Breeze Voice AI

AI-powered voice support for Breeze guests. Built in-house. Ready to scale.

The Problem

No Voice Support

Breeze started with chat and email only, with a strong emphasis on self-service. Voice support has always been missing.

Customer Demand

Our primary demographic has always wanted voice support. It's been a consistent ask that we haven't been able to address.

Too Expensive

Traditional voice support is expensive and difficult to maintain. The economics have never worked for our model.

Vendor Limitations

We've evaluated many vendors. Every demo has been subpar: robotic voices, poor data integration, lack of customization, and high costs.

The Cost Reality

We bought a 3rd-party AI chat tool. Here's how the costs compare:

3rd Party Tool
$0.40
per conversation
  • Limited customization
  • Poor integration
  • Vendor dependency
Our Internal Tool
$0.01
per conversation
  • Fully customizable
  • Deep integration
  • We own it

That's a 40x cost reduction with a better product.

The Solution

Generative AI now makes voice support possible. We built it ourselves.

< 1 week
Build Time
$0.04-0.06
Per Minute
100%
Customizable

Natural Conversations

Human-like voice interactions powered by state-of-the-art speech-to-text and text-to-speech models. No robotic responses.

Deep Integration

Connected directly to Navitaire for real-time flight and booking data. Integrated with Gladly knowledge base for instant answers.

Fully Customizable

We own the code. We can add features, adjust behavior, and respond to feedback without waiting on vendors.

Demo Recordings

Listen to real conversations with Breeze Voice AI

Try It Yourself

Nothing beats experiencing it firsthand. Call now and have a conversation with Breeze Voice AI.

+1 (XXX) XXX-XXXX

This line is live and available for testing

Technical Overview

Built on proven technologies with a focus on reliability and cost efficiency

Voice Pipeline

LiveKit Agents
Real-time voice framework
Deepgram Nova-2
Speech-to-text
Deepgram Aura
Text-to-speech
Silero VAD
Voice activity detection

Intelligence

OpenAI GPT-4o-mini
Language model
RAG + pgvector
Knowledge retrieval
Gladly Integration
Knowledge base sync

Integrations

Navitaire API
Flights, bookings, check-in
PostgreSQL + Neon
Database
LiveKit Cloud
Hosting & telephony

How It Works

When a call comes in, the AI agent joins as a participant. It listens to the guest, processes their request through the language model, retrieves relevant information from our knowledge base and Navitaire, and responds naturally. All conversation state is managed per-call and logged for quality assurance.

Next Steps

Where do we go from here?

1

Internal Testing

Continue gathering feedback and refining the experience. Ensure the system is robust and ready for scale.

2

A/B Testing

Roll out to a small percentage of calls to measure performance against current support channels.

3

Production Deployment

Full rollout to all voice support channels with monitoring and continuous improvement.

What We're Asking For

  • Executive Sponsorship — Clear support to move this forward without bureaucratic delays
  • Timeline Agreement — When do we feel confident to launch?
  • Feedback — Any showstoppers? Concerns we should address?

The phone line will remain live. Test it, share it, and let us know what you think.