An information assistant grounded in the operator's own data

Property operators field a steady volume of repeat questions: rental terms, availability, amenities, procedure for a specific request, who to contact for what. Each one is short, but the cumulative load is real, and the answers all live in the operator's own systems.

Generic chatbots fail this use case because they hallucinate property facts. Operators end up having to fact-check the bot, which is worse than having no bot at all.

Focus AI built a retrieval-grounded assistant that pulls from the operator's own structured property data and approved policy documents. The model is permitted to compose helpful answers but is constrained to only assert facts present in the retrieved context. When the question is outside scope, the assistant says so and offers a handoff to a human.

The data sources are the systems already used to manage the portfolio, so there is no parallel knowledge base to maintain. New properties and updated policies become available to the assistant the moment they land in the source of truth.

Repetitive enquiries are resolved without staff involvement, with the same quality as a written policy answer. The operator's team gets to spend its time on the conversations that benefit from a human, like physical viewings and complex tenant situations.

The answer logs are themselves useful: they reveal the questions prospects ask in their own words, which has fed back into how listings and policies are written.