End-to-End E-commerce AI Automation

The problem

Sena Global ran ManyChat limited to Messenger and Instagram only, with no WhatsApp support. Static decision trees broke on edge cases and the support team was manually answering 15+ questions a day. Responses were generic with no personalization, cart / checkout / browse abandonment signals went uncaptured, and 7,688 customer reviews were scattered across platforms with no unified Shopify presence.

Approach

• Built 5 production flows live across all customer channels (WhatsApp, Messenger, Instagram, Facebook) using n8n as the orchestration layer and Claude API for edge-case reasoning.
• Implemented a bucket-resolver pattern to route inquiries by intent, with Claude handling ambiguous cases that static rules couldn't parse.
• Migrated 7,688 reviews to Shopify via a custom Apify + n8n pipeline with Judge.me integration for $1 in API costs — vs. $1,500+ for traditional migration services.
• Added a per-customer recommendation engine powered by a 14-question diagnostic quiz (RevenueHunt) that feeds results into the AI reasoning layer.
• Deployed cart / checkout / browse recovery flows via WhatsApp Business API with Privyr CRM sync for persistent lead tracking.
• Built with self-hosted n8n to eliminate vendor lock-in. Claude runs as a structured JSON tool, not as the orchestrator.

Results

• ~70% of customer-support inquiries handled autonomously, freeing the team to focus on the 30% requiring human judgment.
• 5 production flows live across all channels — WhatsApp now matches Messenger / IG engagement where competitors are absent.
• 7,688 reviews unified on the Shopify storefront for $1 vs. $1,500+ traditional services.
• WhatsApp-based cart / checkout / browse recovery immediately activated on every trigger.

Tech stack