How to handle delivery problems and lost parcels with an AI agent

Lost and late parcels follow a pattern the agent can run: read the carrier's proof of delivery, guide the customer through the usual checks, open a carrier investigation when your waiting period passes, and draft the replacement or refund. A person approves the remedy, and anything that looks like a repeat claim pattern goes to fraud review instead.

What is the process, and who owns it?

Delivery problems come in a few shapes. The parcel is late and tracking has stopped. Tracking says delivered, but the customer has nothing. It arrived damaged. Part of the order is missing. A delivery was missed and needs rearranging. Each has a standard path, and most firms have policies on waiting periods, evidence and remedies, shaped by their terms, their carrier contracts and consumer rules in their market about who bears the risk until goods arrive.

Fulfilment or logistics owns the carrier relationship and investigations. Customer support owns the conversation. A team with authority owns replacements and refunds. Fraud owns patterns of repeated claims. The agent can run the path but not decide the remedy, consistent with our guide to the AI support agent that knows when to say no.

What are the steps, and at what depth?

1. Agent, depth 1: identifies the problem type: stalled, delivered but not received, damaged, incomplete, or missed.
2. Agent, depth 1: reads carrier evidence. Tracking history and proof of delivery: a photo, a signature, a location, a safe-place or neighbour note. Carrier data is signed by the carrier; the customer's account is declared.
3. Agent, depth 3: shares the proof and the usual checks. Where the photo shows the parcel, around the property, with neighbours, with other household members. It does this without implying the customer is mistaken.
4. Agent, depth 3: rearranges missed deliveries through the carrier's options, where certified for that capability. It does not commit to a delivery time the carrier has not given; can an AI agent commit to a delivery time? covers why.
5. Agent, depth 3: opens a carrier investigation once your waiting period passes, or at once for damaged goods.
6. Agent, depth 1: reads photos of damage and attaches them as external evidence.
7. Agent, depth 1: checks claim history for the customer and the delivery address. Repeat claims are routed to fraud review, not refused by the agent.
8. Agent, depth 2: drafts the remedy: replacement or refund, per your policy, with the evidence and any carrier finding.
9. Person, fulfilment or support lead: approves the remedy within authority.
10. Agent, depth 3: tells the customer what was decided and when the replacement or refund will arrive, from the system.

Which controls matter, and what evidence does each leave?

Proof is shown, not argued. If the carrier photo shows a different door, that matters. Evidence: the proof shared and the customer's response.

Waiting periods come from policy. The agent applies them consistently. Evidence: the policy entry cited when an investigation opens.

Remedies are held. A refund moves money; a replacement ships goods. Both wait for a person unless your limits allow low-value replacements at depth 3. Evidence: held verdicts and approvals.

Repeat claims are routed, not judged. The agent never accuses a customer. Evidence: fraud referrals with the pattern found.

One case per problem. Evidence: a single case linking the order, the carrier investigation and the remedy.

What should you measure?

• Problems by type and carrier. A carrier with many "delivered, not received" claims needs a conversation.
• Time from first contact to remedy.
• Investigations opened and their outcomes.
• Remedies approved and declined, and why.
• Repeat-claim referrals and how many fraud review upheld.

Where does it usually go wrong?

Customers are treated as suspects. Questioning a genuine customer as if they were lying loses them. Share the proof neutrally and follow the policy.

The waiting period is applied rigidly. A parcel containing medicine or something needed for a date is urgent. Route time-critical cases to a person.

Remedies are promised early. "Don't worry, we'll send another" before approval is a commitment. The agent says what it has requested and when the customer will hear.

Date changes are treated as simple. Moving a delivery date involves the carrier's options and sometimes cost. See can an AI agent change a delivery date?

Sensitive situations get the standard script. A parcel for someone who has since died, or a delivery to a care home, needs different handling; handling bereavement notifications with an AI agent covers the first case.

Partial deliveries look like losses. An order shipped in several parcels shows one delivered and one in transit, and the customer reports the whole order missing. Check each shipment before treating anything as lost.

Address errors are blamed on the carrier. The customer typed the wrong house number. Check the order address before opening an investigation.

Where Pryme fits

In Pryme, carrier proof of delivery arrives signed by the carrier and the customer's report as declared, and both are shown on the case. Remedies are held by default; the customer's own limits can allow low-value replacements at depth 3, and the verdict names the limit that decided.