Redfin
genAI assistant on the product page that answers then routes to an agent
Ask Redfin, a genAI assistant launched nationwide in May 2024, answers questions about a listing: 93% of users return to the app within a week and nearly 1 in 10 request to be connected with an agent.
Objective
Answer a buyer's questions about a listing, the local market, and tours instantly, to keep them in the app and convert them into a connection with a Redfin agent.
The deployment
Ask Redfin is a genAI virtual assistant launched in beta on March 7, 2024, then opened nationwide to iPhone app users in the United States on May 2, 2024. It appears on a listing's detail page and answers questions about the property's features, local market conditions, neighboring amenities, and tour availability. For requests that go beyond the listing details, it hands off to Redfin's support team to connect the user with a licensed agent. The tool was designed with a set of instructions to answer in line with fair housing rules and to refuse questions that touch on housing discrimination. After launch, Redfin observes an increase in connections with agents.
Results Proof C
Engagement figures reported by Redfin on its own official channel (T1) and picked up by specialist press (Inman, HousingWire). Usage and conversion metrics, without isolated financial impact, hence a C level.
How it works
Inferred typical approachThe internal detail is not public. Here is a proven approach that leads to the same result, to adapt to your stack.
The stack in detail
- plateforme Sendbird AI conversational agent platform cited as a partner in building Ask Redfin.
- llm LLM de generation de reponses Generated answers grounded in listing details and the local market; the exact model is not publicly named.
- infra Donnees d'annonces et de marche local Redfin Structured listing content and local conditions that serve as grounding for the answers.
- outil Garde-fous fair housing A set of instructions that enforces answers compliant with fair housing rules and refuses questions touching on housing discrimination.
How it runs, concretely
For ops teams-
1Question about the listing customer
The user opens Ask Redfin on a listing and asks a question about the property or the market.
-
2Grounded and filtered answer AI
The assistant answers from the listing details and local conditions, refusing fair-housing-risk questions.
-
3Handoff to an agent human
For a service or tour request, the assistant hands off to the support team, which connects to a licensed agent.
The structured content of the listing and the local market data. If the assistant answers without grounding in this data, it drifts and steps outside the fair housing frame.
How your customers perceive this type of use
Sourced studiesLes consommateurs n'acceptent pas les chatbots par defaut : 64% prefereraient que les entreprises n'utilisent pas d'IA dans leur service client (Gartner, 2024) et pres d'un utilisateur sur cinq du service client par IA n'en retire aucun benefice (Qualtrics, 2025). L'acceptation se construit sur trois conditions mesurees par Salesforce : savoir qu'on parle a une IA, pouvoir escalader vers un humain, comprendre la logique de l'agent.
Acceptance conditions
- Etre informe qu'on parle a une IA et non a un humain (pres de 75% le demandent, Salesforce 2024)
- Un chemin d'escalade clair vers un agent humain (45% plus enclins a utiliser l'agent IA, Salesforce 2024)
- Une logique de l'agent clairement expliquee (44% plus enclins, Salesforce 2024)
Red lines
- Rendre l'humain injoignable : c'est la premiere inquietude des consommateurs sur l'IA dans le service client (Gartner 2024) et 50% craignent que l'IA les coupe du contact humain (Qualtrics 2025)
- Remplacer le service client par l'IA sans alternative : 53% envisageraient de partir chez un concurrent (Gartner 2024)
Sources: Salesforce 2024 · Gartner 2024 · Qualtrics 2025
How to replicate
Inference, not sourcedData prerequisites
- structured listings with usable features
- local market data
- tour availability and routing to agents
Org prerequisites
- compliance and anti-discrimination guardrails
- support team for the human handoff
- engagement and conversion measurement loop
Possible stack
- LLM grounded on the listing (RAG)
- conversational agent platform
- refusal rules on sensitive subjects
The plan, step by step
- Step 1Define the scope of covered questions and write the compliance guardrails (anti-discrimination, refused subjects).Deliverable: Documented answer and refusal rules.
- Step 2Build the assistant grounded on the structured listing content (RAG) via a conversational platform.Deliverable: Internal beta on a subset of listings.
- Step 3Connect the human handoff: escalation to the support team then connection with a licensed agent, and tour availability.Deliverable: End-to-end escalation journey operational.
- Step 4Open a public beta on a single surface (mobile app) with measurement of return to the app and agent requests.Deliverable: Live assistant with an engagement dashboard.
- Step 5Read engagement and conversion (tour requests, connections) and extend the surface.Deliverable: Quantified review and a generalization plan.
First step: Ground the answers on the structured listing content before opening up the field of questions.
The same pattern, proven in other industries
Amazon
genAI shopping assistant
Booking.com
genAI trip planning assistant
Burberry
genAI copilot for consultants (search, image recognition, complete-the-look)
Sources
- S1 Ask Redfin Launches Nationwide, Helping House Hunters Get Quick Answers Primary archive pending
- S2 Introducing Ask Redfin, an AI-Powered Tool to Quickly Answer Questions about For-Sale Homes Primary archive pending
- S3 Redfin Is Beta Testing Its New AI Search Assistant Ask Redfin Secondary archive pending
- S4 Redfin builds a custom AI virtual assistant to answer questions about for-sale homes Interested party archive pending
An error, newer info, a source?
This page lives on its accuracy. If a figure has moved, if the deployment has changed, or if you have a higher-quality source, tell us. Every sourced correction is verified before publication.