Redfin builds a custom AI virtual assistant to answer questions about for-sale homes | Sendbird

Title

Redfin builds a custom AI virtual assistant to answer questions about for-sale homes | Sendbird

Content

【Lead sentence】Redfin built Ask Redfin, a custom AI virtual assistant deployed via Sendbird's in-app chat platform to help house hunters get fast answers about for-sale homes — resulting in 93% of users returning to the Redfin app within a week, with the final chat implementation completed in the equivalent of 3 person-months following a 1-week hackathon. 【Direct quote + attribution】Sebastian Lozano, Sr. Product Manager at Redfin, said: "We received outstanding support during our build-out, with engineers readily assisting us. The onboarding package was exceptionally beneficial and responsive. The team quickly adapted and addressed our evolving needs, ensuring our requirements were met." The case study frames the result: "Redfin has found that Ask Redfin users are some of its most engaged customers: 93% return to the Redfin app within a week and use it near-daily during that period." 【Context / why】Redfin's motivation was to transition house hunters from traditional text and phone calls to a modern, cross-platform, easy-to-deploy chat system. The company had first introduced an "Ask a Question" service that routed consumer questions to Redfin team members; the engagement data on that feature revealed the opportunity to automate and accelerate response time with a custom AI assistant for inbound lead engagement. 【Timeline of evolution】Redfin's path: (1) Initial "Ask a Question" service routed questions to human team members; (2) Hackathon week — basic chat functionalities set up; (3) Customization phase — equivalent of 3 person-months to create the final chat implementation; (4) Launch of "Ask Redfin" as an AI-powered virtual assistant. Sendbird provided highly responsive support throughout, from improving UI Kits to custom feature implementation. 【Technical architecture】Built on Sendbird's chat platform, which hosts 330M+ monthly active users across web and mobile (iOS, Android, React, React Native UI Kits), with a robust chat API and chat SDKs for production customization. Sendbird's underlying RAG model — built on Amazon Bedrock and Amazon OpenSearch, provided by AWS — delivers 96% accuracy. Sendbird operates 45+ AWS native services in 9 regions, leveraging AWS Auto Scaling, EC2, Elastic Load Balancing, Route53 geolocation, CloudFormation, ElastiCache, and Aurora for millisecond latency. Ask Redfin itself covers listings, local market conditions, and touring availability. 【Operational scale】Redfin is "the country's #1 real estate brokerage site" with brokerage, rentals, lending, title insurance, and renovation services. 93% of Ask Redfin users return to the Redfin app within a week and use it near-daily during that period, increasing engagement across the entire platform. The chat platform supports tens of thousands of users across web and mobile. 【Future direction】Per the case study: "Within a few days, Ask Redfin drove an increase in introductions to real estate agents." By engaging consumers on its iPhone app, Redfin provides a branded and convenient experience that matches modern customer service expectations. The article positions this as a model for digital leaders looking to link their most performing digital business channel (the app) with conversational AI to boost engagement, conversion, and loyalty.