How AI resolved 70% of our client's support tickets in 30 days

A behind-the-scenes look at how we built a custom n8n automation workflow that handles customer support emails automatically for a Hertfordshire e-commerce SME — cutting response times from 18 hours to under 4 minutes.

The problem: a full inbox, a burnt-out team

When we first spoke to this Hertfordshire-based e-commerce SME in early 2025, their two-person customer support team was handling over 200 emails per week. Response times had crept up to 18 hours on average. Customer satisfaction scores were sliding.

The owner had tried outsourcing to a VA agency. It helped briefly, then quality dropped and costs rose. She'd heard about AI but assumed it was only for large enterprises with big IT budgets.

200+

emails per week before automation — 70% were the same 12 questions

Our approach: start with the knowledge base

Before writing a single line of automation, we spent two days with the team mapping every inbound email type they received. We categorised 6 months of ticket history and found that 71% of all emails were variations of 12 questions — order status, returns policy, delivery times, product sizing, discount codes and password resets.

That's an important insight: AI doesn't need to handle everything. It needs to handle the repetitive majority so your team can focus on the complex minority.

We built a knowledge base covering all 12 question types, with branching logic for edge cases (international orders, damaged items, etc.) and explicit escalation rules for anything involving legal language, large refund requests or complaints.

The automation build: n8n + Claude AI

We built the workflow in n8n — an open-source automation platform that connects their email (Gmail), their Shopify store data and our AI API into one pipeline. Here's how a typical ticket flows:

  1. Email arrives — n8n picks it up within 30 seconds
  2. AI reads and classifies — identifies the intent and checks against the knowledge base
  3. Confidence scoring — if confidence is above 85%, the AI drafts a reply; below that, it flags for human review
  4. Reply or escalate — auto-replies go out signed from the team; escalations appear in a Slack channel with full context
  5. Logging — every ticket, classification and reply is logged to a Notion dashboard for monthly review

Important: The AI never pretends to be human. Every auto-reply is clearly from the company's email address and signed by the team name. Customers can still reply and their response threads back into the system.

The results after 30 days

We went live in week three of the project. By the end of month one:

70%

of tickets resolved automatically by month one — without a single human reply

What the remaining 30% looked like

The 30% that came through to the human team were almost entirely complex cases: partial refunds on multi-item orders, delivery disputes with couriers, bespoke product requests. These are exactly the cases where a human touch matters — and the team now has the bandwidth to handle them properly.

One unexpected benefit: the escalation summaries the AI generates before flagging to the team save around 3 minutes per ticket in context-gathering. The team arrives at each escalation already knowing what happened, what was ordered and what the customer said previously.

Month three: expanding the automation

Buoyed by the results, we added two more automations in month three:

The second automation alone generated 23 new Trustpilot reviews in its first month — the business had collected only 11 reviews in the previous two years.

What it cost vs what it saved

The setup was on our Growth plan: £3,000 setup + £600/month. The business had previously spent £1,200/month on VA support for email handling. Net saving in month one: £600/month, plus a significant improvement in quality and response times.

By month three, the owner estimated the automation was saving her team around 12 hours per week — time they were reinvesting into proactive customer outreach and new product launches.

Key lessons

Ready to apply this to your business?

Book a free 30-minute strategy call. We'll give you a concrete plan — no obligation.

Book Free Strategy Call →