Zepto
Built an AI powered customer support system that resolved 73% of tickets without human intervention. Faster resolution, lower ops cost and happier customers.
Where they were stuck.
Zepto was handling over 18,000 customer support tickets per day across WhatsApp, email and in app chat. Their human support team was stretched thin with average resolution time sitting at 11 hours. Most tickets were repeating the same 12 issues: wrong order, missing item, refund status, delivery delay. The team was spending 80% of their time on work that required zero human judgment.
What we did about it.
We built a RAG powered support agent trained on Zepto order data, refund policies and delivery SLAs. The agent was connected to their order management system via API so it could pull live order status, trigger refunds automatically and escalate to a human only when the issue required genuine judgment. We deployed it across WhatsApp first using n8n for orchestration, then rolled it out to email and in app within two weeks.
What changed.
Within 30 days the AI agent was resolving 73% of all incoming tickets with zero human involvement. Average resolution time dropped from 11 hours to 4 minutes. Human agents were freed to focus only on complex escalations. Monthly support ops cost dropped by 58% and customer satisfaction scores rose from 3.8 to 4.6 out of 5.
Numbers
that mattered.
The auto up close.
Plotted across the engagement window. Each checkpoint represents a measurable shift in the underlying metric.
We went from drowning in 18K daily tickets to having 73% of them resolved automatically in under 5 minutes. The ops savings alone covered the entire project cost in the first week.
Let's build yours.
Tell us about your brand and we'll send back a 30 days plan with milestones and pricing.