volume from day one

The problem

The account was expected to receive somewhere between 1,000 and 2,500 DMs a month, far beyond what a person can answer promptly by hand.

What we did

The whole pipeline runs as a Make.com scenario triggered by ManyChat, so every new follower and every reply is handled the same way, at any hour, with no one watching the inbox.

The outcome

Every follower gets the same prompt, friendly attention, whether it is the first DM of the day or the hundredth.

Approach

How the work ran

01
Discovery and conversation design
We mapped the conversation the team already has with good leads: how they open, which few questions actually decide fit, the objections that come up, and the point at which to suggest a call. That became the brief for the agent's tone and the order it gathers details in.
02
Prompting and tone
The agent's instructions were written and refined against real-style conversations until replies read as relaxed and human: short messages, natural follow-up questions, no sales pitch, and no sign of a script.
03
Automation and integrations
ManyChat was set to open a conversation on every new follow and pass each message to a Make.com scenario through a custom webhook. Make.com calls the OpenAI API for the reply, works with Calendly for real availability, and pushes the finished lead into Follow Up Boss.
04
Testing and launch
We ran the full journey end to end, from a fresh follow through qualification and booking to the CRM record, including off-script answers and objections, before switching it on for the live account.

Features

What shipped

Instant opener on every follow

A casual first message goes out to each new follower automatically, so the conversation starts while they are still thinking about the account.

Human-sounding free-text replies

Every reply is written by the OpenAI API from the full conversation, in a friendly, casual voice, with no quick-reply buttons anywhere.

Light-touch qualification

Gathers preferred area, condo or house, current renting status, and a soft read on credit, a little at a time as the chat goes on.

Natural objection handling

Picks up on hesitation and follows up with the kind of probing questions a good agent would ask, instead of pushing the next scripted step.

Guided booking in the DM

Offers concrete time windows within the next two days, with Calendly working in the background, then collects phone first and email second.

Every lead in Follow Up Boss

Name, phone, email, a summary of the conversation, and tags including the lead's Instagram handle are pushed into the CRM automatically.

Want Your DMs Answered Like This?

We build AI agents that hold real conversations on Instagram, WhatsApp, and your website, then hand your team qualified leads with the context attached. Tell us where your leads come from and we will scope the build.

Start a project Book a call

More case studies