A project to build with System One models

A returns and refund pre-check that prepares each case for a person

Check every return or refund request against your policy before an agent opens it, with a probability for each rule, so clear cases are ready in seconds and a person still approves every refund.

Support and triageIntermediate5 stepsSuggested by @biplov

The decision

A noul question for each rule of your returns policy that depends on what the customer wrote: "the customer reports damage or a wrong item", "the item was used", "the customer asks for a replacement rather than a refund". Then a choice over the next step: refund, replacement, ask for photos, or pass to a specialist. The model prepares the case; a person decides it.

How to build it

  1. Write each rule of your policy as one plain condition, in the words your agents use.
  2. Check dates, amounts and order history with code, not with the model, and put the results in the state next to the customer's message.
  3. Ask the questions when a request arrives, and attach the answers to the ticket.
  4. Show the agent the suggested step and each rule's probability, with one button to approve it and one to change it.
  5. Track how often agents change the suggestion, rule by rule.

Make it better

The rules agents overrule most are usually the ones your policy words badly: rewrite them, and fine-tune on the corrected cases. Refunds stay a person's decision, however good the model gets.

More projects to build

All projects