Case · Fashion retailer, United Kingdom
Customer care that never closes.
A UK online fashion retailer answered customers from 9 to 6, five days a week. Their customers shop at midnight. Now two agents answer in seconds, every day, and the care team handles only what needs a human.
- Industry
- Retail & e-commerce
- Team
- Customer care
- Agents
- Classifier + Responder
- Live in
- 30 days
Results
Illustrative case based on real projects. Figures are representative.
Before
Monday mornings started with 900 tickets.
Order status, returns, sizes, delivery dates. Most tickets were the same ten questions, across email, chat and WhatsApp. Every weekend left a backlog the team spent Monday clearing.
Slow answers meant abandoned carts and angry reviews. Peak season meant six temporary hires who needed weeks to learn the policies.
What we built
Every channel, one brain, your policies.
The agents work inside Zendesk and Shopify. Customers see the same brand voice, just faster.
Email, chat, WhatsApp
Customers write the way they always did, in their language.
Classifier
Reads the message, finds the order, tags intent and urgency.
Responder
Answers status, returns and sizing from Shopify and your policies.
Care team
Refunds over €500 and upset VIPs go to a person, with the full context.
- Approval rule
- Refunds over €500 and any VIP complaint go to a person
- Shared memory
- Returns policy, size guides, delivery times, order history
- Connected to
- Shopify · Zendesk · WhatsApp
The 30 days
From first call to live, week by week.Week by week.
- WEEK 0
Diagnose
Three months of tickets analysed. The top ten questions covered 70% of the volume.
- WEEK 1
Map
Access to Shopify and Zendesk. Metric signed off: first response time.
- WEEK 2
Build
Classifier and Responder trained on policies and past answers, tone approved by the brand team.
- WEEK 3
Prove
Shadow mode on live tickets. The team edited drafts, agents learned.
- DAY 30
Live
Agents answering 24/7 in five languages. Team trained on the command center.
After
What changed.
Measured from the command center, against the metric agreed in week one.
| Metric | Before | After 90 days |
|---|---|---|
| First response time | 9 hours | 30 seconds |
| Tickets resolved without a person | 0% | 58% |
| Monday backlog | ~900 | 0 |
| Customer satisfaction (CSAT) | 4.1 | 4.6 |
| Languages answered | 2 | 5 |
| Temporary hires for peak season | 6 | 0 |
- First response time9 hours, now 30 seconds
- Tickets resolved without a person0%, now 58%
- Monday backlog~900, now 0
- Customer satisfaction (CSAT)4.1, now 4.6
- Languages answered2, now 5
- Temporary hires for peak season6, now 0
More cases