Back to Blog
· by Alexandra Lartey ·Contact Centre Trends

UK Contact Centre Attrition Is Still Above 30%: Here Is What Actually Reduces It

Illustration reading "UK contact centre attrition in 2026: what the data says" with the Onboardo logo.

Every UK contact centre director has asked the same question this year: how do we reduce attrition before it reduces us?

The numbers haven’t moved much. What has changed is the evidence for what actually works — and it’s not just the pay review.

The 2026 Picture, in Three Numbers

Pay is doing something. The UK Contact Centre Forum’s 2026 Performance Review found that average starting salary rose 9.18% to £26,260 this year — and centres paying above market saw attrition drop to 18%, against 27.1% for the rest.

That’s a real gap. It’s also not the full story, because 18% is still one in five agents walking out the door of a business that has just given its best people a raise.

Zoom out, and the industry baseline barely shifts. Odigo’s analysis of ContactBabel research shows UK contact centre attrition holding at around 20% since 2011 — over a decade with essentially no improvement. Several 2026 reports — including Emite’s contact centre outlook — put the figure above 30% industry-wide, twice the UK’s all-industry average. Onboardo’s own design-partner conversations put BPO and outbound floors even higher: 40–60%+.

So pay helps, but it doesn’t fix it.

Why Burnout Is Driving UK Contact Centre Attrition in 2026

Here’s the shift worth noticing. This year’s research doesn’t blame wages alone — it blames friction.

Emite’s 2026 contact centre outlook ties attrition above 30% directly to burnout from tool sprawl and cognitive overload. CX Today’s February 2026 round-up goes further: the average contact centre now juggles 3.9 different platforms, and CX leaders link that fragmentation straight to training complexity and burnout.

The same round-up found that 89% of CX leaders now expect AI to shorten onboarding — and where it’s already happening, nearly half report new agents reaching confidence in 2–4 weeks. Compare that with the traditional model, where a new starter shadows for 4–6 weeks before anyone decides they’re ready to go live.

It’s worth naming the pattern directly: what looks like a people problem — an agent who “just wasn’t cut out for it” — is usually a process problem. Attrition is the clearest example of that on the floor today.

Why Shadowing Keeps Producing the Same Result

Picture the standard UK SME induction: a PowerPoint deck and printed SOPs in week one, shadowing in weeks two to four, a buddy nearby in weeks four to six, then live calls — often before a single structured practice attempt.

This approach has two structural issues. Shadowing quality depends entirely on who a new starter is paired with, and there’s no consistent way to check readiness before a real customer becomes the test. Every hour a team leader spends shadowing one person is an hour they’re not spending on the other thirty agents already on the floor.

Onboardo’s own market research puts the fully loaded cost of a new UK contact centre hire at £10,000–£12,000 before their first good call. Run that across 40–60% attrition on a 100-seat floor, and you’re carrying an enterprise-scale cost with none of the enterprise-scale infrastructure to manage it.

Three Things That Move the Needle

1. Pay above market, where you can. The evidence is now settled — a nine-point attrition gap is real. But every competitor can pull the same lever, and it doesn’t touch the readiness problem sitting underneath the churn.

2. Cut the friction, not just the workload. Agents rarely quit because they dislike customers. They quit because the job stays stressful in the same way, every day — including the low-grade dread of taking calls they’ve never actually practised.

3. Build confidence before the first real call, not during it. This is the direct read on the 2–4 weeks finding. Structured, repeatable practice beats passive shadowing — and agents who arrive confident are the ones still there in month four.

Where AI Contact Centre Training Fits

This is where the category has moved in 2026. AI contact centre training — specifically AI-powered voice roleplay — gives agents a safe place to practise realistic calls before they speak to a real customer.

An agent takes a simulated call with an AI customer that responds in real time, adapts to what the agent says, escalates when the answer falls short, and stays in character when the conversation goes sideways. The AI holds regional accents, adjusts emotional tone mid-call, and models natural breathing patterns and speech cadence. Leading voice AI now passes as human in 8 out of 10 blind listening tests. The agent knows it’s a simulation. Their nervous system does not — and that’s precisely what makes it effective for building real confidence.

AI roleplay simulations already achieve training completion rates of 80–90%, compared to 15–20% for traditional eLearning. Agents can run the same difficult call ten times in a row, at any hour, without a supervisor in the room and without a real customer on the line. After each session, they get scored on what actually happened — not how they felt it went.

Where Onboardo Fits

Onboardo is an AI-powered contact centre training platform built to sit between your induction deck and a new agent’s first live call — unlimited, realistic roleplay against AI customer personas covering objections, retention conversations, and compliance scenarios, before a real customer is on the line.

It isn’t an LMS, doesn’t need IT sign-off or six-month rollout. A contact centre director or senior team leader can set a new hire up in minutes, with performance data that turns “I think they’re ready” into a measured answer instead of a guess.

Ready to see it in action? Schedule a demo and join our design partner programme.


Sources: UK Contact Centre Forum 2026 Performance Review · Odigo, citing ContactBabel UK attrition research · Emite, “The Contact Centre in 2026” · CX Today, “Contact Center Industry Reports 2026” · Onboardo internal design-partner research.