Ask any contact-centre director where their P&L bleeds and you'll get the same answer: handle time. A 30-second cut across a 200-seat operation is roughly 1,000 agent hours a month back. So why does the average UK contact centre still post AHTs north of seven minutes, with payment-handling calls regularly running past ten?
Because most of the published advice is about coaching scripts and a fresh CRM rollout. Both help — neither cuts AHT by the kind of margin that shows up in next quarter's board pack. The real wins sit somewhere boring: the warm transfer that never happens, the re-keyed reference, the IVR that reads a balance the agent already has on screen, and the payment portion where the agent goes mute for 90 seconds while the customer reads card digits off a wallet.
We've spent the last decade watching contact-centre payment calls, so the payment piece is where we'll go deep. But this post covers the full picture — what we've seen actually drop AHT in 2026 operations, and the bits that get written up as silver bullets but don't survive a real deployment.
What "reducing AHT" actually means in 2026#
Average Handle Time is talk time plus hold plus after-call work (ACW). It's the headline number, but every credible contact-centre director we talk to now slices it three ways:
- Productive talk time — the bit that resolves the customer's reason for calling.
- Friction time — hold, transfer, re-authentication, re-stating the problem.
- ACW — notes, wrap codes, system updates, compliance logging.
The CCMA's 2025 UK Industry Report put average AHT across multi-skilled UK contact centres at 6 minutes 43 seconds, with payment-bearing calls (utilities, telco, retail, charity, public sector) typically 30-40% longer. ACW alone accounts for 80-110 seconds of the average call. That's where some of the easiest wins live, and where most coaching programmes still aim last.
If you don't split AHT into those three buckets, you'll end up optimising the wrong one. Hammering productive talk time is how you get a CSAT drop and a churned agent. Hammering friction time is how you get back six figures a year and keep your CSAT trend up.
The seven AHT levers that pay back in 2026#
We've ordered these by impact in the operations we've worked with — not by how often they get written about. The first three sit above the line in nearly every business case we see; the bottom four are situational.
1. Cut the payment handoff
Here's the one nobody puts at the top of the list and they should. A typical card-not-present payment over the phone in 2026 still runs 90-180 seconds: agent reads the script, customer fumbles for the card, agent types digits as they're read, agent confirms, agent waits for the gateway, agent reads the result. If anything mis-keys, add 45 seconds. If the agent has to take the call off-script for PCI scope ring-fencing, add another 30 to read the privacy notice.
You don't fix this by training agents to type faster. You fix it by getting the agent out of the card-data path entirely. With DTMF masking, the customer keys their own card details on their phone keypad. The agent sees stars on screen, hears suppressed monotone audio, and stays on the line through the whole transaction. There's no transfer, no callback, no "please key in your card and we'll call you back to confirm." Talk time drops because the awkward 30-second silence is replaced by 12-15 seconds of typing. Hold time drops to zero. The reduction we see in operations after deployment is consistent: 45-90 seconds shaved off every payment-bearing call.
The compounding effect matters. A 200-seat contact centre taking 1,200 card payments a day saves roughly 18 agent-hours daily — that's three FTEs back without touching any other metric. And because the agent never sees card data, PCI DSS scope shrinks dramatically, which is why finance signs the business case.
2. Authenticate once, properly
Most contact centres still authenticate at the start of the call (KBA, postcode, date of birth) and then re-authenticate before any sensitive action. That's two minutes of friction you can engineer away with voice biometrics or a one-time push to the customer's app — done in the background while the agent's already talking. Pindrop and Nuance have published case studies showing 30-60 second authentication time savings; we've seen similar numbers when the deployment is genuinely passive rather than "please say your passphrase."
If you can't justify biometrics yet, at least kill the second authentication. There's no security benefit to re-asking someone their mother's maiden name three minutes after they passed the same check. Use a session token internally and move on.
3. Give the agent real agent-assist, not a transcript
The 2024 generation of agent-assist tools was a glorified live transcript with a wiki search bolted on. The 2026 generation is genuinely different — LLM-driven next-best-action prompts that read the customer's intent in 3-4 seconds and surface the right journey. Forrester's late 2025 contact-centre wave showed median AHT reductions of 12-18% in operations that deployed agent-assist with proper workflow integration. Operations that deployed it as a sidebar widget got nothing.
The difference is whether the tool can actually execute the journey or just describe it. If the agent still has to click through six screens after the AI tells them what to do, you've added cognitive load and saved zero seconds.
