The short version
- AHT = (talk time + hold time + after-call work) / calls handled.
- It's a staffing and cost metric first. Used as an agent target, it often makes service worse.
- Automation usually raises AHT for the remaining human calls, because the easy calls leave the queue. That's expected.
Average handle time is one of the oldest numbers in a call center, and one of the most misused. It's essential for planning how many people you need. As a target for individual agents, it can quietly push them to rush customers off the phone and create repeat calls.
This guide covers how AHT is calculated, what drives it, how automation changes it, and how to use it without hurting your customers.
The formula
AHT = (total talk time + total hold time + total after-call work) / number of calls handled
- Talk time: time the agent is connected and speaking with the customer
- Hold time: time the customer is on hold during the call (not time in the queue before an agent answers)
- After-call work (ACW): wrap-up after the customer hangs up, such as notes, dispositions, follow-up emails and form updates
Example: in one day, a team handles 400 calls with 1,600 minutes of talk time, 200 minutes of hold and 600 minutes of wrap-up.
AHT = (1,600 + 200 + 600) / 400 = 6 minutes.
Some teams exclude ACW or measure it separately. That's fine as long as you're consistent and everyone knows which version they're looking at.
What AHT is actually for
AHT's main job is capacity planning. Workforce management uses it along with call volume and your service level target to calculate how many agents you need on shift. If AHT goes up by a minute and volume stays the same, you need more people to answer the same number of calls in the same time.
It's also a cost input. Multiply AHT by your fully loaded cost per agent minute and you have the labor cost of an average call. We build that into a fuller model in cost per call.
What drives AHT up
When AHT rises, the cause is usually one of these:
- Call mix shifted. More complex calls (new product launch, billing change, outage) take longer. This isn't a performance problem.
- Slow or clumsy systems. Agents waiting on screens to load, switching between five tools, or searching for information that should be one click away.
- Missing information at the start. If the IVR or agent didn't capture the account number or reason, the person spends the first minute asking.
- Long wrap-up. Free-text notes, manual dispositions and duplicate data entry pad ACW.
- New staff. Ramp time is normal. Compare new hires against other new hires, not against five-year veterans.
- Transfers. Every transfer adds a greeting, a re-explanation and often a second hold.
When chasing AHT backfires
Put a hard AHT target on agents and they'll hit it. Some of the ways they'll do it:
- Ending calls before the problem is fully solved
- Transferring complex calls instead of handling them
- Skipping notes, which makes the next call about the same issue longer
- Avoiding the questions that would uncover a second problem
All of these lower AHT and raise repeat calls. The customer calls back, so your total time spent on that customer goes up, and their opinion of you goes down.
This is why most experienced operations leads pair AHT with first call resolution. Short calls that resolve the issue are good. Short calls that cause a callback are just deferred work.
A better approach for agents: track AHT at the team level for planning, look at outliers to find coaching opportunities and broken processes, and set individual goals on outcomes like resolution and quality scores.
How automation changes AHT
This catches people out. When you add self-service or an AI voice agent, the calls that get automated first are the quick, simple ones: order status, appointment changes, hours and directions. Those calls had low handle times.
Take them out of the human queue, and the calls left for people are, on average, longer and harder. So human AHT goes up even though the operation is more efficient overall. Total human minutes drop because there are fewer calls, but each remaining call takes longer.
If you report AHT without explaining this, it looks like the team got slower. Report these alongside it:
- Total contact volume, split by handled by agent vs. handled by automation
- Total human talk and wrap-up minutes
- AHT by call type, so you're comparing like with like
Reducing AHT the right way
Most of the real gains come from removing work, not speeding people up.
Capture the reason and identity before the call reaches a person. An agent or IVR that collects account details and a summary means the human starts at minute one, not minute zero.
Fix the tools. Time how long agents spend waiting on systems or searching. It's often the largest single chunk of avoidable time.
Automate wrap-up. Transcription and automatic call summaries can cut after-call work significantly, because the agent reviews a draft instead of typing notes from memory.
Write better knowledge articles. If agents put customers on hold to look things up, the answers are too hard to find.
Route by skill. Sending billing calls to people who know billing shortens every one of those calls.
Benchmarks
You'll find AHT benchmarks for every industry online. Treat them with suspicion. AHT depends heavily on what you count, your call mix, and how much you've already automated. A support line for complex software and a pizza order line will never look alike. Your own trend over time, broken down by call type, is far more useful than someone else's average.
Breaking AHT down by component
A single AHT number hides a lot. Splitting it into its parts tells you where the time actually goes.
Talk time reflects the conversation itself. Long talk time can mean complex issues, agents who struggle to control the call, or customers who need lots of explanation because something upstream (a bill, a product, a policy) is confusing.
Hold time usually points to agents looking for information or waiting for help. High hold time is a strong hint that the knowledge base, tools or escalation paths need work.
After-call work reflects documentation and follow-up. High ACW often means manual data entry, clunky dispositions or free-text notes that could be automated.
Track all three separately, by call type and over time. A rise in hold time after a policy change is a very different problem from a rise in talk time after a product launch.
A worked example
Consider a support team that handles four main call types. Using illustrative figures:
| Call type | Share of calls | Avg talk | Avg hold | Avg ACW | AHT |
|---|---|---|---|---|---|
| Order status | 35% | 2.5 min | 0.2 min | 0.8 min | 3.5 min |
| Returns | 25% | 5.0 min | 0.8 min | 1.7 min | 7.5 min |
| Billing | 20% | 6.0 min | 1.5 min | 2.0 min | 9.5 min |
| Technical | 20% | 9.0 min | 2.0 min | 2.5 min | 13.5 min |
The blended AHT works out to a little over 7.6 minutes. Now suppose an AI agent takes over most order status calls. The calls left for people are returns, billing and technical, and the blended human AHT rises to around 10 minutes, even though nobody got slower and total human minutes fell substantially.
This is exactly the situation where a team gets blamed for "worse AHT" after a successful automation project. Reporting by call type avoids the confusion.
Coaching with AHT, without the side effects
AHT still has a place in coaching, used carefully:
- Look at outliers in both directions. An agent whose calls are much shorter than peers on the same call type may be cutting corners. One whose calls are much longer may need help with tools or call control.
- Listen to the calls. Never coach on AHT from the number alone. The recording or transcript shows whether the time was well spent.
- Pair it with resolution and quality. A coaching conversation that only mentions speed teaches people to rush.
- Fix systemic causes first. If everyone's hold time is high on billing calls, that's a knowledge or tools problem, not a coaching problem.
AHT for AI agents
If you deploy AI voice agents, measure their handle time too, but read it differently. For an agent, longer calls usually mean more turns: the agent asked for repeats, the caller was confused, or a lookup failed and had to be retried. Track turns per call and handle time by call type, and investigate when either drifts upward. Short agent calls that end in hang-ups are a warning sign, not a success.
Frequently asked questions
Does AHT include time in the queue?
No. Queue time is part of speed of answer or wait time. AHT starts when an agent answers.
Should after-call work be included?
Usually yes, because it's time the agent can't take another call. Some centers report it separately so they can target it directly.
Is a lower AHT always better?
No. A lower AHT that comes from unresolved calls creates repeat contacts. Look at AHT together with resolution and customer feedback.
Written by the Voxvencer editorial team. We build and run AI voice agents for call centers and small businesses, and we write about what we see on real phone lines. Questions or corrections: info@voxvencer.com.
