One of the most important finance processes for a consumer products company is the resolution of trade promotion deductions (claims) by their account receivables (AR) department. Trade promotion deductions are special incentives that are offered to retailers to increase the demand for products. For example, a tire manufacturer may send a special offer to tire retailers like buy three, get one free. Rather than pay for all the tires ordered, the tire retailer only pays for a portion of the tires, per the details of the special offer. The AR team for the consumer products company must confirm the payment and reconcile what is due from the tire retailer based on the offer. It is a complex process that involves many parties and systems and requires time to assess and validate. By reducing the time for collecting claims, the AR team can focus on important tasks, such as an analysis of the promotion return on investment (ROI) or process optimization. A faster resolution of deductions can accelerate the consumer product company's access to liquidity. This reference kit may assist you in developing a model to help automate the processing of claims.
In collaboration with Accenture*, Intel developed this AI reference kit to extract information from claims documents to categorize the claims. Paired with Intel® software, this kit may help customers with developing models to accelerate the resolution of accounts receivable claims for trade promotion deductions. This reference kit includes:
- Training data
- An open source, trained model
- User guides
- Intel® AI software products
At a Glance
- Industry: Consumer packaged goods
- Task: Extract information from claim documents using OCR and categorize the claims into categories
- Dataset: Product and pricing information with images
- Type of Learning: Supervised learning
- Models: Convolutional recurrent neural network (CRNN), CNN
- Output: Extracted text, claim identified as valid or invalid
- Intel® AI Software Portfolio:
- Intel® AI Analytics Toolkit (AI Kit)
- Intel® Optimization for PyTorch*
- Intel® Neural Compressor
This model was trained using TensorFlow* 2.9 with Intel® oneAPI Deep Neural Network Library (oneDNN) turned on by default.
Performance was tested on Microsoft Azure* Standard_D8_v5 using 3rd generation Intel® Xeon® processors to optimize the kit.
This AI reference kit demonstrates the advantages of using the Intel® AI Analytics Toolkit for building a pipeline for trade promotion deductions. A two-stage AI solution has been proposed to extract the useful information from claim documents using OCR, and then categorize the claims, with both stages using deep learning. To derive the most insightful and beneficial actions to take, data scientists need to study and analyze the data generated through various feature sets and algorithms, thus requiring frequent reruns of the algorithms under many combinations of parameters. To do this, software optimizations are recommended to use all the hardware resources efficiently. The savings gained by using Intel Optimization for PyTorch and Intel® Optimization for TensorFlow* with oneDNN optimizations may help a data scientist to explore models more efficiently, thus leading to more targeted solutions.