Case Study
Shopify AI Sales Agent | SMS Cart Recovery & Checkout
Abandoned cart emails mostly say the same thing: here is your cart, please come back. They do not answer the question that made the person leave in the first place.
This agent texts instead, and it can actually do something. It answers product questions, edits the customer's live cart through the Shopify API, adds items, applies a discount code, and sends back a checkout link that works.
How it works
It edits the real cart
The difference between a chatbot and a sales agent is whether it can change anything. When a customer says they also want the bindings, the agent calls update_cart with the product IDs and quantities, and the cart changes. The reply confirms what was added and hands back a fresh checkout link.
Structured outputs keep every cart edit valid. A malformed quantity never reaches Shopify, because it never gets past the schema.
Knowing the catalog
An agent that cannot answer a product question is just a reminder with extra steps. I built the Shopify app and the catalog sync, so the agent knows what is in the store, what it costs, and what is in stock when it replies.
On top of that sits campaign scheduling and follow-ups, so a conversation that goes quiet gets picked back up at a sensible time rather than immediately.
Where a person stays in the loop
There is a human approval toggle for anything sensitive. Auto-response can be on for ordinary questions and off for the cases where a store owner wants to read the message first.
It runs in production across stores. The platform it is part of publicly reports $620k in new sales for one customer.
More screenshots