Reducing product description writing time from 5 minutes to 5 seconds with AI

Sellers use Windly to import product data from global marketplaces, localize listing content, and publish it to local channels. Because a single seller manages dozens to hundreds of products at once, publishing speed is a core part of Windly’s value.

Product description writing remained a major manual step. Before publishing, sellers had to turn imported product data into localized copy with accurate details and search keywords—taking up 30% of the listing workflow.

Click to add a description, or generate one with AI

Writing product descriptions remained a major manual gap. If left unresolved, it could delay publishing, interrupt bulk-listing tasks, and drive sellers outside Windly—weakening the value of its broader workflow automation.

To create a product description, sellers copied product data from Amazon, opened an external AI tool, wrote a prompt, and moved the generated copy back into Windly. Sellers still had to coordinate product context, AI setup, content transfer, and final review across disconnected tools.

Key Insight: The delay came from coordinating context, AI, and content across disconnected tools.

I translated the workaround into three design requirements: make description writing simpler, faster, and high-quality while preserving seller control.

01 Simpler

Remove repetitive transfer and setup.

Keep product context, AI generation, and editing in one workflow.

02 Faster

Reuse information Windly already has.

Automatically prepare product data and SEO context instead of asking sellers to copy data and write prompts.

03 High quality

Keep sellers in control of the final copy.

Let sellers review and edit generated content before it enters the listing.

Repetitive setup and content transfer were automated, while review and final approval remained explicit seller actions.

Before: Sellers had to copy product data, switch tools, and rebuild context as a prompt.

After: Since these steps required no seller judgment, I combined them into one Generate action using data Windly already had.

Before: Sellers could only edit and verify the result after moving it back into Windly, forcing them to commit before they could judge its quality.

After: Because quality still required seller judgment, I made the draft editable and moved Review before insertion, separating evaluation from commitment.

Before: Sellers manually copied and pasted approved content back into the editor, adding effort without adding judgment.

After: I automated this mechanical transfer while keeping Apply explicit, so sellers controlled exactly when AI-generated content changed the listing.

The quality of the workflow depended on the quality of the generated copy. I designed the output structure around the information sellers needed to publish a product listing quickly.

Within 30 days of launch, adoption and repeat usage increased while sellers reported positive early satisfaction.

What I learned

AI adoption worked when the feature was placed at the workflow break. Sellers did not need a separate AI tool. They needed a faster way to complete the task they were already doing.