Quick Answer
A UK homewares retailer with 9,000 monthly tickets cut support costs 70% in 12 months with managed AI support. The AI now resolves 78% of conversations, first-response time fell from 36 hours to under 15 seconds, CSAT climbed from 3.6 to 4.8, and annual savings reached roughly £240,000 — with a leaner, specialist human team.
Key Takeaways
- 0178% of inbound conversations now resolved without human involvement
- 02First-response time fell from 36 hours to under 15 seconds
- 03Annual support cost reduction of ~£240,000 — 70% vs prior year
- 04CSAT rose from 3.6/5 to 4.8/5 despite fewer human agents
- 05Proper training data on UK Consumer Rights Act and Section 75 scenarios was the decisive factor
A well-known UK homewares retailer with twelve high-street stores and a thriving online operation came to us in late 2025 with a familiar problem. Their support team of fourteen agents was handling roughly 9,000 tickets per month, response times had stretched to over 36 hours during peak trading, and Trustpilot reviews were starting to slip. Hiring more staff was not financially viable given the cost pressures facing UK retail, and outsourcing had been tried and failed. They needed a different approach.
We began with a four-week discovery and training phase. Our team analysed 18 months of historical tickets, categorised the top 40 intents, and trained a custom AI on their product catalogue, returns policy, delivery partners and brand tone. We paid particular attention to Consumer Rights Act scenarios, Section 75 credit card disputes and the specific handling of Royal Mail and Evri delivery issues, all of which are daily realities for UK retailers. The AI went live in a shadow mode for two weeks before taking its first real conversations.
The results in the first twelve months were significant. The AI now resolves 78 per cent of all inbound conversations without human involvement, and first-response time has fallen from 36 hours to under 15 seconds. The support team has been restructured into a smaller specialist unit of five agents focused on complex returns, VIP customers and supplier escalations. Total support costs have fallen by 70 per cent against the previous year, equivalent to roughly £240,000 in annual savings, while customer satisfaction on post-chat surveys has risen from 3.6 to 4.8 out of 5.
The lessons for other UK retailers are clear. First, invest in proper training data rather than settling for off-the-shelf chatbots, because British consumer protection law creates specific language patterns that generic AI mishandles. Second, redesign your human team around the work AI cannot do, rather than trying to keep everyone doing the same things. Third, measure obsessively: CSAT, deflection rate, escalation quality and cost per resolved ticket should all be tracked weekly for the first six months to catch regressions early.
Frequently Asked Questions
FAQ
Common questions
Generic chatbots don't understand the specific language of UK consumer protection law — Consumer Rights Act 2015 clauses on faulty goods, Section 75 of the Consumer Credit Act on disputed payments, the 14-day cooling-off period under Consumer Contracts Regulations. When a customer writes 'I paid on credit card so I'm raising a Section 75', a generic bot produces a confused or wrong response. A custom-trained UK model recognises the statutory framing and routes correctly. The retailer had tried two off-the-shelf platforms before WayaNerd; both failed specifically on UK legal vocabulary, which is common across every UK retail deployment we've done.
Ten weeks end-to-end. Weeks 1–4: discovery, 18-month ticket analysis, intent mapping for top 40 query types, training data curation. Weeks 5–6: custom model training on brand voice, product catalogue, returns policy, and UK-specific scenarios. Weeks 7–8: Zendesk integration, shadow mode with AI drafting responses that humans reviewed before sending, measurement baseline. Weeks 9–10: phased go-live — website chat first, then WhatsApp, then email triage. Delivery teams joined for the Royal Mail and Evri dispute flows, which required specific handling. The retailer's five remaining agents were given two weeks of dedicated training before any customer-facing traffic switched over.
The support team of 14 became a specialist unit of 5. Three agents left through natural attrition (they'd been planning to leave anyway), four moved into a new customer-success role focused on VIP and B2B accounts, one moved into operations analysis working with the AI data, and one was made redundant — with a full package and external outplacement support. The retailer's public position was transparent from day one: AI would reduce headcount in support and the specific roles affected were communicated before the project started. Support-team Glassdoor scores improved post-transition because the work became more varied and less repetitive.
Post-chat CSAT survey (1–5 scale) sent immediately after every conversation, regardless of whether it was resolved by AI or a human. Sample size ran at roughly 1,200 responses per month during steady state. CSAT rose from 3.6 pre-deployment to 4.8 in month nine. The drivers were speed (single-digit-second response time vs 36 hours before), 24/7 availability, and consistently polite tone — British customers responded well to the AI's crisp, professional voice once it was properly trained on brand guidelines. The biggest surprise: complex escalations to humans scored higher than pre-deployment, because agents now had full AI-gathered context and fresher mental bandwidth.
The percentage gains scale well below the £240,000 headline figure. A smaller UK retailer with 1,500 monthly tickets running WayaNerd's Growth plan at £299/month typically deflects 65–75% and saves £40–70K annually on support capacity — same proportionate ROI. What doesn't scale down is the training data investment; even small deployments need 6–8 weeks of discovery and training data work to get UK legal vocabulary, brand voice and product catalogue right. Starter plan at £50/month is a fine fit for sub-500-conversation-per-month retailers; above that volume, Growth is the economic choice.
Keep reading
AI Customer Support for SMEs: A Complete 2026 Guide
A practical guide for UK small and mid-sized businesses looking to deploy AI customer support in 2026, covering costs, compliance and realistic outcomes.
10 min read
BusinessThe Real ROI of AI Automation: What to Expect in Year One
We break down the actual numbers behind AI automation projects, from implementation costs to the compounding returns most businesses see within 12 months.
6 min read
AI StrategyThe Small Business Guide to AI Customer Support in 2026
You do not need a Fortune 500 budget to deliver instant, personalized support. Here is how small businesses are using AI to compete with the big players.
8 min read