Roots AI

Roots AI is an assistant built for agents at Grassroots Realty, a Canadian brokerage. Agents use it to search listings conversationally, work through marketing ideas, and get grounded answers on real estate situations they're handling. It's a professional tool rather than a consumer app: the audience is the brokerage's own agents, and it's available to them around the clock, which matters in a business where questions come up on a showing or late in an evening rather than during office hours.

An agent's working day generates a constant stream of small questions. What's comparable to this listing, how should I position a property that has been sitting, what's the right approach to a situation I haven't encountered before. Each one individually is minor, and collectively they eat the day. The usual answers are to search a listings portal with filters, dig through brokerage documentation, or ask a colleague who may not be free. Roots AI collapses all of that into one place an agent can just ask, in the language they'd use with a colleague, and get an answer that's specific to how the brokerage actually operates.

I was the front-end developer on Roots AI, building the app in Flutter on Firebase, with OpenAI's models powering the conversational search and advice. My work covered the chat interface, the listing search experience, and the surrounding app: authentication restricted to brokerage agents, session and conversation state, and the screens that turn a model response into something an agent can act on rather than just read.