Reduce average handle time by removing avoidable searching, waiting, repeated explanation and duplicate work before shortening necessary diagnosis. Start with a comparable baseline, identify the work a targeted change should remove and test the result against same-issue repeat contacts and resolution quality.
AHT remains essential for staffing and capacity planning. But its components matter: a successful change can give agents more time to understand the issue while reducing holds and wrap-up enough to lower total handling time.
Establish the baseline before setting a lower AHT target
Separate talk time, hold time and after-call work in the baseline. Those components point to different operating problems and different owners.
For a voice baseline that includes these three components, calculate:
AHT = (total talk time + total hold time + total after-call work) ÷ handled calls
For example, take 600 minutes of talk time, 150 minutes of hold time and 250 minutes of after-call work across 100 handled calls. The total is 1,000 minutes, producing an AHT of 10 minutes. Per call, that means six minutes talking, 1.5 minutes on hold and 2.5 minutes completing records or other wrap-up work.
Reconcile this calculation with your reporting system’s definitions before using it for comparisons. Check how the system treats consultation time, transfers and separate call legs. Confirm which work enters the numerator and which calls enter the denominator. Counting the handling effort of two agents against only one agent’s handled-call population produces a misleading comparison. Queue waiting before an agent answers is separate from handling time.
Build the baseline by contact reason and complexity. Compare similar work before comparing agents or queues. A queue handling exceptions should not inherit the target of one handling straightforward account changes. Retain the component averages and inspect the spread of durations so a few unusually long calls do not become the entire improvement agenda.
Set a provisional reduction target from identified removable effort, such as repeated approval holds, rather than borrowing a universal benchmark. Keep this calculation specific to voice: concurrent chat and asynchronous service require their own definitions of active work and elapsed time.
Define the repeat contacts the improvement must not create
Define same-issue repeat contacts before changing the workflow. Otherwise, a favorable AHT result can influence which returns the team chooses to count.
Link contacts using customer or case identity plus a stable issue classification. Choose a follow-up window that fits the issue lifecycle, including when customers can reasonably discover that the proposed action failed. Apply the same window to the baseline and pilot.
For the pilot, a practical case-level measure is:
Eligible initial cases with at least one unplanned same-issue recontact ÷ eligible initial cases with a fully observed follow-up window
This counting convention measures the share of cases that return, not how often each case returns. Retain the number of subsequent contacts too, since one case returning repeatedly creates more work than one returning once.
Exclude unrelated inquiries from this measure. Separate planned follow-ups from unplanned returns, but keep their handling effort visible. Turning an unfinished call into a scheduled callback does not remove the remaining work. Review why unplanned contacts occurred rather than assuming every same-issue return was avoidable.
Include channel switching wherever records support it. A call followed by an email about the same unresolved issue belongs in the evaluation even if the voice report shows no repeat call. Document gaps where identity or issue linkage is incomplete.
First contact resolution and repeat-contact rate answer related but different questions. FCR asks whether the issue was resolved in the initial contact under the organization’s counting rules. Repeat-contact tracking observes subsequent behavior. No recorded return can also mean abandonment or a customer finding help elsewhere, so pair recontact data with outcome checks. Broader first call resolution guidance can support that verification without making the AHT pilot a full measurement redesign.
Review what the handling time actually contains
Duration identifies cases to inspect; it does not explain them. Review a manageable selection of typical calls, long calls and cases that generated repeat contacts within the same issue group. Include ordinary calls because a small delay repeated frequently can offer a better improvement opportunity than an exceptional call lasting much longer.
Use recordings, screen records, timestamps and workflow histories where available and permitted. Trace customer explanation, information search, holds, approvals, transfers and wrap-up. These activity labels sit underneath the reporting components: searching can occur during talk time or hold time. Do not add it again as a separate block of handling time.
For each interval, ask what it accomplished and what caused it. A diagnostic question that distinguishes between two possible faults advances the issue. Asking again for information already established because it was missing from the handoff record reconstructs previous work. Both consume talk time, but they need different management responses.
Review the sequence with agents. Ask which search terms failed, why approval was needed and which fields they entered twice. Workflow records show the delay; agents often explain the constraint behind it. Also challenge unnecessary explanations or unfocused questioning rather than treating every long conversation as justified.
Prioritize recurring causes by frequency, realistically removable effort and resolution risk. The best first intervention is a common problem with a clear owner and a plausible fix, not necessarily the largest single source of time.
Choose changes that remove a specific source of work
Choose one or a small number of changes that explain exactly which work will disappear. Assign ownership to the team that controls the cause rather than making agents accountable for policy or system delays.
Repeated searches belong with the knowledge owner. Improve article titles, search terms and the path to frequently used answers. Keep exception guidance current and accessible. Expect less search-related talk or hold time, then check answer accuracy so faster retrieval does not spread an incomplete answer.
Repeated explanation during handoffs belongs with operations. Define a concise handoff record covering the issue, checks completed, findings and next action. Establish a receiving-agent practice of reading it before restarting diagnosis. The expected change is less reconstruction during talk time. A longer case history is no improvement if agents cannot find the relevant facts.
