It depends on four things: how many items the agents hold, and when; how fast each kind must be decided; who covers when people are away; and which skills each hold reason needs. Staff from the hold reasons, not from headcount. Every reason needs a named, competent approver at every hour the queue is open.
Why is a review queue different from a work queue?
In a work queue, people do the work. In a review queue, agents have already drafted it, and people decide whether it goes ahead. Each item is quicker to handle, but needs more judgement per minute, and the cost of a careless decision is higher, because the agent's draft looks finished.
That changes the staffing maths. You need fewer hours of preparation and more hours of qualified attention, spread to match when holds arrive.
How do you size the queue?
Volume by reason and time. Count held items per hour of the day and day of the week, split by hold reason. A payments queue may spike before cut-offs; a support queue after a product release. Averages hide the peaks that cause backlogs.
Handling time by reason. A changed bank detail with a call-back takes far longer than a refund slightly above limit. Measure it from the record, where time on item is filed with each decision.
Response targets by reason. Some holds must be decided within minutes, others by end of day. The targets, set by the business and covered in how to set response-time targets, decide how much cover each hour needs.
Depth changes. Certificates lapse after 90 days, and a certification run that clears no depth drops an agent to depth 1 the same day. Work the agent completed alone at depth 3 then arrives as drafts. Plan for that surge, especially near recertification dates.
What skills mix does a queue need?
| Hold reason | Who can approve | Why |
|---|---|---|
| Above the agent's monetary limit | Someone whose own authority covers the amount | Authority |
| New or changed payee details | Verification staff with call-back access | Independent check |
| Fraud or sanctions flag | Financial crime team | Competence |
| Vulnerable customer | Trained specialist | Competence |
| Outside tested cases | Experienced reviewer for that process | Judgement |
| Formal complaint | Complaint handler | Process rules |
Map each hold reason to the roles that can decide it, then to named people. A queue staffed with plenty of reviewers but nobody qualified for fraud flags after 4pm is understaffed.
How do you plan cover?
- Two names per reason per shift, so one absence does not stall a whole category.
- A rota that follows the volume curve. The review queue rota template gives a format.
- An escalation route when a target is about to be missed, which goes to a person with authority, not to bulk approval.
- Certificate expiry dates on the rota. If a firm ties approval rights to a current certificate, a reviewer whose certificate lapses next week is not cover next week.
A worked example
Take a payments firm whose agents hold around 300 items a day. The record shows 60 are bank detail changes, each taking around 15 minutes with a call-back; 200 are payments slightly above limit, taking around 2 minutes; and 40 are fraud flags, taking around 10 minutes, concentrated between 10am and 2pm.
That is roughly 15 hours of bank changes, 7 hours of limit approvals and 7 hours of fraud review a day, before breaks and checking. The firm staffs three verification staff across the day, two payments reviewers, and two fraud analysts on overlapping shifts covering the midday peak. Each reason has at least two qualified names on every shift. When a fraud analyst is on leave, the rota shows the gap a week ahead.
What goes wrong?
- Staffing from averages. The queue copes on Tuesday and collapses before month-end.
- One generalist pool. Everyone can see everything; nobody is qualified for the hard items.
- Speed targets without staffing. Reviewers meet targets by approving faster, which is rubber-stamping.
- No plan for lapses. A certificate lapse doubles the queue overnight.
For how reviewers in one high-volume team work through their decisions, see how payments operations staff review an agent's decisions. The skills each role needs are mapped in AI agent skills by role.
Where the Academy fits
Pryme Intelligence Academy's admin view shows each person's progress, results, certificates and expiry dates, which is the data a queue rota needs if approval rights depend on a current certificate. The Operator track's module on running the rhythm covers the daily and weekly routines of a review queue. Seats are included with paid Pryme Intelligence plans.
Questions people ask
Should reviewers also do other work between holds?
Yes, if the work can be dropped instantly when a hold arrives with a short target. Otherwise, dedicate people during peaks.
How many items can one reviewer handle before quality drops?
It varies by hold type. Watch decision quality in sampled reviews against items handled per hour, and set limits from what you see.
Who owns the review queue?
The process owner for the work the agents do, with the operations manager responsible for staffing and cover.