AI can draft product descriptions, care notes, and size guides from your maker's spec sheet. A two-person label in Melbourne now gets a drop of 30 pieces live in 2 days instead of 6.
AI can write your product descriptions, care notes, and size guides from the spec sheet your maker already sends you. Marlow & Thread, a two-person slow fashion label in Melbourne, used to spend about 6 days writing them for every new drop.
The label releases pieces in drops of about 30. Each piece needed a description, care notes, and a size guide before it could go up on the Shopify store. The founder wrote all of it by hand, so the pieces sat finished and photographed while the words caught up.
What we set up
We set up an AI helper (the industry calls it an agent) that reads the maker’s spec sheet for each piece. The sheet lists the fabric, the measurements, the trims, and the care details. From that, the AI helper drafts the description, the care notes, and the size guide, ready to load into Shopify. The size guide graphics come from the label’s Canva templates, filled from the same measurements.
Write from the spec sheet, not the photo
Most advice on AI product copy says to upload a photo and let the AI describe what it sees. A photo shows colour and shape. It doesn’t show the fibre content, the weight of the linen, or how to wash it, and those are the things a buyer of slow fashion wants to know. The spec sheet has all of them, so that’s what the AI helper reads.
Care notes matter for another reason too. In Australia, clothing has to carry a care label under a mandatory standard, which the ACCC sets out on its Product Safety Australia site. Your online care notes should say the same thing as the sewn-in label, so the founder checks every care note against the label before a piece goes live.
The tone came from examples
The first drafts sounded like every other fashion label, full of soft adjectives and nothing the founder would say. We tried describing the label’s tone in the instructions, and it barely moved.
What worked was better instructions (the industry calls it prompt optimisation) built from examples. We gave the AI helper twenty of the founder’s own past descriptions, plus a short list of words the label never uses. After that, the drafts needed small edits instead of rewrites. If you want AI to sound like you, show it your writing rather than describing your voice.
Where the label landed
A drop of 30 pieces now goes live in 2 days instead of 6.
Fit notes still come from the founder trying on each piece, because a spec sheet can’t tell you how a garment sits on a body. That part of every drop still takes the founder’s time, and only someone wearing the piece can write it.
If your new stock waits on copy, get your AI Starter Plan and tell it where your week goes.
