Order status and returns tickets, answered before your team opens the queue
Where-is-my-order is the most common question in ecommerce support, and return questions arrive every day too. Each answer needs lookups in two or three systems, and the exceptions wait until a lead has time. The agent does the lookups and drafts the reply. Your team approves anything that moves money.
Examples use invented names and figures.
What the agent does
- Step 1Reads and sorts the ticket
From Gorgias, Zendesk, Shopify Inbox or a shared mailbox. It tags each one as order status, delivered but not received, address change, cancellation, return status, a return outside your policy, a damage claim, or something for a person.
- Step 2Looks up the order
It finds the Shopify order and checks the 3PL's shipment record against the carrier's events, then pulls the line of your shipping or return policy that applies.
- Step 3Drafts the reply or the exception file
Routine tickets get a reply in your brand's voice. Exceptions get a short file with the order history, the policy clause, any photos, the carrier evidence, the customer's past returns and a suggested outcome.
- Step 4Queues it for approval
Your team approves in one click. Anything that moves money, such as a refund or a reship, waits for a person. A category goes out without review only after you decide so in writing.
Refunds and reships outside your policy stay with a person, and so does anything above a dollar limit you set. Damage and safety complaints, card disputes, legal threats and messages from press or influencers go straight to a lead. The agent works from order references rather than card details. Card numbers that shoppers paste into a ticket are removed from its inputs, and we test that control before launch.
- Tickets handled per agent per day
- First response time, before and after
- Share of drafts approved without edits
- Exceptions decided per day and their dollar value
- Reopened tickets
Questions about this agent
Gorgias and Zendesk already sell AI agents. What is different?
Use the helpdesk bot for the in-policy answers it already handles well. We build for the work it leaves behind, such as lookups across your 3PL and the carrier or exceptions to your policy, with an approval queue and the evidence attached. The sprint shows which of your tickets fall where.
Will customers know a reply was drafted by AI?
Replies your team reviews and sends go out from your team, in your voice. Set a written policy on whether they mention AI assistance, and confirm it with counsel. A chat bot that talks to shoppers on its own should say clearly that it is AI from its first message. Our guide on chatbot disclosure laws covers the state rules.
Can the agent issue refunds?
It prepares them with the evidence and a suggested amount, and a person approves each one. Later you may release a narrow category in writing, such as a reship for a lost parcel under a set value.