GroveAI
Case Study

Customer Support Transformation for a National Retailer

Fenwick & Cole, a UK retail chain with 85 stores, was drowning in support tickets. We deployed a cloud AI system that triages, routes, and resolves enquiries — cutting response times from hours to minutes.

Client: Fenwick & ColeIndustry: RetailDuration: 8 weeks

Results

Impact delivered

82%

Faster first response

Average first-response time reduced from 4.1 hours to 44 minutes

41%

Tickets auto-resolved

Enquiries resolved without any human agent involvement

+18pts

Customer satisfaction increase

CSAT score improved from 67 to 85 within three months

35%

Agent turnover reduction

Reduced repetitive workload improved agent retention

£320K

Annual cost saving

Reduced need for seasonal temporary support staff

The Challenge

What they faced

Fenwick & Cole receives over 12,000 customer enquiries per week across email, live chat, and social media. Their support team of 45 agents was overwhelmed, with average first-response times exceeding four hours during peak periods. Customers were leaving negative reviews citing slow responses, and agent turnover was climbing due to repetitive workload. The existing ticketing system offered no intelligent routing — enquiries about returns, stock availability, and delivery tracking all landed in the same queue regardless of complexity or urgency.

Our Solution

How we solved it

We built an AI-powered customer support pipeline that sits between the customer and the existing support team. Incoming enquiries are classified by intent, sentiment, and urgency using a fine-tuned model. Simple queries (order tracking, return policies, store hours) are resolved automatically with personalised responses. Complex or sensitive cases are routed to the most appropriate agent with a pre-prepared context summary. The system integrates directly with Fenwick & Cole's Shopify and Zendesk stack, requiring no changes to existing agent workflows.

Approach

Step by step

01

Enquiry analysis and classification design

Analysed three months of historical tickets (over 150,000) to identify the top 25 intent categories and build a classification taxonomy.

02

Automated response generation

Developed response templates for the 12 most common query types, powered by GPT-4o with access to real-time order and inventory data via API.

03

Intelligent routing engine

Built a routing layer that considers intent, sentiment, customer lifetime value, and agent expertise to assign complex tickets optimally.

04

Integration with existing systems

Connected the AI layer to Shopify (orders, inventory), Zendesk (ticketing), and the company's loyalty platform for personalised context.

05

Phased rollout with safety nets

Launched with automated responses on low-risk query types only, expanding coverage as confidence thresholds were validated over four weeks.

Our agents used to dread Monday mornings — the backlog was demoralising. Now the AI handles the routine stuff and our team focuses on the conversations that actually need a human touch. Customer feedback has been overwhelmingly positive.

James Whitfield

Head of Customer Experience, Fenwick & Cole

Technology

Stack used

GPT-4oZendesk APIShopify APIAWS LambdaAmazon SQSPythonLangChainRedis

FAQ

Frequently asked questions

Yes, for pre-approved query types where accuracy is consistently above 98%. All automated responses are branded and follow the company's tone of voice guidelines. Sensitive topics (complaints, refund disputes) are always routed to human agents.

The initial classification model was trained in under a week using historical ticket data. Fine-tuning for response generation took an additional two weeks, including review cycles with the customer experience team.

Want similar results?

Book a free strategy call and we'll assess how AI can deliver impact for your business.