After call work (ACW) is the work an agent does on a call after the caller hangs up: choosing a disposition code, writing notes, updating the CRM or ticket and setting up any follow-up that was promised. It is also called wrap-up time. ACW counts toward average handle time, so staffing math treats every second of it the same way it treats talk time.
This guide covers the settings that decide how ACW is timed in Amazon Connect and Genesys Cloud, a check for after-call work that never reaches the ACW number, a worksheet for deciding which tasks stay, a starter list of disposition codes and what 15 seconds of ACW costs in agents.
What is after call work in a call center?
After call work is the time an agent spends finishing a call once the caller is gone. The call center research literature defines wrap time as "the average amount of time a CSR spends completing service, after the caller has hung up," where a CSR is a customer service representative (Gans, Koole and Mandelbaum, 2003). Their examples of wrap-up are updating the customer's history file and processing an order the customer asked for.
Platforms use their own names for it. Amazon Connect calls it after contact work, and its metric definitions note that some businesses know it as call wrap up time (Amazon Connect metric definitions). Genesys Cloud calls it after call work and ties it to wrap-up codes. In a BPO the term means the same thing as in an in-house center.
While an agent is in ACW, the routing system sends them nothing new. Genesys describes its ACW timeout as the time agents have for after call work "before they receive more interactions" (Genesys Cloud after call work settings). That makes ACW a staffing number as well as an admin one, because an agent in wrap is as unavailable to the queue as an agent on a call.
How do contact center platforms time after call work?
Two settings decide how ACW behaves: whether an agent has to go through it, and how long it can last. Amazon Connect sets the time limit for each agent, while Genesys Cloud sets it for each queue and offers five modes. The table puts the two side by side, from each vendor's own documentation.
| Setting | Amazon Connect | Genesys Cloud |
|---|---|---|
| Where the time limit lives | Each agent's user settings, under Contact handling, as the After Contact timeout | Each queue, as After Call Work Timeout (Seconds) |
| Allowed range | 1 second to 2,000,000 seconds | Up to 3,600 seconds |
| Modes | One timeout per agent; Outbound Campaigns calls and customer-first callbacks take theirs from the Voice settings | Optional; Mandatory, Discretionary; Mandatory, Time-boxed; Mandatory, Time-boxed no early exit; Agent Requested |
| When the time runs out | The agent's ACW lasts no longer than the timeout | The agent becomes available, and a call with no wrap-up code gets the default code ININ-WRAP-UP-TIMEOUT |
The Amazon Connect settings are on its agent settings page, and the averages appear as Avg ACW in its real-time reports and as Average after contact work time and Average outbound after contact work time in its historical reports. The five Genesys Cloud modes, in plain terms:
- Optional: agents may skip the wrap-up code. Genesys suggests it for organizations that don't use wrap-up codes and whose agents don't need ACW time.
- Mandatory, Discretionary: agents stay in ACW as long as they choose and set themselves available when they're done.
- Mandatory, Time-boxed: ACW has a ceiling. Agents who finish early can go available, and at the ceiling they become available automatically.
- Mandatory, Time-boxed no early exit: agents stay in ACW for the full time even when they finish early, which Genesys describes as a "cool down" period between interactions.
- Agent Requested: agents ask for ACW before they disconnect, and those who don't ask go straight back to idle. Genesys recommends against it for outbound dialing, because skipped after call work can make campaigns perform inaccurately.
Pick the mode from the work. Discretionary has no ceiling, so watch the spread by agent and not only the average. A time box helps only when it is set from the measured task time; set it shorter than the work and the work moves somewhere else, which the next section shows. No early exit spends a fixed number of seconds on every call, a deliberate trade of agents for rest between calls. Optional suits a line where nothing reads the codes.
Why is my ACW lower than the real after-call work?
ACW measures time spent in the ACW state, and the work can leave that state. When the time box runs out before the notes are done, agents finish them in a not-ready status, during the next call or after the shift. The ACW average falls while the amount of work stays exactly where it was, and the staffing forecast inherits the lower number.
Take one agent's hour on a queue set to time-boxed ACW of 45 seconds. The agent takes 12 calls. On 8 of them the notes are done in 30 seconds and the agent goes available early. On the other 4 the timer runs out first, the platform applies its default timeout code, and the agent switches to a not-ready status for another minute to finish.
| Calls | Time in ACW | Not-ready time afterward | Code on the call |
|---|---|---|---|
| 8 calls, notes done inside ACW | 30 seconds each | None | The one the agent chose |
| 4 calls, timer ran out first | 45 seconds each | 60 seconds each | The default timeout code |
That hour gives three different readings:
- Reported average ACW: (8 × 30 + 4 × 45) ÷ 12 = 35 seconds.
