Best Practices and Instructions

Quantum Leap is a document digitization and traceability tool designed to extract, structure, and standardize supply chain data from existing documentation (e.g. waybills, purchase orders, invoices, import/export documents). This guide outlines the best practices and workflow for beta testing.

Access and Setup

• Login method: Use a Google or Microsoft email account to access the Quantum Leap platform.
  • Ensure the correct email address is shared with IFT so it can be added as an approved email to access to the tool.
• Read and accept user agreement.

File Upload

• The current upload limit is 50 documents per session
• Start with a small batch of 5-15 documents that describe a representative supply chain journey (harvest → import) of one product
• Upload any document containing traceability data, such as:
  • Harvest logs
  • Waybills/packing lists
  • Import or customs documents
  • Invoices

Data Extraction Workflow

Step 1 – Document Classification

• Once files are uploaded, the system automatically classifies each document by type (e.g. invoice, purchase order, import document, transport document, etc.)
• Users may be prompted to manually classify documents if the system is unable to make the determination.
• User should verify the automated classification before proceeding.

Step 2 – Data Extraction Validation

• The AI extracts KDEs such as product description, location name, volume, lot number, etc.
• Review the extracted data in the “Document Preview” window

• The system highlights uncertainties or missing data for user confirmation
• If prompted:
  • Validate/correct misread or misclassified data directly in the interface.
  • Manually add data the system is unable to identify.
• Do not add data that is not explicitly included in the document. Adding external information will cause increasingly inconsistent results for future extractions.

Step 3 – Entity Matching Verification

• Drag and merge unmatched references from left side of the window to the right if they refer to the same product, location, or event.
  • Collapse entities on the right for an easier view.
• Create new entities when an extracted entity doesn’t match an existing one. Do this by scrolling to the bottom of the right side of the screen.
• If two documents refer to the same product or location using slightly different names, merge them and select the most accurate or preferred name (e.g., “Apple Granny Smith” → "Granny Smith Apple”).

Traceability Graph

This view shows the supply chain mapped by the system. The floating menu on the left has a few more features:

• Documents: View and edit the data extracted directly from the uploaded documents.
• Traceability Data: View and edit traceability data, and link any unmatched references.
• Analyze: Displays compliance with regulatory traceability frameworks.
  • In the upper right corner under Type, users can choose between Core Traceability, GDST and FSMA requirements.
  • Missing KDEs are flagged by event.
• Export: Users can download results in Excel or JSON formats.
  • Each export includes metadata on document sources and extraction results

General Best Practices

• Start small: Begin with 5-15 documents from one shipment to simplify verification steps and familiarize yourself with the application.
• Include Transformation documents: For each processing or packaging step that converts one product or lot into another, include a record that explicitly identifies input and output product data (i.e. input lot code, input quantity, output lot code, output quantity, etc). Transformation records create the lot-to-lot links Quantum Leap needs to connect events and build a complete traceability graph.
• Prioritize accuracy during data validation; AI learning depends on clean feedback.
• Don’t over-edit: Only correct what’s incorrect; avoid adding inferred data.
• Record feedback: Keep a log of any issues, bugs, unclear processes, or missing fields to share in our next feedback meeting.