If you run or manage a contact centre, you already know the AHT pressure cycle. The board wants it down. The agents say they're already moving as fast as they can. The QA team flags rushed calls. CSAT wobbles. Six months later you've shaved twelve seconds and lost three points of first-call resolution.
Average handling time (AHT) is the total time an agent spends on a single customer interaction, measured as talk time plus hold time plus after-call work (ACW). UK contact centres typically run between 4 and 6 minutes per call. The trick to reducing it without wrecking customer satisfaction is to cut the components that don't involve listening to the customer — payment collection, manual lookups, system switching, post-call admin — rather than rushing the conversation itself.
The problem with most AHT-reduction programmes is that they squeeze the wrong end of the call. If you push agents to talk faster, you trade average handle time for first-call resolution and end up with a worse cost-per-resolved-issue, just spread across more calls. The actual wins come from removing the friction the customer never sees: how long it takes to load a record, capture a card number, type up a disposition note, or transfer to the right team.
What AHT actually measures#
AHT is a composite metric, and that matters because each of its three parts responds to different fixes:
- Talk time — the agent and customer are actively speaking. Mostly determined by the nature of the query and the agent's skill. Hard to compress without hurting quality.
- Hold time — the customer is waiting while the agent looks something up, asks a colleague, or transfers. Highly compressible with the right tooling.
- After-call work (ACW) — the agent is finishing notes, updating CRM records, or completing disposition codes after the customer has hung up. Often the most overlooked component.
Some operations report AHT without ACW. Don't. You'll end up rewarding agents for offloading work to the wrap-up phase, which makes the metric look healthier while doing nothing for cost-per-call.
Six fixes that genuinely reduce AHT#
These are ranked roughly by impact-per-effort for a typical UK contact centre handling mixed inbound work. Your mix will shift the order around.
1. Screen-pop and CTI integration
If your agent has to ask a customer for their account number, postcode, or date of birth before they can do anything useful, you're burning 30 to 60 seconds of talk time per call on identity-shuffling. A computer-telephony integration (CTI) layer that pops the customer's record onto the agent's screen the moment the call connects — driven by the inbound CLI or an IVR-captured reference — removes that step entirely.
Most CCaaS platforms support this natively now. The work is in the CRM mapping, not the telephony.
2. A knowledge base agents actually use
Most contact centres have a knowledge base. Most are dreadful — out of date, badly organised, slow to search. Agents stop using them and either guess, put the customer on hold, or escalate.
The fix isn't a new platform; it's editorial discipline. Pick a small team to own knowledge content. Set freshness SLAs (anything not reviewed in 90 days flags). Track which articles actually get opened during calls. The investment is boring but the AHT impact is real — a well-maintained KB cuts hold time meaningfully because agents trust it enough to use it instead of asking a team leader.
3. Reduce transfers and escalations
Every transfer adds at least 30 seconds of dead air, often a minute, and an unhappy customer who's about to repeat themselves to a stranger. The cause is usually one of three things: routing logic that sends calls to the wrong skill group; agents who aren't trained on adjacent topics; or escalation paths that agents reach for too early because no one ever told them off for it.
Audit your top five transfer reasons quarterly. Most of them are fixable with training, not technology.
4. Eliminate manual lookups during the call
If your agents are alt-tabbing between three systems to handle a typical call, you've got an integration problem, not an agent-speed problem. Unified agent desktops are an investment, but the AHT and CSAT case usually pays for them inside a year. If a full unified desktop isn't on the cards this year, even small wins help — a single sign-on, a shared search bar, automated cross-system note copying.
5. Streamlined payment collection
This one's specific but it shows up on a lot of contact-centre call mixes. If your team takes card payments by phone, manual collection — agent reading digits back, customer correcting, pause-and-resume on the recording, retyping into the CRM — typically eats 90 to 120 seconds per transactional call. That's a third of a 5-minute average call gone on a single sub-task.
An agent-triggered DTMF masking flow drops that to 20 to 30 seconds. The customer types their card details into the keypad, the agent never hears or sees them, and the payment confirmation lands on the agent's screen automatically. It's faster because there's no read-back, no error correction, and no manual data entry.
The bonus is that it removes the agent and the call recording from PCI DSS scope, which is its own conversation. We'll come back to that.
6. Smarter ACW automation
After-call work is where a surprising amount of AHT hides. If agents are typing free-text notes, picking from 40-item disposition dropdowns, and updating two systems by hand, you can probably cut 20 to 40 seconds off every call with structured wrap-up forms, AI-generated call summaries from the recording, and automated CRM field population.
Be careful here — automated summaries are getting good but still need a human pass on anything that affects a customer's account. The wins are real but they're not free of QA.
The honest caveat#
Chasing AHT as a primary metric will eventually hurt you. We've seen contact centres cut AHT by 18% over a year and lose 11 points of CSAT in the same window. The agents who survived the squeeze did so by skipping discovery questions, rushing customers off, and getting through the queue instead of solving problems.
The teams that win on cost-per-resolved-issue treat AHT as a diagnostic, not a target. They look at the components — hold time, ACW, transfer rate, repeat-call rate — and fix the friction that's invisible to the customer. Talk time gets left alone. If anything, it sometimes goes up, because agents who aren't fighting the clock can ask better discovery questions and resolve the issue first time.
Where to actually start#
If you're staring at a 5-minute AHT and a board mandate to bring it under 4, here's the order we'd suggest:
- Break AHT down by call type and queue. The average is hiding a bimodal distribution.
- Identify the call types where a non-conversational component (payment, lookup, transfer, ACW) is more than 30% of the call. Those are your fast wins.
- Pick one component and fix it across the board before moving to the next. Three half-finished initiatives produce zero results.
- Watch CSAT and first-call resolution alongside AHT. If either moves the wrong way, you're squeezing the wrong end.
For most teams that take card payments, the payment-collection component is the cleanest first move because it's a non-negotiable part of every transactional call, the time saving is consistent, and the PCI compliance side-effect makes the business case easier to defend.
If you're taking card payments by phone#
We build telephone payment infrastructure, so we'd be daft not to flag this. If your contact centre handles card payments and you're still doing it manually — agent reads digits back, customer corrects, pause-and-resume on the recording — you're paying for that twice. Once in AHT and once in PCI compliance scope. Moving to DTMF masking with an agent-triggered IVR typically returns 60 to 90 seconds per transactional call and takes the call recording out of audit scope. It's not the answer to every AHT problem, but if it fits your call mix it's one of the few interventions that improves AHT, CSAT, and compliance at the same time.
If you'd like to see what that flow actually looks like — agent's screen, customer's experience, the masked recording — book a fifteen-minute demo. We'll show you a live call, walk through the integration, and you can decide if it's worth a deeper conversation. If it's not, no harm done.
