Book a call
All work

Case study · Real estate

LIVA.

An AI-powered conversational video platform for real estate businesses that uses a live AI avatar to engage property buyers, capture qualified leads, and deliver them to agents.

What we built
White-label multi-client SaaS with live AI video avatar and admin dashboard
Stack
React.js, TypeScript, Tailwind CSS, Node.js
Services
AI video avatarLead qualification automationCRM integrationAdmin dashboardWhite-label SaaS
Admin sign-in screen over a home interior on a laptop on a wooden table

About the project

What is LIVA?

This project involved developing an AI-powered conversational video platform for real estate businesses. The system uses a live AI avatar to engage property buyers, answer property-related questions, identify buying intent, capture qualified leads, and automatically deliver those leads to real estate agents. The platform was designed as a white-label, multi-client SaaS system, allowing different real estate businesses to deploy branded experiences with isolated data, custom knowledge bases, and client-specific configurations.

The problem

What was slowing them down.

  1. 01

    Lost leads

    Real estate businesses often lose valuable leads because website visitors are left with static property information and traditional contact forms.

  2. 02

    Provide instant assistance

    Provide instant, conversational assistance to property buyers.

  3. 03

    Answer questions

    Answer questions using each client's property knowledge base.

  4. 04

    Identify buyer readiness

    Identify when a buyer is ready to contact an agent.

  5. 05

    Collect information with consent

    Collect personal information only after obtaining proper consent.

  6. 06

    Qualify and distribute leads

    Automatically qualify and distribute leads.

  7. 07

    Support multiple clients

    Support multiple real estate clients without rebuilding the platform.

  8. 08

    Reduce manual work

    Reduce manual work for agents after each buyer interaction.

The solution

What we built.

Conversational AI avatar

SprintX developed a complete AI-driven lead qualification and automation system combining live video avatars, AI processing, CRM integration, and automated workflows.

  • A HeyGen-powered live avatar interacts with buyers in real time, answers questions using client-specific knowledge, asks qualifying questions, supports multilingual conversations, and maintains context for returning visitors.

Conversational lead capture

  • Instead of sending users to a traditional form, the avatar remains visible while collecting buyer information conversationally.
  • Before information is shared with an agent, the system explains the process and obtains explicit consent.
  • Consent records are stored with the property, timestamp, source, checkbox states, and exact wording shown to the buyer.

AI lead qualification

After a session, the transcript is processed with OpenAI to:

  • Score leads as Hot, Warm, or Cold.
  • Identify buying intent and urgency.
  • Generate structured summaries for agents.
  • Highlight important buyer requirements and concerns.

Automated lead operations

Post-session workflows automatically:

  • Send the buyer their conversation transcript.
  • Send the agent a lead summary and score.
  • Create or update the lead in HubSpot.
  • Provide calendar booking links when viewing intent is detected.
  • Log sessions in Google Sheets.

White-label architecture

  • Approximately 60% of the platform is reusable across clients. New deployments can be configured with branding, knowledge bases, email settings, CRM credentials, and property information without modifying the core application.

Admin dashboard

  • The platform owner can manage clients, properties, knowledge bases, avatars, trigger phrases, and platform activity from a centralized dashboard.
Avatar and voice settings with a grid of presenter avatars on a laptop

Results

What changed.

Key business value

  • An automated pipeline

    The solution created an automated lead-generation and qualification pipeline from the first buyer interaction to agent handoff.

  • 24/7 conversational engagement

    24/7 conversational engagement with property buyers.

  • Faster identification

    Faster identification of high-intent prospects.

  • Automated lead qualification

    Automated lead qualification and summarization.

  • Reduced manual follow-up

    Reduced manual follow-up for agents.

  • Consistent lead delivery

    Consistent lead delivery through CRM and email workflows.

  • Reusable white-label architecture

    Reusable white-label architecture for multiple clients.

  • Centralized visibility

    Centralized visibility into sessions, leads, and performance.

  • An intelligent sales experience

    The platform transforms a property website interaction into an intelligent sales and qualification experience.

Key challenges

The hard parts, and how we solved them.

  • Maintaining a natural conversation: traditional forms interrupt the connection between the visitor and avatar.

    Built conversational data capture while keeping the avatar visible throughout the process.

  • Privacy & consent: personal data could not be shared with agents without explicit permission.

    Implemented a consent explanation, independent consent controls, and an immutable audit record containing the exact wording and timestamp.

  • Real-time AI experience: live avatar interactions require low latency and reliable streaming.

    Integrated HeyGen real-time avatar technology with WebSocket communication and optimized the surrounding application workflow.

  • Multi-client scalability: each real estate business requires separate branding, properties, knowledge, and integrations.

    Created reusable white-label architecture with isolated client configurations and data.

  • Valuable conversation data can become difficult for agents to act on manually.

    Automated scoring, summarization, CRM synchronization, email delivery, and calendar-booking workflows.

Sessions list of buyer conversations on a laptop on a dark sofaKnowledge base and property cards with QR codes on two laptops

Tech stack

What it runs on.

Frontend
React.jsTypeScriptTailwind CSS
Backend
Node.jsExpress.js
Database
PostgreSQL
Caching
Redis
AI
OpenAI GPT-4
Avatar
HeyGen Live Avatar API
Automation
n8n
CRM
HubSpotMeetings API
Email
SendGrid
Reporting
Google Sheets API
Storage
AWS S3
Infrastructure
AWS EC2Cloudflare CDNDockerNginx
Realtime
WebSocket / Socket.io

Process

How we got there.

  1. 01Buyer arrives

    Buyer scans a QR code or opens a property link. · Branded landing page loads with property information.

  2. 02Conversation

    Buyer starts a conversation with the live AI avatar. · Avatar answers questions using the property's knowledge base. · Buying intent is detected from the conversation.

  3. 03Consent and capture

    Avatar explains data sharing and requests consent. · Buyer provides contact information through the conversational interface. · Consent is permanently recorded.

  4. 04Qualification and handoff

    Session ends and transcript is processed by OpenAI. · Lead receives a Hot/Warm/Cold score and structured summary. · Buyer receives the transcript while the agent receives the qualified lead. · Lead and session data are synchronized with HubSpot and reporting systems.

Leads list with a lead detail summary panel on a laptopAdmin dashboard with lead quality breakdown and sessions chart on a laptop

More work.

All projects

Plant Style AI Chatbot

Built a private AI-powered business chatbot that centralizes company procedures, policies, templates, planners, and operational documents into a searchable knowledge system for secure and consistent management support.

End to End Catering Automation System

Built an end-to-end catering automation system that connects AI voice reception, web and email inquiries, scheduling, invoicing, payments, staff notifications, and customer follow-up into one structured workflow.

Building something like LIVA?

Book a 30-minute call More work