Voice Based Customer Support Agent

Overview

A voice based AI customer support system for RelayPay that lets customers ask support questions through a web interface and receive accurate spoken responses grounded in approved company documentation.

Problem

Support agents spent significant time on repetitive questions about onboarding, payouts, transactions, and compliance, even when answers existed in internal docs. Response quality varied and scaling became difficult.

Solution

Built a voice agent using VAPI, n8n, Claude, and RAG architecture with a custom React portal. Speech is converted to text, matched against a vector knowledge base, and returned as spoken responses with escalation logic.

1. 1.Customers interact via a branded web portal; speech converts to text in n8n.
2. 2.RAG retrieval matches questions against approved RelayPay documentation.
3. 3.Claude generates grounded responses with escalation and fallback rules.
4. 4.Responses convert back to speech; interactions log to Airtable for tracking.