Automatic Customer-Service Rollout With Our Own Model
Gorgias stays the inbox; our own model writes the replies for many Shopify stores, using each store's macros and knowledge. It only sends on its own above 0.80 confidence and never touches money.
362 tickets closed by the bot in 30 days at one brand · ~7 cents per closed ticket vs ~90 cents for Gorgias AI Agent.
TL;DR
- Gorgias stays the inbox. Our own model writes the replies, for many Shopify stores, using each store's macros and knowledge base.
- A reply sends on its own only when the category is allowed and the confidence is at least 0.80. Refunds, cancellations and order changes go to a person.
- At The Love Blanket, the bot closed 362 tickets in 30 days, at an estimated seven cents per closed ticket, against about ninety cents for the Gorgias AI Agent.
The problem
After setting up Gorgias for brand after brand (see the setup process), the same pattern kept repeating: most tickets are the same handful of questions, and helpdesk AI add-ons are priced per resolution and configured from zero for every brand. That gets expensive as volume grows.
How it works
- 1Read the whole ticket, including what was already promised and who is assigned.
- 2Match a macro first. If the store's macro library has a fitting answer, it is used.
- 3Draft from the store's knowledge if no macro fits: tone, shipping, returns, the macro library and the rules already on the account.
- 4Score every draft. Each draft gets a confidence score.
- 5Auto-send only when it's safe: the category is allowed and the confidence is at least 0.80.
- 6Queue the rest for a person. Refunds, cancellations and order changes always go to a person, who still sends those.
Rolled out store by store
Before anything goes live, the tickets people have already closed are run as a dry run, and the sheet of drafts is reviewed. Each store then moves from dry run, to shadow, to live, on that store's own voice and policies.
Why our own model
- One pipeline across stores. The inbox is Gorgias; the model and the rules are ours.
- It can't invent a second policy. It is trained on that store's documents, macros and rules.
- Money stays with a person. Anything that touches money or the order is escalated.
- Cost. About seven cents per closed ticket, against about ninety cents for the Gorgias AI Agent.
Results
| Metric | Value |
|---|---|
| The Love Blanket, last 30 days | 1,091 tickets created · 1,112 closed · 362 closed by the bot |
| People on the same account | about 20 tickets an hour; the native Gorgias AI Agent row was empty |
| One dry run of 142 tickets | 71 auto · 58 queued · 2 escalated · 11 assigned · nothing sent |
| Estimated cost per closed ticket | ~7 cents vs ~90 cents for Gorgias AI Agent |
| Live stores | Clear Drop runs live on a schedule. Saber X is in dry run until a person approves it. |
Connected projects
- Builds on → Gorgias AI customer-service setup
- Store data access shares tooling with → Supplier → Shopify inventory sync
Images.
Ready to build your AI project?
Free 30-minute discovery call. No pitch. Just a conversation about what AI could do for you.
- Free · 30 min
- A concrete first step you can act on
- In German or English, online