- Actual after-call work: (8 × 30 + 4 × 105) ÷ 12 = 55 seconds.
- Calls with a real disposition: 8 of 12. The other 4 carry the timeout code, so a third of the hour is missing from the outcome report.
The 240 seconds of not-ready time still appear somewhere. They show up as lost availability in schedule adherence, where they look like an agent problem when the cause is a setting. Meanwhile the 20 seconds missing from every call's handle time go into the staffing forecast, which comes out short.
To check your own center:
- Export one day of agent state history for one team.
- For each ACW period, add any not-ready time that starts the moment ACW ends, before the agent goes available again.
- Count the calls closed with your platform's timeout code or with no code at all.
- Sit beside two agents for 20 calls each and time every after-call task with a stopwatch.
If steps 2 and 4 come out close to the reported ACW average, the number is honest. If they don't, raise the time box or cut tasks before anyone sets a target on it. Twenty calls per agent is a rule of thumb for a first look, not a statistical standard.
Which tasks belong in after call work?
The tasks that belong are the ones somebody reads or acts on later. Most ACW is made of a handful of repeating tasks: the disposition code, notes, CRM or ticket fields, a follow-up and sometimes processing the request itself. List them for one call type, time each one, and write down who uses the result. A task with no reader is the first one to cut.
| After-call task | Who uses it later | Usual fix |
|---|---|---|
| Disposition (wrap-up) code | Outcome reports, outbound list rules, workflows | Keep it, shorten the list, and let agents pick it on the call screen before the caller hangs up |
| Free-text notes | The next agent who takes this customer's call | Turn the sentences people always write into fields or a template, and use an automatic summary where the platform writes one |
| CRM or ticket fields | Other teams and reporting | Capture them during the call, on the same screen, prefilled from the caller's record |
| Follow-up (callback, email, ticket to another team) | The customer and the team that acts next | Keep it, and create it from the disposition code instead of by hand |
| Processing the request (an order, a refund, an address change) | The customer | It is part of serving the caller, so do it on the call where you can, and send it to a back-office queue when it needs no phone skills |
| Verification or consent note | Audit | Make it a required field with set answers instead of free text |
| Waiting on a slow screen or save | Nobody | Time it separately, since the fix is in the system rather than the process |
Fill it in one call type at a time, because a billing call and a password reset wrap up differently. Put the median seconds per task next to each row, then start with the rows that have the most seconds and no reader.
How should disposition codes be set up?
A disposition code answers one question: how did the call end? Keep the list short enough to read at a glance, make the codes mutually exclusive, and give each one a consumer, meaning a report or a workflow that reads it. Record the reason for the call in a separate field, so outcome and reason don't get crammed into one list.
A starter set for an inbound line:
| Code | Use it when | What it should trigger |
|---|---|---|
| Resolved on the call | The request was finished before the caller hung up | Nothing |
| Transferred | The call went to another team or number | Nothing, if the receiving team codes its own part of the call |
| Callback scheduled | A time was agreed to call the customer back | The callback itself, created from the code |
| Follow-up owed | Something was promised that isn't done yet | A task with an owner and a due date |
| Escalation or complaint | The caller asked for a supervisor or raised a complaint | The complaint workflow |
| Not for us | Wrong number, wrong department or an unsolicited sales call | Nothing, though a rising count points to a routing problem |
| Disconnected before done | The call dropped or the caller hung up mid-request | A review of recordings if the count climbs |
Outbound programs add their own codes: reached and completed, declined, voicemail left, no answer, wrong number, and a request to stop calling, which should take the number off every future list.
Two numbers tell you whether the list works. The first is the share of calls coded Other, or closed with the timeout code; when it climbs, agents either can't find the right code or don't have time to pick one. The second is one code taking most of the calls, which usually means it is hiding several outcomes that need codes of their own.
If an IVR or an AI agent takes part of the volume, map its outcome values onto the same list, so one report covers every call no matter who took it.
What does after call work cost in staff?
ACW is part of handle time, so queueing math staffs it exactly like talk time. In the example below, with 240 calls in a half hour, 4 minutes 30 seconds of talk and hold per call, and a target of 80 percent answered within 20 seconds, each 15 seconds of ACW adds about two agents on the phones in that half hour, before shrinkage.
| ACW per call | Handle time | Agents needed on the phones (Erlang C) |
|---|---|---|
| 60 seconds | 5:30 | 50 |
| 45 seconds | 5:15 | 48 |
| 30 seconds | 5:00 | 46 |
| None | 4:30 | 42 |
Our Erlang C calculator runs the same math for your volume and handle time. Erlang C assumes nobody abandons, so treat the result as an approximation, and add shrinkage on top to get from agents on the phones to agents on the schedule, which the workforce management guide works through.
