Ask anyone who has worked a Shopify inbox what is waiting on Monday morning and you get the same short list. "Where's my parcel?" "Can I fix my address?" "Cancel it, please." "Does this come up small?" "Any codes?" Gorgias can take most of that away from your team, once someone writes the rules that say what it may do, with which data, and when it has to stop.
Below are the nine automations we build first when we set up Gorgias for a Shopify brand. They're grouped by how much the AI is allowed to change. The first three only read data and reply. The middle ones edit orders inside limits you write down. The last one is aimed at sales.
What has to be in place before rule one
The Gorgias Shopify integration, connected. Without it the AI Agent can't see orders, fulfillment status or customer history, and every rule below depends on that data.
AI Agent switched on and fed. Add your Help Center, your shipping and returns policies and your tone guidelines as knowledge. It answers from what you give it.
A tagged sample of recent tickets. Read a few hundred before writing a trigger. Customers describe problems in their own words ("my stuff never came", "wrong size again"), and those words rarely match the labels your team uses internally. Triggers built from your own tickets catch more and misfire less.
Then resist the urge to switch everything on at once. Start with the three automations that cover your largest ticket tags, watch them, and add the rest in batches. A rule that fires on the wrong message costs you more than having no rule at all.
Level one: replies that change nothing
This is the safe place to begin. The AI looks things up and answers. Nothing in Shopify moves, so if a rule gets it wrong, a person corrects the reply and no order is affected.
1. Order status
Order tracking is usually the largest and most predictable category in a DTC inbox. When a message asks about delivery or includes an order number, the AI Agent pulls the order from Shopify, reads the latest tracking event and tells the customer where the parcel is, in your brand's tone.
If the parcel is moving normally, the ticket closes there. If tracking hasn't updated for longer than the limit you agree with your logistics team, the AI hands the ticket to a person with the order and the tracking history attached. A stuck parcel is a job for a human to chase with the carrier.
2. Sizing and fit
Half of this automation lives in your Help Center. The AI Agent answers from the articles it has, so a vague size guide produces vague replies. When a shopper asks whether a style runs small, it quotes the relevant line from the guide, links the article and offers a person if that doesn't settle it.
For fashion and footwear brands the question has a second cost, because a fit doubt left unanswered before checkout can come back as a return. Write the size articles from the questions customers actually send, by product line where fit differs. In our team's track record, a Help Center built that way deflects 25-40% of repeat questions.
3. Discount requests
People ask for a code for different reasons. Some are trying their luck. Others have a case your team would want to hear, such as a loyal customer or someone with an open complaint about their last order.
The rule tells them apart. Before replying, the AI Agent checks what Gorgias knows about the customer, including their Shopify tags and any open tickets. If one of those reasons applies, a person gets the ticket with that context on top. If none does, the AI replies with your discount policy, points to the official route you offer (a newsletter sign-up, a loyalty programme) and closes. Track these requests month by month to see whether one consistent answer changes the volume.
Level two: changes in Shopify, inside written limits
From here the AI Agent edits real orders. Each rule needs two conditions agreed before it goes live: the one that allows the action and the one that sends the ticket to a person.
4. Address changes before fulfillment
An address can be changed until the order reaches the warehouse. After that it means a call to the warehouse or the carrier. Agree that window with your fulfillment team, in writing, and build the rule on it.
Inside the window, the AI Agent updates the shipping address in Shopify, confirms with the customer and closes the ticket. Outside it, operations gets the ticket with the order number and the new address ready to copy. When fulfillment moves the cut-off, update the rule the same day.
5. Cancellations
Same family, same logic. If the order hasn't been fulfilled and its value is under the limit you choose, the AI Agent cancels it in Shopify, refunds the customer and confirms. If it has shipped, or it's above the limit, a person decides. While an order is still waiting to be picked, these requests rarely need judgement, and taking them off the queue gives your team time for the ones that do.
6. Refunds for defective or wrong items
This is the rule that makes finance nervous, and with tight limits it's also one of the safest. The AI Agent may issue a full or partial refund only when every condition holds. The order is under the amount you set. The ticket is tagged as a defect or a wrong item. The customer has uploaded a photo of the problem. The account has fewer previous refunds than your limit.
If any condition fails, a person takes it. Keep the photo step even when it feels like friction. A customer with a broken product can usually send a picture in seconds, and that small step is what makes the rule safe to run unattended. Watch refunds issued by the AI, disputes and chargebacks.
Level three: actions in the apps around Shopify
Many brands run subscriptions and returns in separate apps. When those apps are connected to Gorgias, the AI Agent can act in them too.
7. Recharge subscriptions
Pausing, skipping a delivery, swapping a product and cancelling are routine for any brand on Recharge. The AI Agent checks that the person writing owns the subscription, makes the change and confirms it. Cancellations deserve one extra step. Have the AI offer a skip or a swap first, cancel only if the customer still wants out, and measure how many subscriptions that step keeps.
8. Loop returns and exchanges
With Loop, starting a return is already self-service, so the AI Agent's job is to get the customer there. It sends the link to the returns portal, answers policy questions from your returns article and passes the ticket to a person only when the customer says the portal didn't work for them.
9. The one that sells: Shopping Assistant
The eight automations above save time. Gorgias Shopping Assistant is pointed at revenue. It answers questions in chat before the purchase, on product pages and in the cart, and helps the shopper get to checkout.
It's only as good as what you load into it: the full catalog, the size guides, shipping cut-offs and stock information. Brands that switch it on without that get generic answers, shoppers leave, and the team concludes the tool doesn't suit them. Fix the knowledge base before judging the tool. From the first month, compare the revenue Gorgias attributes to these chats with the full cost of running them.
A rollout that doesn't break the inbox
Choosing the first three is a counting exercise. Group last month's tickets by tag and start where the volume is. Then keep to a few habits:
- Add one automation at a time. Let it run on real volume and decide whether it stays, changes or goes before you add the next.
- Give every rule a clean exit. A customer who asks for a person should reach one, with the whole conversation attached, after one bot reply at most.
- Review the rules every week. Catalogs, carriers and promotions change, and a trigger that matched well at launch can start misfiring months later.
If you'd rather have this built for you, that's what our CX automation and AI Agent setups cover. Our team's implementations take 4-8 weeks, depending on scope.
What a well-tuned account looks like
Across properly configured accounts, our team's median is 50-60% of repeat tickets handled by the AI Agent without a person touching them. The rest reaches your team with the order, the history and the reason for the handover already in the ticket, so the people on shift spend their day on the cases that need them.
For where the AI Agent falls short, read our honest take on Gorgias AI Agent.
Our team was named Best European Gorgias Partner 2024. Book a demo, bring your ticket tags, and we'll tell you on the call which three automations to build first.
