Average Handle Time (AHT): Formula, What to Include, and Why It Rises When AI Takes the Easy Contacts

Average handle time (AHT) is the average time an agent spends on one contact, from the moment it starts to the moment the agent is free for the next one. The standard formula is:

AHT = (Total talk time + Total hold time + Total after-call work) ÷ Number of contacts handled

If a team handles 500 calls in a week with 2,500 minutes of talk, 500 minutes of hold and 500 minutes of after-call work, AHT is (2,500 + 500 + 500) ÷ 500 = 7 minutes.

AHT is the single most important input to a staffing calculation after volume, so small errors in it turn into real headcount errors. This guide covers the formula, what to include, what 30 seconds of AHT costs, and a trend that is catching many teams out in 2026: human AHT rises when AI takes the easy contacts, even when the operation as a whole is getting more efficient.

What goes into AHT

Three parts make up handle time:

• Talk time: the live conversation with the customer.
• Hold time: time the customer waits while the agent looks something up or consults a colleague.
• After-call work (ACW): wrap-up such as notes, CRM updates, disposition codes and tagging.

Leaving out after-call work is the most common definition mistake. It makes AHT look lower than the real workload and hides the documentation effort, so your staffing plan comes out short. Decide your definition once, write it down and apply it everywhere. If your phone system reports talk and hold but not wrap-up, add ACW from the agent state data before you plan.

Two related measures are worth separating:

• Handle time vs talk time. Talk time is only one part. Plan with handle time.
• AHT vs cost per contact. AHT measures effort, not outcome. A fast call that does not resolve the issue comes back as a repeat contact, which adds workload. See Why Chasing a Lower Average Handle Time Might Be Costing You More Agents and Cost Per Contact Is Misleading.

What good AHT looks like

Published benchmarks vary widely, and most are not tied to a stated source. Common ranges quoted for phone support sit somewhere around 3 to 10 minutes depending on the industry, with retail and ecommerce at the short end, and technical support and complex healthcare at the long end. Treat any figure as a rough direction, not a target. One knowledge hub notes that no authoritative benchmark by industry exists, because call complexity, channel and regulatory load change what a good number is. Your own trend by contact type is a far better guide than an industry average.

Note also that average hides the spread. Two teams with a 6 minute AHT can have very different workloads if one has many 2 minute calls and a few 25 minute ones. Look at AHT by contact type and by interval before you plan.

What 30 seconds of AHT costs

Here is a standard Erlang C example. It is illustrative, with no shrinkage and no abandonment.

• 300 calls in a 30 minute interval
• 80/20 service level target (80% answered within 20 seconds)

Agents on the phones needed at different AHT values:

Every extra 30 seconds of AHT adds about 5 to 6 agents here, which is roughly 10% more headcount for a 10% longer call. Add shrinkage on top: see how to calculate shrinkage in a BPO for how to gross up properly. You can test your own numbers in our free Erlang C Calculator and see the service level effect with the Service Level Calculator.

The reverse is true too. A 30 second AHT reduction that does not hurt resolution saves a similar number of agents, which is why AHT projects attract so much attention. The risk is a rushed call that comes back as a second contact.

Why human AHT rises when AI takes the easy contacts

As bots and AI agents resolve the simple contacts, the contacts that reach people are the harder ones: longer, more emotional and more multi-step. Human AHT goes up even when the operation as a whole is more efficient. Nothing has gone wrong. The mix changed.

An illustrative example, with all figures made up for the example:

• Before: 300 calls in an interval. 120 are simple calls (3 minutes each) and 180 are complex (8 minutes each). AHT is (120 × 3 + 180 × 8) ÷ 300 = 6.0 minutes. Required agents at 80/20: about 67.
• After: AI resolves 90 of the 120 simple calls. People now get 210 calls: 30 simple and 180 complex. AHT becomes (30 × 3 + 180 × 8) ÷ 210 = about 7.3 minutes. Required agents: about 58.

Volume to people fell by 30%, human AHT rose by about 21%, and required agents fell by only about 13%. A report that shows "AHT up 21%" would look like a performance problem, but the real picture is less workload with harder work. If you plan headcount on volume alone, you will cut too deep. If you judge agents on the old AHT target, you will push them to rush the complex calls.

This is why we recommend:

1. Track AHT by contact type, not only as one blended number.
2. Set separate targets for AI and human handling, paired with a first contact resolution floor so that nobody rushes.
3. Re-baseline AHT after each automation change, because history describes work that no longer exists. See Why Your AI Deflection Numbers Are Quietly Breaking Your Staffing Model.
4. Watch occupancy. Longer, harder calls are more tiring. See The Occupancy and Retention Link.

Some writers have gone further and proposed a blended "net AHT" that weights human and bot time together, to show the hybrid picture and compare cost per contact across human and AI handling. It is an interesting idea for reporting, but for staffing the human workload is what you must plan, so keep human AHT, by contact type, as your planning input.

Common AHT mistakes

1. Leaving out after-call work. Your plan is short by the wrap-up time.
2. One blended AHT for every channel. Voice, chat and email behave differently. For messaging, see Async Messaging Staffing: Which Handle Time to Plan With.
3. Planning on last year's AHT. New products, new policies and new automation all change it.
4. Using the average for a skewed mix. Plan on the mix of contact types.
5. Pushing AHT down without watching repeat contacts.
6. Ignoring new-joiner AHT. Newer agents are slower. Plan the AHT of your actual tenure mix.

Quick checklist

• Define AHT as talk plus hold plus after-call work, in writing.
• Calculate by contact type and by interval.
• Re-baseline after any automation or process change.
• Pair AHT targets with a resolution measure.
• Check headcount impact with an Erlang C calculation before you set a target.

Want a quick, printable view of where your planning is weakest? Try the free WFM Health Check.

FAQ

What is the formula for average handle time? AHT = (Total talk time + Total hold time + Total after-call work) ÷ Number of contacts handled.

Does AHT include after-call work? Yes. Leaving it out understates the real workload and leaves your plan short.

What is a good AHT? It depends on the industry, channel and call complexity. Published ranges are only a rough guide, so track your own trend by contact type.

How does AHT affect staffing? Directly. In the example above, every extra 30 seconds of AHT needs roughly 5 to 6 more agents on the phones at 300 calls per half hour.

Why is my AHT rising after we added AI? Usually because the simple, short contacts now go to the AI, so the contacts that reach people are longer on average. Look at AHT by contact type before treating it as a performance issue.

What is the difference between AHT and talk time? Talk time is just the conversation. AHT also includes hold time and after-call work.

Sources: Parloa: Average handle time, Featurebase: Average Handle Time: Formula, Benchmarks and How to Reduce It and CustomerThink: Rethinking AHT: A New Formula for AI-Hybrid Contact Centers. The staffing and mix examples use the standard Erlang C formula and invented figures calculated by the WFM Helper team.