AI-Assisted Value Mapping

When you have unmapped values in a value mapping table, FeedPrep's AI can suggest the correct mappings for you. It understands product data conventions, translations, abbreviations, and synonyms across 20+ languages — so it can handle supplier data from virtually anywhere.

How It Works

1. Open a value mapping page for any attribute (e.g. color, material, size).
2. Click the "AI Suggest" button.
3. AI analyzes your unmapped values against your existing approved values and suggests the best match for each one.
4. Suggestions appear inline next to each unmapped value. Review them and click Accept or Dismiss per value.

What Can AI Map?

The AI understands a wide range of real-world product data patterns:

• Translations: "Schwarz" → "Black", "Stahl" → "Steel", "Bois" → "Wood"
• Abbreviations: "BLK" → "Black", "SS" → "Stainless Steel", "XL" → "Extra Large"
• Synonyms: "Midnight" → "Black", "Crimson" → "Red"
• Multi-language: "Stål" → "Steel", "Cuero" → "Leather", "Holz" → "Wood"
• Internal codes: "CLR_029" → "Navy Blue", "MAT-OAK-S" → "Solid Oak"

Bulk Actions

When AI returns suggestions, you don't have to review them one at a time. Two bulk actions are available:

• Accept all mapped — accepts every suggestion that maps to an existing approved value.
• Accept all — accepts every suggestion, including ones that would create new approved values.

Batching

Values are batched in groups of 25 to handle large sets efficiently. If you have 75 unmapped values, the AI processes them in 3 batches. Each batch costs 1 credit.

Caching

AI suggestions are cached for 1 hour. If you refresh the page or navigate away and come back, the same suggestions will appear instantly without using additional credits. You only pay once per batch.

Cost

1 credit per batch of up to 25 values. For example, 50 unmapped values = 2 credits. See Understanding AI Credits for more details on how credits work.

Privacy

Your data is never used for AI training. AI suggestions are generated on-demand for your specific mapping context and are not retained beyond the cache period.