For an outside reference point, the same 2003 review describes a large US catalog retailer whose 10:00 to 11:00am peak ran a service time of about 3.75 minutes with 30 seconds of after-call work (Gans, Koole and Mandelbaum, 2003). That is one dated operation with its own call types, so the task time you measure on your own calls makes a better target.
How do you reduce after call work?
Remove work before you shorten the timer. Cutting the time box alone pushes the same work into not-ready time, as the example above shows. These are the changes that take seconds out, in the order we would try them:
- Drop the tasks nobody reads, using the worksheet above.
- Move data capture into the call, on the screen the agent already has open.
- Replace repeated free text with fields, templates or an automatic summary.
- Create callbacks and follow-up tasks from the disposition code.
- Send processing that needs no phone skills to a back-office queue.
- Fix slow screens and saves.
- Set the time box from the task time you measured.
Software that drafts the summary and suggests a disposition code after a human call is a category of its own, covered in the call center automation guide. If you adopt it, check a sample of its summaries and codes against the recordings every week, the way you would score calls in quality assurance, since a wrong code is harder to spot in a report than a missing one.
What happens to after call work when AI agents take calls?
Calls an AI agent completes on its own need no wrap from your team, because the platform writes the record when the call ends. Calls it transfers to a person still get that person's normal ACW. Teams usually move the shorter, high-volume call types to an AI agent first, so the calls left for people run longer, and ACW per human call can rise for the same reason handle time does.
Keep ACW split by call type before and after the change, the way the average handle time guide recommends for handle time. Map the AI agent's outcome tags to your wrap-up codes before it takes a live call, so the outcome report still adds up, and read a sample of its summaries against the recordings as you would for any automatic summary. What AI agents handle in a call center covers which call types usually go first, and the call center metrics guide covers the other numbers that shift when part of the volume moves.
What a Telvana agent leaves behind after each call
Every call a Telvana agent handles is recorded, transcribed, summarized and tagged by outcome, with no wrap-up time from your team. Turn on the post-conversation webhook and, after each call, your system receives the transcript, summary, sentiment, tags and a disposition of COMPLETED, VOICEMAIL, TRANSFERRED, APPOINTMENT_BOOKED, SCHEDULED_CALLBACK, ABANDONED or FAILED (webhooks). Those values map onto a wrap-up code list like the one above. With HubSpot connected, the call summary is added as a note on the matching contact after every call.
A transfer is an ordinary call to the number you choose, so the person who picks up does their own wrap-up as usual and can read the agent's transcript and summary in the portal afterward. Our team builds the agent from your scripts, FAQs, policies and transfer rules and tests it on your call types before it takes a live call. Pricing starts at $399 a month based on call volume, plus a one-time setup fee (pricing).
Frequently asked questions
What does ACW mean in a call center?
ACW stands for after call work: the time an agent spends finishing a call after the caller hangs up, on tasks such as the disposition code, notes, CRM updates and follow-up. It counts toward average handle time, and while an agent is in ACW the routing system sends them no new calls.
What does ACW mean in a BPO?
It means the same as in an in-house center: the agent's post-call work on disposition codes, notes, system updates and follow-up. Measure it for each client program separately, because each program has its own wrap-up tasks and code list.
What is wrap-up time?
Wrap-up time is another name for after call work. Wrapping up a call means finishing everything the call requires once the caller has gone, such as choosing the wrap-up code, writing notes and setting up any follow-up.
How do you calculate average after call work time?
Divide the total time agents spent in after call work by the number of calls handled in the same period. Check which calls your platform counts in that denominator, and if agents finish notes in a not-ready status after ACW ends, add that time back or the average will read low.
How long should after call work take?
There is no standard figure. Set it from how long the tasks take on each call type, measured by timing a sample of real calls. As one dated reference, a 2003 review describes a catalog retailer's peak hour with about 30 seconds of after-call work against 3.75 minutes of service time.
What happens when the after call work timer runs out?
It depends on the platform. In Genesys Cloud, an agent on a time-boxed queue becomes available at the end of the timeout, and a call with no wrap-up code gets the default code ININ-WRAP-UP-TIMEOUT. Frequent timeouts usually mean the time box is shorter than the tasks.
What is the difference between after call work and not-ready time?
After call work belongs to a call and counts in handle time. Not-ready or auxiliary time is any other time an agent is logged in but unavailable, such as breaks, training or meetings. When agents finish call notes in a not-ready status, that work drops out of handle time and shows up as lost availability instead.
Bring your wrap-up codes to a demo
Book a demo and bring your wrap-up code list along with your top call types. We'll show which calls an agent could take and how its outcome tags line up with your codes, and you can hear it take a call.