AI enrichment

How enrichment works

For products with content gaps, ShelfReady sends the existing product data to Claude in small batches and asks for compliant rewrites. Each proposal comes back with a rationale explaining why the change was drafted, and is stored as a pending proposal — never applied automatically.

What gets enriched

A product is an enrichment candidate when any of these hold:

• the description is missing or thin (under 80 characters),
• the brand is missing,
• the title is missing or over the 150-character ACP cap.

These mirror the content rules in the audit, so approving proposals directly improves your readiness score.

{
  "field": "description",
  "proposedValue": "The Ridgeline 2P is a freestanding two-person tent built for three-season backpacking…",
  "rationale": "Description was missing or thin.",
  "status": "pending"
}

Reviewing & approving

Review proposals in Dashboard → Enrichment. Each shows a before/after view. Approve to apply the change to your canonical catalog (feeds pick it up on the next render); reject to discard it. Rejected proposals are not re-drafted unless the underlying product data changes.

Guardrails

• Enrichment only proposes title, description, and brand. Prices, GTINs, SKUs, and URLs are never invented or altered.
• Proposals use structured output — they arrive validated, never scraped from free text.
• Every proposal cites a rationale so you can audit the reasoning before approving.