Voice multi-agent at 50,000+ outbound calls per day
Outbound voice agents that qualify, transfer and book — at the scale of an entire BPO, with one natural-sounding voice.
- Calls per day
- 50,000+
- Marketing-qualified leads
- +65%
- Voice latency
- Sub-second
Overview
Simbiosia builds AI-powered sales assistants for B2B teams across Spain and Latin America. We designed and shipped the multi-agent voice infrastructure behind their outbound channel: specialized agents that qualify prospects, transfer to humans and book meetings, all running through a single, natural voice with sub-second latency.
Challenge
To deliver on its promise of warm, human-feeling AI sales conversations, Simbiosia needed a voice infrastructure that could ramp from a few thousand to tens of thousands of calls per day without losing voice quality, latency or qualification accuracy. Off-the-shelf voice bots burned trust within seconds.
What we built
- ›Designed a multi-agent architecture with specialized roles — qualifier, transfer agent, booker and knowledge agent — orchestrated per call.
- ›Built a low-latency voice pipeline: speech-to-text, LLM reasoning and a single, custom natural voice tuned to Simbiosia's brand.
- ›Wired a RAG layer over each client's offer, pricing and qualification criteria so every agent answers grounded, on-brief.
- ›Deployed an autoscaling telephony layer that handles 50,000+ daily calls with retries, transfers and meeting-booking integrations.
Outcome
- ›Sustained 50,000+ outbound conversations per day across multiple Simbiosia clients.
- ›Marketing-qualified leads grew 65% versus the prior baseline.
- ›Sub-second response time made the experience feel like a human conversation, not a bot — the whole point of Simbiosia's product.
Next project
WhatsApp concierge that automates 8,000+ tickets per month
