Walk any contact centre floor and you’ll see the same thing: an agent finishes a call, and instead of taking the next one, they spend two, three, sometimes five minutes typing notes, updating systems, tagging dispositions, and copying information between screens that don’t talk to each other.
Across a 200-seat operation taking 8,000 calls a day, wrap-up time quickly becomes one of the largest controllable costs. If average after-call work (ACW) falls from three minutes to 90 seconds, that releases around 200 agent-hours of capacity every day — before accounting for occupancy, scheduling and demand patterns.
I spent a decade inside one of the UK’s largest contact centre operations — on the phones, as a subject matter expert, and leading teams. I’ve done the wrap-up. I’ve coached people on the wrap-up. And I’ve built the systems that measure it. So this article is about what actually reduces ACW, what only looks like it does, and why the difference matters enormously for your agents.
First, understand what wrap-up time actually is
Before you automate anything, break ACW down. In most operations it’s some mix of:
- Call notes and summaries. The agent recounting what happened, usually in free text and usually inconsistently. This is often the single biggest component.
- System updates. Changing account statuses, logging outcomes and raising follow-up tasks — often across two or three systems that each want the same information entered slightly differently.
- Disposition coding. Picking the reason for the call from a list that may no longer reflect the work people are actually doing.
- Recovery time. The bit nobody puts on a dashboard: a few seconds to breathe after a difficult call. This matters, and we’ll come back to it.
If you don’t know your own split, that’s your first job. Listen to calls, sit with agents and time the components. When I’ve done this properly, the findings almost never match what the leadership team assumed.
What actually works
1. Automated call summaries — often the highest-impact place to start
Modern speech analytics and LLM-based tools can transcribe a call and produce a structured summary before the agent has taken their headset off. The agent’s job changes from writing the note to checking it, turning a lengthy typing exercise into a short review.
In many operations this is the intervention I would examine first. But there are two important caveats.
The summary must be checkable. If agents can’t quickly see whether the AI got it right, they’ll either rubber-stamp errors — a compliance risk — or rewrite everything from scratch, saving no time at all. The review step needs to be designed, not bolted on.
Quality varies enormously by configuration. A generic summarisation prompt produces generic summaries. The system needs to know what your operation considers a good note: which fields matter, what your compliance team needs to see, and what a downstream colleague needs to pick the case up cold. That’s prompt-design work, and it’s where many off-the-shelf deployments fall short.
2. Kill the duplicate data entry
A large share of ACW isn’t note-taking — it’s re-keying. The same customer detail is entered into the CRM, then the workflow tool, then the ticketing system. Integration and robotic process automation are less glamorous than AI, but if your agents are copying and pasting between systems, fixing that may beat any AI tool on pure return on investment.
During Covid, I built an automation in 24 hours that captured and logged critical account actions that had previously been done manually. It wasn’t clever. It was a form that wrote to the right places. It saved hours a day and eventually justified a dedicated role.
Boring automation is often the best automation.
3. Fix your disposition codes before you automate them
If your reason-for-call list is long and agents regularly fall back to “General enquiry”, no AI will fix the underlying problem — it will simply automate the same weak classification. Rationalise the list against real demand first, then let AI suggest the most appropriate code from the transcript with the agent confirming it. Suggestion plus confirmation can be faster than manual selection and safer than full automation.
4. Give guidance during the call, not after it
Some wrap-up work exists because the agent couldn’t complete it during the call — they were hunting for information or holding details in their head to enter later. Agent-assist tools that surface the right knowledge article or auto-fill fields mid-call can shrink the after-call backlog before it forms.
What doesn’t work — and quietly makes things worse
Setting an ACW target and pushing it down. If you simply tell agents to wrap up faster, they will — by writing worse notes. Six months later your complaints team can’t reconstruct what happened on a call, QA scores drift and regulated record-keeping has holes in it. You haven’t reduced the work; you’ve deferred it to someone more expensive.
Removing wrap-up time entirely through automatic call delivery. Back-to-back calls with zero buffer are how you manufacture attrition. Remember recovery time? Agents need it, especially after vulnerable-customer or complaint calls. The operations with the best sustained ACW numbers I’ve seen protect a small human buffer deliberately — and their attrition and quality figures are the payoff.
Buying a tool without redesigning the process. An AI summary tool layered on top of three disconnected systems and a bloated disposition list saves a fraction of what it should. The tool is only part of the result; the process around it determines whether the saving survives.
The agent-trust problem nobody budgets for
Here’s the thing I saw over and over from the inside: frontline teams don’t reject tools because they fear technology. They reject tools that make their job harder while claiming to make it easier — the summary that’s wrong just often enough to be dangerous, the “assistant” that’s really a surveillance dashboard, or the new system that adds a login without removing one.
If agents don’t trust the summaries, they rewrite them, and your business case evaporates silently. The deployments that work involve agents early, show them exactly what the AI does with their calls, give them a fast way to correct it and — critically — hand the time saved back as breathing room and better work, not just a higher call quota.
That last point is a leadership decision, not a technology one. Make it deliberately.
Where to start: a practical sequence
- Measure the real split. Spend two weeks listening to calls and shadowing agents. Find out where ACW actually goes.
- Fix the plumbing. Address duplicate entry and disposition bloat first. These changes are usually cheaper, faster and make every later step more effective.
- Pilot automated summaries with one team. Co-design the summary format with agents and your QA and compliance people, then measure both time saved and note quality.
- Expand on evidence. Roll out when the pilot numbers hold, not when a vendor’s slide deck says so.
- Reinvest some of the saving in your people. Protect recovery time, then monitor attrition, wellbeing and quality alongside productivity.
In my experience, a well-run programme can deliver a meaningful reduction in ACW — sometimes 30–50% — while note quality improves rather than declines. The outcome depends on the starting point, what the operation counts as ACW and how much of the surrounding process is redesigned. Treat larger headline claims carefully and check exactly what is being measured.
The honest summary
Wrap-up time is one of the contact centre problems where the technology is genuinely useful, the return is measurable and the main failure mode is avoidable — because that failure mode is ignoring the people who do the work.
Get the process right, involve your agents and deploy summaries that compliance can stand behind, and this can be one of the most reliable operational improvements available today.