Case study 02 · E-commerce
Four out of five support tickets were the same three questions about an order that the systems already knew the answer to. We connected those systems to a chat agent that answers from live data and escalates the rest with context attached.
- Industry
- E-commerce
- Engagement
- 9 weeks to launch
- Channels
- Website chat, WhatsApp, email
- Status
- Live in production
Direct-to-consumer retailer, ~12k orders a month
The challenge
Customer teams were spending time handling repetitive order and support questions.
Where is my order, can I change the address, and how do I return this made up the large majority of the queue. Every one of them was answerable from the order management system, but the answer had to be looked up by a person, copied into a reply and sent — a few minutes each, thousands of times a month.
The volume also hid the queries that genuinely needed judgement. A damaged delivery or a payment dispute sat in the same queue as a tracking question and waited its turn behind it.
What we built
- AI Chat Agent
- Order lookup
- Customer support automation
- WhatsApp integration
- Notification workflows
How it works
- ICustomer asks about an order
- IIAI identifies the customer and order
- IIIAI retrieves live status from the systems
- IVAI resolves or escalates with context
- VCustomer receives proactive updates
Results
- 72%
- Order and returns queries resolved with no human involved
- 41s
- Median first response across chat and WhatsApp
- 31%
- Lower support cost per order
The outcome
Customers receive instant answers while the team focuses on higher-value issues.
The agent identifies the customer, reads live status out of the order system and answers in the channel the question arrived on. Address changes, return labels and delivery windows are handled end to end, inside the business rules the team already had written down.
What it cannot settle, it hands over with the order, the history and its own reading of the problem already attached, so an agent starts from context rather than from scratch. The queue is now mostly the cases that deserve a person, and it is short enough to work properly.
Stack
Have a process worth automating?
Tell us what your team is doing manually. We'll show you what AI can take over.