AI Innovation

ETIM AI Classifier: Automated Product Classification

AI-powered tool that automatically classifies electrical products according to ETIM international standards from PDFs, images, and datasheets.

Published November 6, 2025
6 min read
Sivert Kjøller Bertelsen
AI
ETIM
Product Classification
Automation

The Challenge

Electrical product data comes in unstructured formats - PDFs, datasheets, images. Converting this into standardized ETIM classifications requires:

  • Deep knowledge of ETIM's 5,640 product classes
  • Manual reading of technical specifications
  • Understanding of 159 product groups and hierarchies
  • Hours of work per product

Manual classification doesn't scale.

The Solution

The ETIM AI Classifier uses GPT-4 to automatically classify products from any source:

  • PDF datasheets - Extract and classify technical specs
  • Product images - Analyze photos using Vision API
  • Text descriptions - Process natural language
  • JSON data - Structured product information

Result: Complete ETIM classification in 30-60 seconds.

ETIM AI Classifier Demo

Watch the tool classify an electrical product in real-time - from upload to complete ETIM classification with features.

How It Works

1. Input Processing

Upload any format:

  • PDFs are parsed for text extraction
  • Images analyzed with GPT-4 Vision
  • Text and JSON processed directly

All inputs normalized to unified format.

2. Smart Classification

Hierarchical approach:

Step 1: AI selects from 159 ETIM groups
        ↓
Step 2: AI narrows to ~40 relevant classes
        ↓
Step 3: AI picks exact ETIM class
        ↓
Step 4: Extract 20-50 feature values

This narrows from 5,640 classes to 1 accurate match.

3. Feature Extraction

AI reads the document and extracts values for ETIM features:

  • Validates against ETIM API - Uses official feature definitions
  • Confidence scoring - Each value includes certainty percentage
  • Smart suggestions - AI infers missing values from context
  • ETIM compliance - Respects allowed values and units

Example extraction:

FeatureValueConfidenceSource
Nominal voltage230V95%Datasheet page 2
IP ratingIP2088%Description
Number of poles390%Product specs

Real-Time Progress

Watch the classification happen live with Server-Sent Events streaming. See AI reasoning, group selection, and feature extraction in real-time.

ETIM classification results showing extracted features

Complete ETIM classification with confidence scores

Business Value

Unlock Data Trapped in Documents

1.1M deeplink documents × 30 seconds = Massive time savings

Cross-Validate Existing Data

Compare SAP classifications against vendor documentation automatically.

Automated Compliance

Check product data completeness against ETIM requirements.

Technology

AI Model: GPT-4o-mini for fast, cost-effective classification

ETIM Data:

  • Official ETIM API for feature definitions
  • Offline database (ETIM 10.0) with 5,640 classes
  • 159 product groups

Architecture:

  • Next.js 15 with Server-Sent Events
  • Vercel deployment (5-minute timeout)
  • Real-time streaming progress

Accuracy & Cost

Per classification:

  • ~8,000 tokens
  • $0.0024 USD
  • 30-60 seconds

Accuracy:

  • Group selection: ~95%
  • Class selection: ~85%
  • Feature extraction: 60-80% of features found

The hierarchical approach ensures high accuracy by narrowing from 5,640 classes to ~40 candidates before final selection.

"The key insight is using ETIM's hierarchical structure. Instead of asking AI to pick from 5,640 classes, we narrow it down step by step. This matches how humans classify products - you start broad and get specific."
SB
Sivert Kjøller Bertelsen

Use Cases

Product Data Management Enrich existing catalogs with ETIM classifications

E-commerce Integration Standardize product data across multiple platforms

Supplier Onboarding Automatically classify incoming product sheets

Quality Assurance Verify and validate existing ETIM data

Migration Projects Convert legacy product data to ETIM standard

Example: Circuit Breaker Classification

Input: Schneider Electric MCB datasheet PDF

Process:

  1. AI identifies: "Miniature circuit breaker"
  2. Group: EG000020 (Circuit breakers and fuses)
  3. Class: EC000042 (Miniature circuit breaker)
  4. Extracts 9 features including poles, current rating, voltage

Output:

{
  "class": "EC000042",
  "confidence": 85,
  "features": {
    "poles": "3",
    "rated_current": "20A",
    "rated_voltage": "230V",
    "release_characteristic": "C"
  }
}

Time: 45 seconds Cost: $0.0024

Future Enhancements

  • Batch processing for product catalogs
  • Additional product categories beyond electrical
  • Multi-language support
  • Custom feature extraction rules
  • Integration APIs for ERP/PIM systems

Try It Now

The ETIM AI Classifier is available at etim.sivertbertelsen.dk. Upload a product PDF and see classification happen in real-time.

Launch Classifier

Conclusion

The ETIM AI Classifier demonstrates how AI can transform unstructured product documentation into standardized, structured data at scale. By combining GPT-4's understanding of technical specifications with ETIM's international standard, the tool makes product classification accessible, fast, and cost-effective.

Built for: Product managers, data teams, e-commerce operations Powered by: OpenAI GPT-4 + ETIM International Cost: $0.0024 per classification

About This Article

Category: AI Innovation

Review Status: Published

Sivert Kjøller Bertelsen

Ready to Transform Your Product Data Management?

Let's discuss how Impact Commerce can help you achieve your digital commerce goals.