Compose the complete product record — the matrix and every variant. Read the full enrichment instructions before drafting. Supplier export is attached.
Delivered to the agent
Case StudiesLuregod
Turning supplier spreadsheet chaos into a structured product catalog.
Luregod is a German fishing-tackle e-commerce store. Its catalog is assembled from dozens of supplier and manufacturer feeds that arrive as ad-hoc Excel files with almost no shared structure. Two AI agents turn that mess into a production-ready product matrix — every product lives on a complete landing page with canonical names, variant tables, attribute-rich filters, FAQs, and the navigation links needed to rank in search and surface in the right category slices.
002Enrichment Agent
Inside one enrichment pass.
The enrichment agent receives a prompt to compose the complete product record — matrix and every variant — and is told to read the full instructions before drafting. The matching supplier export is attached as source data. From there it reads, writes, and validates, finishing only when the record passes.
Reads the enrichment instructions to learn what a complete product record requires.
read_fileinstructions/enrich-product.md
Awaiting
Reads the supplier export and the existing catalog filters so the new record lines up with variants in the data and conventions already in use.
read_filesupplier/shimano-stella-fj-4000.xlsx
read_filecatalog/filters.json
Awaiting
Writes the enriched product JSON — matrix, variants, attributes, filters.
write_fileproducts/shimano-stella-fj-4000.json
Awaiting
Runs the validation script to confirm the enriched product.
run_scriptscripts/validate-product.sh
Awaiting
Enriched product ready for the catalog.
Awaiting
003The Pipeline
Multiple suppliers in. One landing page out.
Multiple supplier exports are merged on EAN into one canonical record. Two bounded agents take it from there — the first places the product on the right shelf; the second turns that draft into the ten fields a landing page actually needs.
SUP-01Source
Supplier 01
- Format
- Excel .xlsx
- Volume
- ~12,400 rows
SUP-02Source
Supplier 02
- Format
- Excel .xls
- Volume
- ~8,200 rows
SUP-03Source
Supplier 03
- Format
- CSV dump
- Volume
- ~5,900 rows
supplier_productfrom SUP-01
“Shimano Stella FJ 4000 Spinnrolle”
- EAN
- 4969022356789
- Shape
- 12 cols · DE
supplier_productfrom SUP-02
“STELLA 4000XG (Shimano)”
- EAN
- 4969022356789
- Shape
- 8 cols · EN
supplier_productfrom SUP-03
“Rolle Stella FJ 4000”
- EAN
- 4969022356789
- Shape
- 15 cols · DE
luregod_product
Merged canonical record
- EAN
- 4969022356789
- Sources
- multiple supplier rows attached
CATEGORISE
Agent 01AI Agent
Classification agent
Attaches a category and brand. Defines ruleset for the next agent.
luregod_product
Draft product
- Category
- Rollen / Spinnrollen
- Brand
- Shimano
ENRICH
Agent 02AI Agent
Enrichment agent
Category-aware reasoning that turns a linked raw supplier data into cohesive product data.
luregod_product
Enriched product
01Name
02Variant names
03Description
04Gallery
05Attributes
06Filters
07FAQ
08Comparison
09Translations
10Tags
004Continue
Want to see the service this pattern demonstrates?
Find the workflow that should be your first controlled AI pilot.
Start with one high-value process where manual coordination is slowing the business down.