E-commerce
Product catalogue enrichment
Descriptions and attributes are completed from supplier data and approved before they are published.
Fictional scenario. An example situation that illustrates the approach. It is not a client project and describes no measured results.
Business problem
A fictional shop imports thousands of items from suppliers with incomplete data. Descriptions are inconsistent and filters do not work.
Assumptions
- Suppliers provide files or APIs.
- Attribute rules are written down.
- The catalogue has an API for drafts.
Constraints
- Prices and stock are never changed by the AI.
- Nothing is published without approval.
- The brand voice is respected.
Reference architecture
Components by stage — an illustration, not a running system and not a certified design.
01Input
- Supplier data
- Current catalog
02AI and rules
- Attribute normalisation
- Draft descriptions
- Contradiction check
03Review
- Approval queue
04Output
- Publishing to the shop
- Change log
The diagram as a list
| Stage | Components |
|---|---|
| Input | Supplier data, Current catalog |
| AI and rules | Attribute normalisation, Draft descriptions, Contradiction check |
| Review | Approval queue |
| Output | Publishing to the shop, Change log |
Interfaces
| From | To | Via |
|---|---|---|
| Suppliers | Intake | Files or APIs |
| Enrichment | AI model | Adapter |
| Queue | Shop | Draft API |
Data flow
- Supplier data is normalised.
- Missing attributes are suggested with their source.
- Descriptions follow the style guide.
- A person approves in batches.
- Changes are recorded with a version.
Security boundaries
- The shop key can only write drafts.
- No customer data in the process.
- Every publication is logged.
Where a person approves
- Every batch of changes before publishing.
- Changes to the attribute rules.
Delivery phases
01 · Discovery
Categories, attributes, style.
02 · Pilot
One category, drafts only.
03 · Expansion
More categories and suppliers.
04 · Operation
Evaluation set and quality control.
Limitations
- Wrong supplier data is carried over.
- Technical specifications need checking.
Trade-offs
- Batch approval — faster but less attention per item.
- Strict style — less creativity.
Related services
Related assessments
Integration patterns used
Workshops and assessments
Facing something similar?
We start with an assessment of where you are and propose a realistic first phase.