01 / The problem
What the system needs to solve
Small teams miss calls, and a voicemail box loses context. A business needs a receptionist that answers from its own knowledge, routes callers by intent and hands live calls to the right person, while each tenant's data stays separate.
02 / The approach
How the pieces fit together
- Configure an AI voice receptionist per business with its own knowledge base, instructions and ElevenLabs voice.
- Build IVR menus and conditional routing rules that end at the AI, a team member or voicemail.
- Transfer live calls to team members by phone or browser-based calling over Plivo telephony.
- Isolate tenants with PostgreSQL row-level security, encrypted credentials and two-factor authentication.
03 / Key features
What it does
- AI receptionist with a customizable knowledge base and voice
- IVR menus, conditional call routing and automated voicemail
- Real-time call testing and monitoring
- Call recordings, activity logs, compliance auditing and CSV exports
- Admin dashboard, team management and Stripe subscription billing
04 / Results
What it delivered
- Deployed to production and tested with real phone calls, including AI-assisted conversations and live call transfers
05 / The tradeoffs
Engineering is a set of choices
Enforcing tenant isolation with database row-level security is stricter than filtering in application code, but every query and migration then depends on the correct tenant context being set.
06 / Validation considerations
What to test before relying on it
Place real calls through the full path (AI conversation, IVR branch, live transfer and voicemail fallback) and check the recording, activity log and audit trail after each one.