Base digital

Base digital

AI, Real Estate

AI, Real Estate

Turning property search
into one conversation

Turning property search
into one conversation

Designing and building Base Digital, an AI-first property search platform, from the first screen to the live product.

Designing and building Base Digital, an AI-first property search platform, from the first screen to the live product.

My role

Designer, Frontend

The Challenge

The Challenge

Looking for a property is slow and discouraging. Portals hand you a filter bar and expect you to translate what you want into fields: a price range you are not sure you can afford, a neighbourhood you have never visited, a number of bedrooms standing in for the life you want to live there. The buyer ends up doing the work the platform should be doing. And the information that would actually decide the purchase, what the area is like and whether it is growing, is nowhere on the page.

Looking for a property is slow and discouraging. Portals hand you a filter bar and expect you to translate what you want into fields: a price range you are not sure you can afford, a neighbourhood you have never visited, a number of bedrooms standing in for the life you want to live there. The buyer ends up doing the work the platform should be doing. And the information that would actually decide the purchase, what the area is like and whether it is growing, is nowhere on the page.

The solution

The solution

A platform where the search is a conversation. The person writes what they want in their own words and the model reads the database and comes back with real listings. Under the chat sits a layer of neighbourhood intelligence built from several data sources, so the answer carries the history of the area, the businesses around it, the construction nearby and the money moving through the region. Someone buying to live and someone buying to invest ask different questions and both get an answer in the same place.

A platform where the search is a conversation. The person writes what they want in their own words and the model reads the database and comes back with real listings. Under the chat sits a layer of neighbourhood intelligence built from several data sources, so the answer carries the history of the area, the businesses around it, the construction nearby and the money moving through the region. Someone buying to live and someone buying to invest ask different questions and both get an answer in the same place.

What i did

What i did

The decisions that shaped the product.

Made the results part of the conversation.

Made the results part of the conversation.

A chat that answers in prose is a worse portal. Every reply returns listing cards with photo, price, area and stage, so the person can act without leaving the thread.

Gave the model tools instead of asking it to improvise.

Gave the model tools instead of asking it to improvise.

The platform runs on GPT with property-specific tools wired to the database. The model does not invent inventory, it queries it. That is the difference between a demo and something a buyer can trust.

Put neighbourhood intelligence inside the answer.

Put neighbourhood intelligence inside the answer.

Population around the block, active businesses, capital in the radius, foreign investment, construction underway, visitors per year, the history of the area. Data that normally sits in reports nobody reads, delivered at the moment someone is deciding.

Turned campaigns into entry points.

Turned campaigns into entry points.

Minha Casa Minha Vida and financing for app drivers became ways into the product, with a simulation inside the tool. Someone who had never considered buying could arrive through a question about their own situation and leave with a number.

Designed it and shipped it.

Designed it and shipped it.

Screens, front-end and the interaction with the model were all mine. The pacing of a reply, the moment a card appears, how a wrong answer recovers, none of that survives a handoff document.

Outcomes

Outcomes

Conversations with the AI

Conversations with the AI

500+

Properties listed

Properties listed

500+

Neighbourhoods covered

Neighbourhoods covered

95+

Bootstrapped, zero marketing spend

Bootstrapped, zero marketing spend