Case study: automating ~64% of support messages in made-to-order e-commerce

Summary: After connecting SemAgent to its Shopify store, made-to-order t-shirt brand Tonmode has roughly two-thirds of support messages resolved before they ever reach a human. This post walks through the setup steps; full case: Tonmode case study.

The problem: paying for questions you've already answered

In a made-to-order model, delivery takes longer than in standard e-commerce, and customers can't help asking. Tonmode's inbox came down to three questions: order status, size and print, and returns. Every answer was already written down — in the FAQ, on the returns page, in the Shopify dashboard. The information wasn't missing; a way to deliver it instantly was.

Setup: 3 steps, ~1 day

1. Knowledge base (first hour): the site was crawled; the size chart and shipping-returns pages were indexed; the automatic FAQ generator produced ready answers to the classic questions.
2. Shopify connection: "where is my order?" is now answered live — carrier, tracking number, estimated delivery.
3. Escalation rule: a damage/payment/anger signal → hand off to a human with a conversation summary.

Results

• ~64% of messages resolved without reaching a human, with sources cited.
• First response is instant — the time-zone problem is gone.
• The ~2 hours of weekly reporting effort dropped to zero.

Three questions for your store

• What percentage of your inbox is questions "whose answer is already on the site"?
• How much piles up overnight, waiting for someone in the morning?
• Who compiles the weekly reports, and in how many hours?

For a similar setup: e-commerce solution · pricing.