From regulatory compliance to AI powered CX: a decade long partnership reinvigorated by Copilot and QA.

Nearly ten years of Continuous Improvement on Zendesk culminating in a Copilot and QA rollout that cut full resolution time by two thirds and made quality assurance forty times more efficient for the team behind 22.9 million annual journeys.

  • -66% full resolution time since QA rollout
  • 40x more efficient automated vs. manual QA
  • +21% complaint CSAT since Copilot & QA rollout
  • 1 min average first reply time for messaging
The challenge

A small team, a regulated industry, and demand that arrives in waves.

Chiltern Railways has been moving people in and out of London for nearly three decades, connecting the capital to Oxford, Birmingham, Stratford-upon-Avon, and beyond. As a seven day a week operator, it serves a diverse passenger base: weekday commuters, weekend travellers, and tourists, every day of the year, from 6am to 10pm.

Behind every journey sits a contact centre of just ten agents handling communications across phone, email, social and messaging. That team faces two structural challenges. The first is regulatory: as a rail operator, Chiltern must comply with strict reporting requirements set by the Office of Road & Rail (ORR), with resolution time allowances that can't be missed. The second is volatility: demand is unpredictable. "Our two most popular channels are web forms and X. We get our ebbs and flows based on what's happening on the ground. If we're in major disruption (broken down trains, signal failures), then we get high volumes of contacts, and then they drop off once things are running normally again," explains Contact Centre Manager Marc Costello.

The team had a working platform, but by their own admission, they'd been "coasting along for a little while" on a setup that had stayed stagnant. To unlock the next step, bringing AI powered assistance, automated quality management, and the productivity gains both could deliver, Chiltern needed a partner to lead the implementation and a foundation strong enough to support it.

Our approach

Build the foundation, then add the AI.

  1. 01

    Foundation

    A unified Zendesk implementation in 2017, built with bespoke operating tools to fulfil ORR reporting requirements. Resolution times dropped from around 13 days to 6, a 54% improvement that transformed operational efficiency and gave the team a single view of every passenger interaction.

  2. 02

    Content & consistency

    Over 200 knowledge base articles built across ticket refunds, service disruption procedures and policy, giving passengers self service options and creating the structured, well defined content that AI tools would later depend on. Consistent ticket structure and a clear QA rubric were established in parallel.

  3. 03

    Copilot

    Copilot was configured to leverage Chiltern's full knowledge base: external articles and internal documentation alike. The "Customer Relations Bible", hundreds of how-to guides scattered across Microsoft Teams folders, was consolidated into Copilot's internal knowledge base, so agents could find answers by keyword instead of hunting through folders.

  4. 04

    QA

    Zendesk QA replaced manual sampling with automated review across every interaction. The team can now track previously hard to measure dimensions, like empathy, alongside greeting quality, readability, tone and grammar. Copilot and QA together create a feedback loop: QA identifies what to improve, Copilot helps agents deliver it consistently.

Outcomes

A reinvigorated team and measurable service gains.

  • 66% reduction in full resolution time since the QA rollout, with a 21% rise in complaint CSAT over the same period.
  • Automated QA running 40 times more efficiently than manual reviews, with the overall Internal Quality Score up 6.3% in five months.
  • Quality scores up across the board: greeting +4.2%, readability +3.9%, tone +1%, grammar +0.8%.
  • First reply times of 1 minute on messaging, 2 minutes on Facebook, 5 minutes on X, providing rapid acknowledgment precisely when passengers need it.
  • 7.8% drop in overall ticket volume year on year, with one touch tickets down 8.6%, as self service and automation deflect routine enquiries so agents can focus on complex cases.

"Copilot and QA have enabled us to unlock far greater benefits from Zendesk in terms of agent productivity and, ultimately, passenger experience. We got used to what we had, we were happy with it, and we just coasted along for a little while. These tools reinvigorated the team's approach to customer service."

Marc Costello Contact Centre Manager, Chiltern Railways
What's next

Toward a fully unified omnichannel contact centre.

The foundations are now strong enough to keep compounding. Copilot is in agents' hands. QA is running automatically across every interaction. The knowledge base is doing real work for both customers and the team. The next steps, telephony integration, further automation, and expanded language support, will close the remaining gaps and create a fully unified omnichannel contact centre.

"For Zendesk Copilot and QA to succeed, it's imperative that you have clean, well defined knowledge, consistent ticket structure and policies, a clear QA scoring rubric, and strong feedback loops so we can trust and continuously improve AI."

Rick Hutchinson, Appamondo co-founder

If you'd like to do the same, whether your stack is Zendesk Suite, Copilot, QA, or somewhere in between, we'll happily explore the path with you.

  • rail
  • transport
  • regulated-industries
  • copilot
  • quality-assurance
  • continuous-improvement