4. Kill the after-call wrap by writing it during the call
ACW is the easiest 30-second win in the building. Modern speech-to-text plus an LLM summariser can produce a call summary, suggested wrap codes, and a CRM update by the time the customer hangs up. The agent reviews, edits if needed, and clicks save. We've seen ACW drop from 95 seconds average to 22 seconds in operations that deployed this properly in 2025.
The catch: don't let the AI auto-save. Agent review stops the compliance and accuracy issues that will otherwise eat the savings back when QA finds them in month three.
5. Stop transferring calls you could have routed
Transfer rate is the metric that hides in plain sight. Every transfer is roughly 45-60 seconds of dead time plus an authentication re-do plus the customer re-stating their problem (which Forrester data says is the single biggest CSAT killer in contact-centre interactions). If 12% of your calls transfer, that's 6-7% of your total AHT spent on a process that loses customers.
The fix is intent classification at the IVR stage that's good enough to route correctly the first time. If your IVR misroutes more than 4-5% of calls, replace it. The tech to do this well costs less than the headcount you're burning.
6. Train for talk time, not against it
This is the contrarian one. Coaching agents to "keep calls short" produces shorter calls and worse outcomes — they push customers off the line, repeat callback rates spike, and your real cost-per-resolution goes up. Coach for first-call resolution instead. AHT will drop on its own because the customer doesn't call back two days later.
Sycurio, Genesys and Five9 customer-success teams all publish data on this and they all say the same thing: operations that target FCR see AHT improvements within 60 days. Operations that target AHT directly see a one-month drop followed by a churn spike.
7. Audit your IVR's read-back
Half the IVRs we audit read information to the customer (account balance, order status, delivery date) that the agent will read back again 30 seconds later. Either the IVR resolves the call without an agent, or it shouldn't be quoting the figure in the first place. Pick one.
What doesn't move AHT in 2026#
A short list of things we keep seeing pitched as AHT levers that don't deliver:
- Bigger knowledge bases. Agents don't search them mid-call. They ask the team lead.
- Mandatory call-flow scripts. They add 20-30 seconds of preamble and frustrate experienced agents.
- CRM rebuilds. Two-year projects that maybe shave 5 seconds. Do agent-assist first.
- "Faster" headsets. Marketing slide, not a metric.
- Outsourcing. Different cost line, not an AHT lever — and CSAT usually takes a hit.
The payment-handling angle nobody puts in the AHT plan#
We keep coming back to this because it's the one with the biggest single AHT delta we can prove, and because contact-centre directors routinely leave it out of the plan. The reason: payment compliance lives under the finance or security team, AHT lives under the ops team, and the conversation doesn't happen.
Three patterns we'd push on:
- Mid-call DTMF capture. The agent stays on the line, the customer keys in card details directly to the payment processor, AHT drops 60-90 seconds. DTMF masking is the cleanest version of this — fully PCI DSS v4.0.1 compliant, no card data ever reaching your network. We deploy this in 3-5 days, not the three-month CCaaS integration the platform vendors quote.
- Self-service payment links for warm leads. For non-time-sensitive payments, secure SMS or email payment links drop the payment off the call entirely. Customers pay in their own time, agent moves to the next call. Works brilliantly for collections and renewals.
- Channel separation when audio recording is mandated. Some sectors (financial services, regulated debt collection) need full audio capture. Channel separation keeps the customer-side audio out of the recording during card entry, so you keep your QA recording without the PCI scope.
None of this is new technology. What's new in 2026 is that the integration story has matured — contact-centre platforms like Genesys Cloud, Five9, NICE CXone and Talkdesk all have proper APIs now, so you're not stitching together a Frankenstein deployment.
The business case to put in front of your CFO#
If we were building the AHT business case for a 200-seat UK contact centre handling 1,200 card-not-present payments a day, the rough numbers we'd put on a page:
- DTMF masking deployment: 75-second average AHT reduction on payment calls → 25 agent-hours/day saved → ~£275k/year at fully-loaded agent cost.
- Auto-wrap (agent-assist generated): 60-second ACW reduction across all calls → ~£420k/year.
- Intent routing improvement: 6% transfer rate reduction → ~£180k/year and a CSAT bump worth more.
- Authentication consolidation: 45-second auth reduction → ~£155k/year.
That's roughly £1m of run-rate saving from four projects that each pay back inside 12 months. None of it requires you to push agents harder. All of it requires you to stop making them do work the technology should be doing.
If you want to see how the payment handoff piece works in practice, our DTMF masking glossary entry walks through the call flow, and our team can show you a live demo in about 20 minutes — no slideware, just a real call.
