How we built an AI customer support platform for Shopify merchants that brings emails, order data and store policies into one workflow, with human agents reviewing every reply.
Project overview
Zenva is an AI customer support platform developed by NeuraDynamics for ecommerce businesses, particularly Shopify merchants. It brings customer emails, order information, and business documents into one workflow to streamline support, generating context-aware responses while keeping human agents involved.
The challenge
Ecommerce businesses handle a continuous flow of inquiries about orders, refunds, exchanges, shipping, and store policies. Support teams spend significant time reviewing emails, finding the right order information, checking policies, and preparing repetitive responses. The challenge was to automate routine tasks without losing accuracy, context, and human oversight.
Project objectives
Reduce manual support effort by using AI to handle routine customer inquiries. Improve response speed, accuracy, and consistency across customer interactions. Connect customer emails, ecommerce data, and business knowledge in one workflow while keeping humans involved in final customer interactions.
Our approach
Context-driven automation brings customer emails, Shopify order data, and relevant business documents together so the AI has the context it needs. AI agents classify incoming requests, retrieve relevant information, and generate responses. Support agents stay in control by reviewing AI-generated responses and deciding how each interaction should proceed.
How Zenva works
Merchants connect their Shopify store and email account, then add company information, store policies, and business documents as a knowledge source. Zenva analyses incoming emails, retrieves relevant order details and knowledge, and generates response drafts. Support teams edit and approve responses, take the required action, or route unresolved requests for further handling.
Technical architecture
An AI evaluator identifies genuine customer requests before response generation. RAG-based retrieval brings in policies, FAQs, and support documents, while Shopify data retrieval pulls order and product details. Each draft is reviewed against customer, Shopify, and policy information and regenerated up to three times before being routed for human review. AI-powered policy scoring flags missing information and contradictions in uploaded policies, and the DREAMS mechanism learns from previously edited emails.
Business impact
Based on one store processing 2,000 customer emails per month, Zenva saves approximately 50 hours of support effort and $1,750 each month, and support teams operate at roughly 2× the speed.