Real estate · NDA

AI qualification for inbound sales leads

On busy days this real estate agency was handling well over a hundred enquiries. We designed an AI layer to take the first pass at qualifying them, and showed it running on a demo.

Design work
AI agent Knowledge base First-pass qualification
100+
leads a day: the flow they were sorting by hand
100+
new leads a day at peak
3
reps taking the whole flow by hand
1 stage
we automated only the first pass
01 Context

Leads arrived faster than the team could answer

During discovery, a real estate agency came to us. At peak they were taking more than 100 new leads a day.

Every enquiry landed with one of three reps. They answered the same questions by hand, asked about budget, timing and what the buyer wanted, and only then got round to selling.

When the volume spiked, new leads just sat in the queue.

02 Why we started here

The money wasn't leaking out of sales

Sales had nothing to do with it. Neither did the CRM.

It leaked at first-pass qualification, the stage where a rep repeats the same thing to every new lead.

So we automated that one stage and left the rest of the sales floor alone.

03 What we designed

An AI layer between the ads and the sales team

We proposed an AI layer sitting between the ad channels and the sales team.

After the first message an assistant picked up the conversation. It answered the usual questions from a knowledge base, asked about budget and timing, and collected everything a rep normally has to dig out in the first few minutes.

By the end of the conversation the lead reached a rep as a finished record.

Knowledge base for the AI agent: a table of categories, questions and answers used to advise leads
The knowledge base the agent answers from. The development name and address are hidden.
04 What the client saw

One enquiry, start to finish

On the demo we walked through a single enquiry end to end.

The client sent the first message. The assistant handled the whole conversation: answered questions, checked details, and at the end handed the rep a finished lead record.

A person only stepped in once qualification was done.

Screen recording
Demo: the AI agent talking to an inbound lead. Personal details are blurred.
Screen recording
Collected lead records in the database.

Alongside the chat we built a dashboard and analytics for the whole flow: the funnel, the sources, and what a hot lead tends to look like.

Lead qualification dashboard: funnel, traffic sources, budgets and the profile of hot leads
Qualification dashboard: funnel, sources, profile of hot leads. The agency name is hidden.
Analytics: the AI's read on the lead flow by temperature, urgency and traffic source
Analytics: the AI's read on the flow — quality, urgency, sources.
05 Why we drew the line there

We didn't take the human out of the sale

We deliberately didn't try to replace the reps. Selling property still needs a person.

What we did take away was the first pass, the repeated questions, and filling in the CRM by hand.

06 How it ended

What actually happened

Honestly, about the result

After the demo the client decided to use this architecture on one of their next projects. It never got deployed: the deal fell through.

We're keeping this one as design work — an architecture we built and showed running.

Get started

We'll show you a solution built for your problem

Let's find the process that takes time from your team every day.