Recurring approval holds belong with policy owners. Clarify what agents can decide, the evidence required and the boundary for escalation. Predict a reduction in approval-related hold time and review decision quality. If hold time stays unchanged, investigate whether the real constraint is unavailable approvers or unclear evidence requirements. If inappropriate approvals rise, the control design needs correction.
Misdirected calls belong with routing owners. Review contact classification and routing rules to remove avoidable handoffs. Check for new misroutes and inspect the complexity of calls left in each queue. Moving difficult calls elsewhere changes queue averages without necessarily reducing handling effort.
Duplicate documentation belongs with workflow owners. Identify where the same information is entered more than once and which record actually needs it. Remove redundant entry while preserving accurate notes. Expect less after-call work, but check that entry has not moved into the conversation or been left for another team.
Missing customer-side detail calls for a different remedy from internal workflow friction. Where agents repeatedly reconstruct a device condition or screen problem through verbal description, selective visual assistance is worth testing. Grypp supports this human-assisted use case as an enterprise visual layer within existing workflows. Test whether the visual context reduces repeated explanation or diagnostic uncertainty and helps the advisor choose an appropriate next action. Offer customer choice, protect sensitive information and retain an accessible nonvisual route.
Coach concise conversations that still establish the outcome
Conversational coaching should reduce repetition and ambiguity. Clarify the actual issue and intended outcome early, then use concise explanations and purposeful signposting to keep the conversation directed toward that outcome.
Asking a customer to repeat information already established adds work; checking whether the proposed action produced the intended result tests whether the interaction is ready to close. Coaching should distinguish their purpose rather than treating both as expendable talk time.
Coach the closing check around the outcome, remaining steps, ownership and expected timing. If the task is incomplete, explain what happens next and who is responsible. Preserve required identity or compliance checks, accurate notes and appropriate escalation. Assess those behaviors alongside handling speed.
Useful talk time can increase while total AHT falls. Using the earlier example, increasing average talk time from six to 6.5 minutes while reducing hold time to one minute and after-call work to 1.5 minutes produces a nine-minute AHT, down from 10 minutes.
The extra conversation is justified only by what it accomplishes. Review whether it improves diagnosis or confirmation, and whether the lower hold and wrap-up components reflect work genuinely removed. This gives coaching a more precise goal than reducing every component independently.
Pilot the change and test whether time was removed or displaced
A pilot should test a specific explanation for lower handling time. Write that explanation before launch: for example, clearer approval authority should reduce approval-related hold time while preserving decision quality. This makes the result more informative than simply asking whether aggregate AHT fell.
Start with one contact reason. Record baseline talk, hold and after-call work alongside the agreed repeat-contact measure and resolution-quality checks. Assign an owner, describe the change precisely and keep unrelated changes separate where possible. Reviewers should use the same quality criteria across comparison groups, including whether the intended action worked or the next step was appropriate.
Prefer a concurrent comparison between equivalent cases where feasible. Allocate comparable work to pilot and comparison groups, with random assignment where practical. Account for complexity, agent experience and routing differences rather than comparing a highly experienced pilot team with the rest of the center. A before-and-after comparison is easier to run but more exposed to unrelated changes in demand, policy or staffing.
If the pilot coincides with a self-service change that removes simpler contacts from the agent queue, the remaining calls can become more complex. A higher aggregate AHT would not by itself show that agent handling deteriorated. Equally, a pilot receiving easier cases can look successful without improving its workflow.
Wait until the repeat-contact window has matured before comparing cohorts. Cases opened near the end of the pilot need the same opportunity to return as baseline cases. Set the observation period and required evidence with an analyst based on contact volume, normal variation and the deterioration the operation needs to detect.
Read component changes alongside case-level recontacts and resolution quality. Use complaints and customer feedback as supporting signals. Where records permit, inspect associated transfer and follow-up effort, including planned callbacks, to identify work displaced beyond the initial call.
Agree decision rules before reviewing results:
- Scale when the predicted component improves, the cases are comparable and mature outcome evidence is strong enough to support adoption.
- Revise or extend when the mechanism is unclear, follow-up data are immature or uncertainty remains too large to judge the outcome.
- Stop or reverse when unresolved work or quality failures deteriorate beyond the locally agreed threshold, even if AHT falls.
Set those thresholds around operationally meaningful changes and acceptable uncertainty. A small pilot may be unable to detect a meaningful increase in repeat contacts. An inconclusive result requires more evidence, not a declaration of success.
Scale the process improvement, not a blanket call-time limit
Rollout should reproduce the mechanism that worked. Update the relevant knowledge guidance, approval authority, routing rule or documentation process. Give the owner responsibility for maintaining it and monitor whether the component gain and outcome safeguards persist at greater scale.
Keep necessary exceptions visible. Broader rollout can introduce different case complexity or reveal gaps that the pilot did not encounter. Use those differences to refine the process rather than forcing every interaction toward the pilot average.
If clearer approval authority reduced hold time, the durable improvement is that agents can make the right decisions without unnecessary waiting. Preserve the authority, evidence requirements and coaching that made that possible. The pilot average is evidence of the process change; the process is what leaders need to manage.
