IA Djust

Extraction

The Extraction phase is responsible for transforming unstructured or semi-structured input data into usable and standardized information.

Before any processing occurs, a data preparation step is applied to ensure privacy and compliance by anonymizing sensitive information. Once secured, the system uses AI-based models to extract relevant business data.

This phase includes:

Identification and extraction of key information (e.g., products, quantities, references)
Structuring of extracted data into a normalized format
Validation of extraction completeness to ensure no critical data is missing

The result of this phase is a clean, structured dataset that can be reliably used for product matching.

Matching

The Matching phase aims to identify the most relevant products based on the information extracted in the previous stage.

This phase relies on artificial intelligence to analyze product references, names, and contextual information in order to accurately associate extracted data with existing products. The AI model evaluates multiple signals simultaneously, including known identifiers, textual similarities, and contextual relevance, to determine the best possible match.

Each potential product candidate is evaluated using a unified confidence scoring mechanism. This score reflects the overall relevance and reliability of the match produced by the AI.

Products that meet or exceed the defined confidence threshold are automatically validated and selected. When the confidence level is insufficient, the system proposes alternative product recommendations to ensure accuracy while avoiding incorrect selections.

This AI-driven approach provides a balance between efficiency and precision, enabling reliable product identification while minimizing manual intervention.

Order Creation

The Order Creation phase is the final step in the workflow. Once products have been successfully validated, the system generates the corresponding order automatically.

This phase includes:

Adding validated products to the cart
Verifying quantities, pricing, and availability
Creating the final order in the system

Only products that meet the defined confidence and validation criteria are included in the order. This ensures a high level of reliability and reduces the need for manual intervention.