PrymeIntelligence

AI agent skills by role·Playbook

How to build a reviewer rota for contact centre

How to build a reviewer rota for a contact centre: live handoffs staffed by interval, back-office holds by day, response targets and the skills mix.

Split the contact centre rota into two lines. The live line takes handoffs while the customer is still on the chat or call, so it is staffed interval by interval like any front line. The back-office line decides holds the customer is not waiting on in real time, such as refunds above limit and account changes.

Each line gets its own response targets and skills mix.

Why does a contact centre need two lines?

A contact centre agent answers customers directly. When it reaches the edge of its bounds mid-conversation, it hands off: the customer asks for something it may not do, or becomes upset, or raises a complaint. Those handoffs are live, and a customer waiting three minutes for a person feels it. Other holds arise after the conversation: a refund the agent drafted for approval, an address change awaiting checks. Those can wait hours.

Mixing the two in one queue makes both worse. Advisers abandon back-office holds whenever a live handoff arrives, and live handoffs wait whenever advisers are deep in a back-office case. The contact centre manager owns the rota, working with workforce planning. For the skills involved, see the guide to AI agent skills by role, and for how managers read the agent's decisions, how contact centre managers review an AI agent's decisions.

What are the steps?

  1. Agent, depth 1: reports handoffs and holds separately. Handoffs by interval and reason; holds by reason and age.
  2. Person (workforce planner): forecasts live handoffs by interval. The same way contact volume is forecast, from the agent's history.
  3. Person (contact centre manager): sets response targets. Live handoffs answered within a short target measured in seconds or minutes; refunds above limit within a working day; account changes within the time policy promises customers.
  4. Person (workforce planner): staffs the live line by interval. With advisers trained to pick up a conversation from the agent's summary.
  5. Person (contact centre manager): staffs the back-office line by day. Named reviewers, with team leads for refunds above the advisers' authority.
  6. Agent, depth 2: drafts a summary for every handoff. What the customer asked, what the agent did, why it handed off, and what it refused.
  7. Agent, depth 3: completes routine requests within bounds. Balance queries, address changes after checks, small refunds.
  8. Person (contact centre manager): reviews intraday and weekly. Moving people between lines as the day unfolds.
Step The agent does Depth Stays with a person
Routine requests Completes within bounds 3 Sampling quality
Live handoff Summarises and transfers 2 Taking over the conversation
Refunds above limit Drafts and holds 2 Approval by an authorised reviewer
Account changes Holds pending checks 2 Confirming identity and approving
Complaints Flags and routes 2 Logging and handling
Refused requests Refuses and explains n/a Reviewing refusals in sample

What skills does each line need?

The live line needs advisers who can read a handoff summary in seconds and continue without making the customer repeat themselves. They also need to recognise what the agent refused and why, so they do not grant on the phone what the agent correctly refused. Refusals are often right: the request was outside policy, not outside the agent's ability.

The back-office line needs reviewers who can read the record properly: sources, provenance, the bound that held the action. Team leads take refunds and credits above advisers' authority. The AI agent competency matrix for the contact centre helps decide who is ready for which line.

Which controls matter, and what evidence do they leave?

  • Handoffs have a named taker. Evidence: the adviser's name on the conversation from the moment of transfer.
  • Refusals are not reversed without a reason. Evidence: an adviser's override of a refusal is recorded with a reason, and sampled by team leads.
  • Authority matches the hold. Evidence: refund verdicts show the approver and the limit.
  • Complaints are logged as complaints. Evidence: the complaint record linked to the conversation.
  • Lines are staffed separately. Evidence: the rota shows who is on which line each interval.

What should be measured?

  • Live handoff answer time by interval, against target.
  • Back-office response targets met, per hold reason.
  • Customer repeat rate after handoff: how often the customer had to explain again.
  • Refusal overrides, by adviser and reason.
  • Movement between lines intraday, which shows whether the forecast is right.

Where does it go wrong?

One queue for everything. Already covered, and still the most common design.

Advisers reverse refusals to please customers. A pleasant override can breach policy. Team leads should review overrides daily in the first weeks.

The agent's hours exceed the rota's. If the agent answers chats at night and hands off to nobody, customers are left waiting. Either staff the hours or set the agent to tell customers when a person will reply.

Lending and collections teams run similar two-speed queues; compare the rotas for the lending team and the collections team.

Where the Academy fits

Contact centre advisers and team leads who review agent work are the people the Academy's Operator track is written for. Its modules on human review and running the rhythm cover reading handoffs, respecting refusals and keeping held work moving. Tests are set in the learner's own industry and role, and Certified Pryme Operator is valid for 12 months.

Questions people ask

Should advisers on the live line also see back-office holds?

They can see them, but should not be assigned them while on the live line. Visibility helps; assignment splits attention.

How do we forecast handoffs before the agent goes live?

Start from the contacts you expect the agent to hand off, based on the bounds you set, and staff generously for the first weeks. Replace the estimate with real data as soon as it exists.

What if a customer insists on a person from the start?

Hand off. Customers should be able to reach a person, and they should be told when they are talking to an AI system. The live line exists for exactly this.

Train the people who run your agents.

Pryme Intelligence Academy: free courses on the real product, audio lessons, labs in a free workspace, and certificates anyone can check on VerifyDoc.ai. Teams get seats, an admin view and a free readiness assessment.