AI Voice ·Confidential, AI voice agent platform
After-hours calls stop going to voicemail and start arriving as booked appointments.
Small businesses losing revenue every night, because a missed call goes to voicemail and the caller simply dials the next number on the list.
Where this fits
Best for sales and front-desk teams losing after-hours calls to voicemail, where every missed call is a lost lead.
12+ months live in production. Around 18 recovered bookings a month per client.
- Twilio + ElevenLabs voice
- CRM and calendar integrations
- Python and FastAPI on AWS
What we did
- Built the real-time voice pipeline on Twilio and ElevenLabs
- Wired booking straight into Google Calendar and the client CRM
- Ran call handling on Celery and Redis to keep latency down
- Added an integrations layer so each business connects its own tools
Result
Calls that used to hit voicemail after hours are answered, qualified and booked while the caller is still on the line. The platform has run in production for over 12 months, and clients recover around 18 bookings a month that previously went to whoever picked up next.
The hard part
In voice, everything is decided in the gap between the caller finishing a sentence and the agent starting to speak. Speech recognition, the model, speech synthesis and the phone network each want a few hundred milliseconds, and past about a second the caller talks over the agent or hangs up. Checking a calendar slot or writing to a CRM has to happen without spending that budget, which is why the work lives in queueing and streaming rather than in prompts.
- Stack
- PythonFastAPIPostgreSQLGraphQLRedisCeleryTwilioElevenLabsAWS
- Focus
- AI Voice AgentLead EngagementSales AutomationTwilioCRM Integrations