E-Commerce
Product Description Generation
DTC brand, 200+ SKUs per seasonal launch. 30–60 minutes per SKU, 55% Creative Director revision rate, contractor copywriters every launch cycle. A prior AI tool had failed. The difference this time was implementation, not technology.
The situation
The content team was producing product descriptions for 200+ SKUs per seasonal launch. Each SKU required four content types: a long-form description, a short-form version, a bullet list, and an SEO meta description. The process took 30–60 minutes per SKU from brief to copywriter draft — before Creative Director review. The Creative Director was revising more than half of what came to her.
The company had tried an AI writing tool before. It produced generic output that didn't sound like the brand. The Creative Director had killed it within two months. That experience was in the room throughout this engagement — it was not a compliance issue or a tech failure, it was a trust failure. The prior tool had been deployed without training it on who the brand actually was.
The copywriter had built informal workarounds: category-specific phrase lists, a personal note file with the Creative Director's most common redlines. These weren't a problem. They were a proof of concept — the institutional knowledge existed, it just wasn't in a form the tool could use.
The approach
The core design decision was to make the Creative Director the system's governing authority. She co-authored a brand voice governing document — a structured file that defines tone by product category, lists approved and prohibited phrase patterns, and specifies the approach for each of the four content types. That document became the system prompt. When the brand voice evolves, she updates the document and the system prompt follows.
The tool — ContentDraft — pulls product data from the Shopify catalog, takes keyword input from the Marketing Manager, and generates all four content types per SKU in a batch. Output lands in an Excel review file. The copywriter reviews each SKU — editing, approving, or flagging for rewrite. The Creative Director reviews the copywriter-approved content. Nothing publishes automatically.
One finding from the pilot changed the revision rate more than any prompt change: context notes. When the Marketing Manager added brief positioning notes for key products before the generation run, the Creative Director's revision rate on those products was 9%. On products without context notes, it was 21%. Structured input investment reduces review time downstream.
The result
The seasonal launch batch — 188 SKUs — was completed in 10 working days. The prior launch had required four-plus weeks with contractor copywriter support. No contractors were used. Average time per SKU, including generation, copywriter review, and Creative Director approval, was 8.4 minutes.
The Creative Director's revision rate on the full launch batch was 19% — down from 55% at baseline, and below the 20% target. Her assessment: "It's more consistent than what we were doing before. The brand voice is actually more uniform across SKUs now." An SEO audit 30 days post-launch found 96% of new product pages included the primary target keyword, compared to an estimated 50% baseline.
The Creative Director had been the most skeptical stakeholder. By the end of the pilot she had submitted process improvement suggestions and signed off on full adoption. Her sign-off: "The tool isn't replacing the copywriter's judgment — it's giving her a better starting point and making my review faster because the baseline is higher."