- 01
Market
Saudi Arabia
- 02
Written for
COO, Head of CX
- 03
Decision
Reduce contact centre queue time without losing service quality
- 04
Languages
EN · AR
Selection
What makes a call reason a good candidate
Judged on three properties. A reason that fails the third should be automated last however high its volume, because that is where an automated mistake is expensive.
| Option | Volume | Does a system already hold the answer? | Cost of being wrong | Best for |
|---|---|---|---|---|
| Status and balance enquiries | Usually the largest single reason. | Yes — one lookup after identity is established. | Low. A wrong answer is corrected in the same call. | Automate first. This is where the queue time is. |
| Booking, rescheduling, cancelling | High, and heavily peaked at predictable times. | Yes, if the calendar can be written to safely. | Moderate — a bad write shows up as somebody arriving to nothing. | Automate second, with slot holds in place. |
| Complaints and disputes | Low volume, high handling time. | No. The answer requires judgement and often discretion. | High, and paid in the relationship rather than in the call. | Route, do not automate. Give the agent the context instead. |
Containment
The metric that can be gamed by making service worse
Containment counts calls that ended without a person. It rises when the agent is good and it also rises when escalation is made difficult, and those two look identical on a dashboard. A deployment optimised on containment alone will eventually hide the handover option, and the callers who needed it will simply hang up — which also counts as contained.
Read it against repeat contact within seventy-two hours and against satisfaction on the automated reasons specifically. A contained call the customer had to make again two days later cost more than a transfer would have, and it is invisible unless the two numbers are read together.
Sequence
How a rollout that survives contact with the floor runs
Read what the calls are actually about
Dispositions are usually wrong or unused; transcripts are not. Two weeks of transcripts clustered by reason gives a distribution nobody disputes. A ranked list with volume and handling time against each.
Automate one reason, completely
Including the exception path and the out-of-hours case. A reason handled ninety per cent of the way generates more work than it removes. One reason off the queue, with its own numbers.
Run in parallel before switching
A share of traffic, with agents able to take over live. The failures found here are the ones that would have been incidents. Evidence rather than confidence.
Give the agents what the system learned
Every handover arrives with the transcript and the identified intent. This is the part that makes floor staff advocates rather than sceptics. Shorter handled calls on the reasons that stay human.
Our role
What we take on in a contact centre programme
- 01
The analysis that picks the target
Clustering real transcripts into reasons, with volume and handling time, so the first automation is chosen from evidence rather than from whoever complained loudest.
- 02
The build, including the exception path
Integration into the telephony platform and the systems of record, with the escalation route built and tested before the happy path is polished.
- 03
Instrumentation that cannot be gamed
Containment, handover rate, repeat contact and satisfaction reported together, because any one of them alone can be improved by making the service worse.
Staffing levels and service targets stay a management decision; what we provide is a defensible read of which calls a system should be taking and what happens to the rest.
Common questions
What buyers ask
What containment rate should we expect?+
It depends almost entirely on which reasons you automate, so a number quoted before the call analysis is a number about somebody else's contact centre. A single well-chosen reason with a system-held answer contains most of its own volume; a broad deployment across every reason contains far less and annoys more people doing it.
Do we need to replace our telephony platform?+
Usually not. Most contemporary platforms expose the integration points an agent needs — media streaming, transfer with attached data, and a way to record the interaction. Where a platform genuinely cannot hand over with context, that limitation is worth surfacing early, because a transfer that drops the conversation undoes most of the benefit.
How do agents react to this?+
Better than expected when the handover carries context, because the calls they receive arrive half-solved, and worse than expected when it does not, because they receive an annoyed caller repeating themselves. The floor's reaction is a good early indicator of whether the deployment is designed properly.
In this cluster
Next step
Pilot it on one queue
One call reason automated end to end, measured against the queue it came from, before a second one is scoped.

