Intelligent Customer Support Automation for FinServ Pro

Buraq designed and deployed an AI-first omnichannel support platform combining large-language-model intent classification, retrieval-augmented generation (RAG), and secure transactional automation. Architecture highlights: • Omnichannel Ingestion Gateway, A unified API layer built with Node.js on Azure Functions normalises inbound messages from email (SendGrid webhooks), in-app chat (WebSocket), WhatsApp Business API, and Twitter/X DMs into a common ticket schema stored in MongoDB Atlas. • Intent Classification & RAG Engine, Incoming messages are classified by a fine-tuned GPT-4o mini model into 48 intent categories with 97.3% accuracy. For knowledge-based queries, a RAG pipeline retrieves context from a vector store (Azure AI Search) containing 1,200+ help articles and regulatory FAQ documents, generating accurate, citation-backed responses. • Transactional Action Layer, Unlike typical chatbots, our AI agent executes real actions: password resets via the identity API, transaction-status lookups from the payments microservice, KYC document re-submission triggers, and refund initiations (up to $50, auto-approved by policy). All actions are logged with full audit trails for compliance. • Smart Escalation, When confidence drops below threshold or the query involves disputes over $50, the system seamlessly hands off to a human agent with full conversation context, sentiment analysis, and a suggested resolution. Agents see a co-pilot sidebar with AI-recommended responses they can send with one click. • Analytics Dashboard, A real-time Next.js dashboard tracks deflection rate, CSAT per channel, agent utilisation, and trending issue categories. Automated alerts flag sudden spikes (e.g., a payment-gateway outage causing a ticket surge).