Portfolio ProjectEvidence attached

Ecommerce Product Copy Engine

A generation engine that writes on-brand copy per product from the real attributes, with a human review step before anything reaches the storefront.

DR

Daniel Roth

Support automation builder

Business problem

A store with 4,000 SKUs had supplier-provided descriptions on most products. They read identically to three competing stores and converted badly, but rewriting them by hand was never going to happen.

Solution

A generation engine that writes on-brand copy per product from the real attributes, with a human review step before anything reaches the storefront.

What was built

  • •A brand voice profile built from 20 hand-written examples
  • •Bulk generation runs by category with cost per run visible
  • •A review interface showing old and new copy side by side
  • •Publish-back with a one-click revert

Architecture / workflow

  1. 1Catalogue synced from the store nightly
  2. 2Copy generated per SKU from attributes plus the brand profile
  3. 3A reviewer approves, edits or rejects in a queue
  4. 4Approved copy is published back and the previous version stored

Integrations

ShopifyAirtable

Tools / stack

OpenAINext.jsSupabase

Screenshots

Screenshot slot
Screenshot slot

Demo

No demo added yet. Builders can attach a short video.

Results

No verified results shared. AgeBridge only displays real, builder-reported outcomes.

Evidence

  • Screenshots · Review queue, before and after copyReal screens from the build, not mockups.

Proof

Proof score

15/100

Add demo video — 60 seconds showing the thing actually running. The single strongest proof.

Not reviewed yet. Evidence is shown exactly as the builder attached it.

Decisions and tradeoffs

  • •Generation uses product attributes only. No invented materials, certifications or claims.
  • •Everything is revertible, because copy changes affect search rankings.
  • •Review is per category, so a reviewer builds rhythm instead of context switching.

What went wrong

The first run invented care instructions for fabrics. The prompt now refuses any attribute not present in the source data.

What I'd do next

Compare conversion on 200 rewritten products against 200 untouched ones.

Need something like this?

Talk to the builder behind the project, or build your own version.