How to build a reviewer rota for claims team

Plan two rotas, not one. The base rota staffs the claims agent's normal flow of holds: coverage questions, settlements above authority, fraud indicators and declines. The surge rota names who steps in when a storm or flood multiplies claims overnight. Both set response targets per hold reason and match each reviewer's settlement authority to the holds they will see.

What is the claims rota for, and who owns it?

A claims agent takes first notification, gathers evidence, checks the policy, sets an initial reserve and drafts settlements. It settles small, clear claims within bounds and holds the rest: settlements above a reviewer's authority, claims where cover is in doubt, claims with fraud indicators, declines, and claimants who appear vulnerable. Each hold reason needs a different skill and a different response time.

The claims manager owns the rota. The head of claims agrees the authority each handler holds. For what each role needs to know, see the guide to AI agent skills by role. For the terms used here, see what a reviewer rota is and what an AI agent reviewer is.

What are the steps?

1. Agent, depth 1: reports holds by reason, size and age. Pull enough history to include at least one weather event, so the surge shape is visible.
2. Person (claims manager): sets response targets per hold reason. Vulnerable claimants the same day; settlements above authority within a working day; coverage questions within two; fraud referrals passed to the investigation team the same day.
3. Person (claims manager): staffs the base rota by authority band. Each day needs handlers who can decide settlements up to the levels the agent typically holds, plus a senior for larger ones.
4. Agent, depth 4: settlement payments stay held above the bound. Paying a claim is hard to undo; it waits for a handler within authority.
5. Agent, depth 2: drafts declines. Every decline is held for a handler, with the policy wording relied on.
6. Person (claims manager): writes the surge plan. A trigger (for example, claims volume above a set multiple of normal), a named surge lead, and a reserve list of trained handlers from other teams who can join.
7. Person (handlers on the rota): decide each held claim. Approve, adjust, decline, refer to investigation, with reasons.
8. Person (claims manager): reviews after every surge. What waited, what was decided in haste.

How should the surge rota work?

Surges are when quality falls. Handlers clear backlogs fast, the agent's holds pile up, and settlements get approved on thinner evidence. A surge plan protects against that in three ways.

First, it brings in people already trained to review the agent's work, not just people who can handle claims. A reserve handler who has never read an agent's record will approve its drafts on trust. Second, it keeps authority bands intact: a surge is not the moment to let junior handlers release large settlements. Third, it separates the queue: weather claims go to the surge team, other claims stay with the base rota, so a fire or a theft claim does not wait behind hundreds of storm claims.

The agent also changes behaviour in a surge only if its bounds change. Raising its settlement bound to clear volume is a depth decision that needs Attest evidence, not a rota decision.

Which controls matter, and what evidence do they leave?

• Authority matches the hold. Evidence: each settlement verdict names the approver and the limit that applied.
• Fraud referrals are not decided by the handler alone. Evidence: the referral and the investigation team's acknowledgement.
• Declines carry policy wording. Evidence: the wording cited on the verdict.
• Surge activation is recorded. Evidence: the trigger, the time, the surge lead's name.
• Vulnerable claimants are prioritised. Evidence: time from flag to first contact.

What should be measured?

• Response targets met per hold reason, in normal weeks and in surges, reported separately.
• Settlements later reopened or adjusted, by the week they were approved; a rise after a surge is a warning.
• Fraud referrals made per surge, compared with normal weeks; a fall during a surge suggests indicators are being missed.
• Reserve handlers called, and how quickly they were productive.

Where does it usually go wrong?

No surge plan until the surge. By then the trained reserve does not exist. Train it in quiet months.

Authority creep under pressure. Managers lift limits informally to clear queues. The record will show approvals above authority; better to plan enough senior cover.

Declines rushed. Declines take longer to write well. Under pressure they become generic. Keep declines with experienced handlers even in a surge.

The method applies with different peaks to the rotas for the IT team and the compliance team. For more on the reviewer's own role, see what an agent reviewer is.