Gorgias out of the box is a good helpdesk. Gorgias configured is a profit centre. The difference is somewhere between forty and sixty automation rules, a tuned AI Agent, and a macro library mined from your real ticket history.
This is the playbook we ship on every Growth or Enterprise setup. Nine automations. Each one shaves either time, money, or both off the inbox.
A note before we start
These automations assume you're on Shopify with the Gorgias-Shopify integration enabled. They assume your AI Agent is licensed and on. If either is missing, fix that first — these don't work without the foundation.
Also: don't ship all nine in week one. Pick the three that match your biggest ticket category, ship them, measure for two weeks, then layer the rest. Automation that fires on the wrong trigger is worse than no automation.
1. "Where is my order?" auto-responder
The biggest single ticket category in DTC is order status. Usually 25–40% of inbound. It's also the easiest to kill.
The rule: when an incoming ticket contains "where", "order", "status", "tracking" or the user pastes an order number in the body — the AI Agent looks up the order in Shopify, finds the latest tracking event, and replies in your voice. If the order is delivered or in transit normally, it closes the ticket. If it's stuck (no movement >72h), it escalates to a human with all the context attached.
What we measure: category close rate without human touch. Bar is 70%+ on a properly tuned rule.
Common pitfall: rules that fire on "tracking" but not on the customer's actual phrasing. Mine your ticket history for the words your real customers use. They don't say "tracking number" — they say "where's my stuff".
2. Address change inside the SLA window
Every brand has a window — usually 1–4 hours after order — when an address change is operationally feasible. After that, the order is in fulfillment and changes need warehouse coordination.
The rule: AI Agent reads "wrong address" or "change address". If the order was placed less than X hours ago, AI Agent updates the Shopify shipping address, replies confirming, closes the ticket. If older, it escalates to ops with the order ID and the new address pre-formatted.
What we measure: % of address-change tickets resolved without human touch. Easy 60%+ with a clean rule.
Tuning note: define "X hours" with your fulfillment team, in writing, before you ship the rule. If they change SLA, you change the rule.
3. Refund inside policy rails
This is the one that scares CFOs the most and saves the most money. Done right, it's also the safest.
The rule: AI Agent has permission to issue full or partial refunds only under specific conditions: order under €X, item is defective per a confirmed photo upload, ticket category matches "defect" or "wrong item", and the customer has fewer than Y prior refunds. Outside the rails, escalate.
What we measure: AI-issued refund volume, dispute rate, downstream chargebacks.
Where this fails: brands that try to skip the photo-upload step. Don't. The friction is the safety. Customers who have a real defect upload a photo. Bad actors don't.
4. Order cancellation pre-fulfillment
Same logic family as #3. Cancellations land all day. Most are inside the fulfillment window. Letting AI Agent handle them frees your team for the harder cases.
The rule: "cancel order" or "I want to cancel". If order is unfulfilled and under €X, AI Agent cancels in Shopify, refunds, replies. If fulfilled or above the threshold, escalate.
What we measure: cancel-and-close rate. 50%+ on a clean rule.
5. Sizing and fit deflection on PDP
This one's a Help Center play, but it shows up as automation because the Help Center articles are fed to the same AI Agent that powers chat.
The rule: when a chat or ticket contains sizing language ("does this run small", "what's my size in", "fit"), AI Agent surfaces the right size guide article first, summarizes the relevant point, and offers to bring in a human if needed. Most customers click through to the article and never reply.
What we measure: size-question chat resolution without agent. 65%+ once your size guide articles are mined from real ticket data.
Why this matters: sizing is the dominant return reason in fashion and footwear. Every sizing question you answer well is a return you don't process later.
6. Discount-request triage
A meaningful chunk of inbound is people asking for a discount. Most of it is bad-faith negotiation. Some of it is legitimate (loyalty members, post-purchase issues).
The rule: "discount", "promo", "code". AI Agent first checks if the customer is a loyalty member, on a subscription, or has a post-purchase issue ticket open. If yes, escalate to a human with context. If no, AI Agent replies with the brand's official discount policy in your voice, surfaces the newsletter signup as the only authorized path, and closes.
What we measure: chat-to-discount ratio over time. Should drop quickly when the policy is consistent.
7. Subscription management flows
For brands on Recharge: pause, skip, swap and cancel are routine. They don't need a human.
The rule: AI Agent handles "pause my subscription", "skip next order", "swap", "cancel". It validates the customer is the subscription owner, performs the Recharge action, replies confirming. For cancellations, it offers a save flow first (one skip, one swap, then cancel).
What we measure: AI-handled subscription action volume, save-flow conversion.
8. Returns initiation via Loop
For brands on Loop: returns initiation is a self-service motion that should never touch CX.
The rule: "return", "exchange", "send back". AI Agent surfaces the Loop self-service portal with a deep link to the customer's order. If the customer says they've already tried the portal and it didn't work, escalate. If they ask a return-policy clarification question, AI Agent answers from the policy article and offers the portal.
What we measure: % of return-related tickets that close without agent touch. 70%+ when Loop is properly wired.
9. Pre-purchase Shopping Assistant
This is the one that flips CX from cost to revenue. It's also the one most brands ignore.
The setup: Gorgias Shopping Assistant runs on PDPs and the cart. It answers product questions, checks live stock by variant, and pushes to checkout. Trained on your full catalogue, your size guides, your shipping cut-offs.
What we measure: revenue attributed to Shopping Assistant chats. We've measured a Spanish DTC account at €163K/month from this single layer, against fully-loaded cost of about €8K/month. 20× ROI documented. Not all accounts hit those numbers — but every account we've shipped this on has cleared 5× ROI inside three months.
Where it goes wrong: brands that bolt on Shopping Assistant without giving it a real knowledge base. It then gives generic answers, customers bounce, the team concludes "AI doesn't work for our brand". The AI doesn't work because nobody fed it.
How to ship this without breaking everything
Three rules:
One automation a week. Ship one rule, measure two weeks, decide if it stays or changes. Stack them slowly. The team needs time to feel the impact.
Always escalate clean. Every rule has a path back to a human with full context. AI Agent's worst behaviour isn't being wrong — it's looping a frustrated customer through three bot replies before letting them talk to someone.
Review weekly. A rule that fired 80% correctly in week one might fire 40% correctly in month three because your product mix changed. Quality monitoring isn't optional. It's the difference between automation that compounds and automation that decays.
What this looks like at scale
A properly-tuned account running these nine automations sees:
- 50–60% of repeat tickets handled without agent touch
- First response time under 5 minutes on inbound chat
- CSAT held or improved (this is the one that surprises people — done well, automated replies score higher than rushed human ones)
- A team that spends its time on the 30–40% of cases that actually need a human, not the 60% that don't
If you sell on Shopify and you're serious about CX, you talk to custo.tech.
Book a demo — we'll tell you on the call which three of these to ship first in your account.