---
title: "Orders, returns and refunds with AI"
description: "An agent answers on orders and refunds and prepares returns and exchanges under the brand's rules, and refunds and exceptions stay with the operator."
url: https://www.joinx.it/en/industries/use-cases/orders-returns-refunds
---

# Requests about orders and returns, closed with data from every channel

The customer asks where the parcel is or is waiting for a refund, and the answer depends on systems that customer service checks one at a time and that do not always agree.

## The questions repeat, and so do the searches

The most frequent requests to customer service concern the status of an order, a late delivery, a change of size or a refund that has not arrived. To reply, the operator opens the orders, the courier's tracking, the returns and the payments one after another, and sometimes each shows a different date. The time goes on these repeated searches, on requests the customer sends again because the first reply was incomplete, and on refunds held up awaiting a check.

## The refund and the cases outside the rules stay with the operator

The agent identifies the order, reads the status, tracking and payment and replies to the customer when the answer is a fact, such as the delivery date. For a return or an exchange it checks the brand's returns policy and, for online purchases, the right of withdrawal under the Italian Consumer Code (Codice del consumo), then prepares the authorisation. The refund, which cannot be reversed, is approved by the operator, who also receives the disputes and the cases outside the rules.

1.  The customer writes about an order or a refund
2.  The agent links the order, tracking and payment and replies on the status
3.  For returns and exchanges, it checks the rules and prepares the authorisation
4.  The operator approves the refund and decides the exceptions

Illustrative example

## Data, modules and responsibilities across orders and refunds

| Item | In this case |
| --- | --- |
| Data | Website and store orders, courier tracking, returns, payments and refunds, the returns policy in force |
| Modules | Data Foundation, Agents, Workspace |
| Who decides | The operator on refunds and exceptions; the head of customer service on the agent's rules and on the replies it sends on its own |
| What we measure | Time to close, repeated requests, refunds held beyond the returns policy's time limits and replies to correct, also measured before release so there is a comparison |
| AI Act risk class | Because the agent talks to the customer, the transparency obligation applies, and the customer must understand they are talking to an AI system. The case is outside the Annex III high-risk uses most common in our industries, such as the creditworthiness of individuals, and we confirm the class with those who look after risk and compliance at the start of the project |
| Where to start | With Deployment, because the requests are repetitive, the returns rules are written down and the starting point can be read in the customer service systems |

## Order, return and payment have to be linked before the agent

If the website, the store and the payment system do not share an order number, or if the refund status is updated once a day, the agent replies with out-of-date information and the customer calls back. Before building it, we link the sources and establish how often each piece of information is updated. With modest volumes, a written procedure and an experienced operator remain the cheapest option.

The sheet comes from the Consumer page.

[Data and AI for consumer businesses](https://www.joinx.it/en/industries/consumer) · [All use cases](https://www.joinx.it/en/industries/use-cases)

## Let's start with the requests that come back twice

Starting from the requests a customer had to repeat, we look with you at which system was missing the answer.

[Contact us](https://www.joinx.it/en/contact?settore=consumer)
