AI-native CRM with configurable inbound/outbound voice agents and a self-learning knowledge base.
Self-learning KB - agents get smarter after every call automatically
01 / Context
A Salesforce/HubSpot competitor built AI-first - voice agents handle inbound/outbound calls, take notes, schedule meetings, and the system learns from every conversation.
02 / The Problem
Existing CRMs bolt AI on top. This needed AI at the core - agents with real tool access, per-agent knowledge bases, and automatic learning from transcripts.
03 / Key Decisions & Tradeoffs
Chosen
ElevenLabs voices + Twilio call routing
Alternative
VAPI or Bland.ai
Rationale
ElevenLabs gives best voice quality and native tool-calling support. Twilio handles phone number provisioning separately.
Chosen
Agents synthesize transcripts into KB articles
Alternative
Manual KB maintenance
Rationale
Manual KB maintenance does not scale. Synthesis means the system gets smarter with zero human effort.
Chosen
RabbitMQ message queues
Alternative
Direct service calls
Rationale
Lead events fan out to multiple consumers. Queues give reliable at-least-once delivery without tight coupling.
04 / Hard Parts
Wiring LLM tool calling to live CRM actions (meeting scheduling, note-taking). Building a self-learning KB that synthesizes call/meeting/debrief transcripts into new articles automatically. Real-time meeting agent delivering live suggestions to reps mid-call.
05 / Outcome
Configurable voice agents (ElevenLabs + Twilio), full lead-timeline event tracking, self-learning KB, real-time debrief agent that auto-creates LMS skill tasks, SMTP/SMS for follow-ups.
Tech Stack
Full walkthrough available on request — reach out.