Average handle time (AHT): formula and AI-era cohorts
AHT is the average time a contact takes from first word to finished wrap-up. It is the input every staffing model depends on, and it is easy to distort once an AI agent handles part of the queue.
The formula
AHT = (talk + hold + after-contact work) ÷ contacts handled
Use contacts handled, not contacts offered: abandoned calls have no handle time. Keep hold and wrap-up in the numerator. Leaving them out is the most common reason AHT looks better than staffing reality.
Worked example
75 contacts with 21,000 s of talk, 3,000 s of hold and 6,000 s of wrap-up: AHT = 30,000 ÷ 75 = 6:40 (400 s), of which 20% is wrap-up.
Try it
AHT after an AI rollout
When an AI agent resolves easy contacts on its own, a single AHT figure mixes very different work. ahtByCohort reports AI-only, handoff and human-only contacts separately.
Example: 400 AI-only contacts, 100 handoffs and 500 human-only contacts, against a 380 s baseline.
- Blended AHT across all 1,000 contacts: 264 s, 31% below baseline. This is the number that gets put on slides.
- Human AHT on the 600 contacts a person touched: 370 s, only 2.6% lower. Staffing should use this number.
Plugging the blended figure into Erlang C would understaff the human queue.
Usage
import { ahtFromTotals, ahtByCohort, mmss } from 'contact-center-math/aht'
const { aht } = ahtFromTotals({ talk: 21000, hold: 3000, acw: 6000, contacts: 75 })
mmss(aht) // "6:40"