Contact centre managers rarely approve an individual refund, yet they decide whether an AI agent deployment works. They staff the queue of held items, read the workspace for trouble, decide when to pull an agent back and answer for the outcomes. That is why they should hold a current certificate before agents go live: their decisions shape every reviewer's.
The manager's decisions are system decisions
A frontline adviser who reviews a held item makes one decision. A contact centre manager makes decisions about how hundreds of those decisions get made. Who is on the review rota today? How long can a held refund wait before a customer chases? Is the rise in refusals since Tuesday a sign the agent is working or a sign something upstream broke?
None of those decisions appear as verdicts in the record. They appear as patterns: held items waiting too long, the same correction made forty times, a reviewer approving everything in seconds. A manager who cannot read those patterns will see a dashboard of numbers and react to the wrong ones.
So the case for certifying managers is different from the case for advisers. Advisers need to judge single items. Managers need to judge the system that produces them.
What does a contact centre manager need to read in the workspace?
The workspace shows, among other things, what each agent is certified to do, at what depth, and when that certificate lapses. It shows the held queue and how long items have waited. It shows which bound decided each verdict. It shows corrections reviewers have banked.
A certified manager reads it as an operating picture:
| What the manager sees | What it may mean | What they decide |
|---|---|---|
| Held refunds ageing past the service target | Too few certified reviewers on shift | Change the rota, not the agent's limits |
| A spike in refusals on account changes | A source went stale, or a limit was narrowed | Find the cause before anyone asks for more depth |
| One reviewer approving held items very fast | Possible rubber-stamping | Sample their decisions and talk to them |
| Agent certificate due to lapse in a week | Recertification needs fresh cases | Make sure the team has banked corrections |
| Same correction made repeatedly | The agent's brief or sources need updating | Route to whoever owns the agent |
An uncertified manager tends to respond to the first row by asking for the agent's refund limit to be raised so fewer items are held. That is the wrong lever. It moves risk from the queue to the customer.
Why does depth matter so much to a manager?
Because depth changes the shape of the work. If the agent loses its certificate on a capability, its depth drops to 1 the same day, and everything it used to handle alone arrives as held items. A manager who does not understand that will see a sudden queue and assume a fault.
The agent's certificate lapses after 90 days, so this is a recurring event, not a rare one. Managers should know when each agent's renewal falls and plan staffing around it. The guide on preparing contact centre managers for the 90-day recertification cycle covers how to build that into the rota.
Depth also decides who may approve what. If a business ties approval rights to current reviewer certificates, the manager is the person who keeps that list true. They need to understand the rule well enough to explain it to an adviser who has lost rights because their own certificate lapsed.
What does certification prove for a manager?
It proves the manager can do four things under test:
- Read the workspace and name the likely cause of a pattern.
- Distinguish a refusal (outside this agent's bounds) from a withheld action (blocked for every agent at the platform floor), and explain why neither is a bug.
- Decide when to lower an agent's working depth for the team, and when to leave it.
- Run the review rhythm: rota, sampling, banking corrections, escalation.
A practicum adds what a written test cannot: two weeks of real decisions in a live workspace, observed. For a manager, the most useful practicum entries are the moments they changed the rota, paused a capability or challenged a reviewer. The practicum log template for contact centre managers gives a structure for recording those.
What is the trade-off?
Managers are scarce and busy, and certifying them before go-live takes them away from running the floor. Some will feel that a certificate for a manager is overkill when the advisers do the reviewing.
There is a fair counter-argument that managers learn best once the system is live. They do learn faster then. But the first weeks after go-live are when the patterns are least familiar and the wrong reactions do the most damage, and a manager who raises a limit to clear a queue can create harm that takes months to unwind.
Then there is accountability. When something goes wrong, the question of whether contact centre managers are accountable for an agent's mistakes turns on what the manager knew and what they did. A certified manager who followed the rhythm is well placed. One who was never trained to read the workspace is not.
What should an operations director do differently?
Certify managers first, not last. Make it a go-live criterion alongside reviewer certification. Give managers ownership of the review rota and the certificate list. Teach them that the answer to a growing held queue is people and process, not wider limits. And compare your plan with how finance teams approach the same question for financial controllers and how collections teams approach it for collections agents. The underlying principle is set out in the guide to AI literacy and certification for people who work with agents.
Questions people ask
Does a contact centre manager need to review held items personally?
Not routinely, but they should be qualified to, and they should sample reviewers' decisions. A manager who never looks at individual verdicts loses the feel for what the numbers mean.
Who decides whether an agent's depth is lowered?
The depth an agent may use is set by evidence and the business's own limits. The manager can narrow the business's limits for their team, for example by holding a capability that was previously allowed, and should record why.
What if the held queue grows faster than the team can review it?
Treat it as a staffing and process issue first. Add certified reviewers, check whether a source or limit changed, and only then ask whether the agent could earn more depth through recertification.