AI Voice Agent — 20K Calls/Month, 80% Automation
Voice AI agent automating inbound customer support calls for a micromobility startup. ~20,000 calls/month. Delivered in 24 days.
Operators spent 80% of their time on 4 recurring scenarios: trip completion, lock failures, red-zone violations, app freezes. The bot now handles all four autonomously — queries CRM in real time, takes action, closes the call.
Result: ~80% of calls resolved without operator involvement.
Architecture:
Binotel webhook → Whisper STT (UA/RU) → GPT-4o-mini → CRM API → ElevenLabs / Azure TTS
Key solutions:
- Latency: all fixed phrases pre-generated via ElevenLabs and cached. Azure TTS only for dynamic responses
- CRM queries run async during audio playback
- Out-of-script → transfer to operator with full call context
Stack: Go · Binotel · OpenAI Whisper · GPT-4o-mini · ElevenLabs · Azure TTS · Redis · PostgreSQL
Operators spent 80% of their time on 4 recurring scenarios: trip completion, lock failures, red-zone violations, app freezes. The bot now handles all four autonomously — queries CRM in real time, takes action, closes the call.
Result: ~80% of calls resolved without operator involvement.
Architecture:
Binotel webhook → Whisper STT (UA/RU) → GPT-4o-mini → CRM API → ElevenLabs / Azure TTS
Key solutions:
- Latency: all fixed phrases pre-generated via ElevenLabs and cached. Azure TTS only for dynamic responses
- CRM queries run async during audio playback
- Out-of-script → transfer to operator with full call context
Stack: Go · Binotel · OpenAI Whisper · GPT-4o-mini · ElevenLabs · Azure TTS · Redis · PostgreSQL
Kyiv 
Available for hire
Offer work