Building advanced machine learning algorithms to streamline AP processes for a financial services enterprise
| Client | Industry | Solution Provided | Technologies Used |
|---|---|---|---|
| Financial services enterprise | Financial Services | Data & AI | Optical Character Recognition algorithms; Amazon Textract, Amazon S3, AWS Lambda, AWS Glue, AWS Cloudwatch, Snowflake, Apache Airflow, Tableau, TensorFlow, PyTorch, Scikit |
The Challenge
This global financial services provider processes an average of 1,500-2,500 invoices daily across five global locations, and human errors in data input were leading to delays and costly penalties. The client sought to automate invoice data extraction and interpretation with advanced machine learning algorithms to streamline the AP process, reduce errors, and avoid late fees. Finally, the client also required a solution that could integrate with data-driven workflows and custom ERP systems to expedite AP processing and manager approvals.
The Gorilla Logic Approach
- Gorilla Logic developed advanced Optical Character Recognition algorithms and implemented other tools for data extraction, validation, and transformation, including:
- Amazon Textract, Amazon S3, AWS Lambda, AWS Glue, & AWS Cloudwatch for automated text and data extraction, automated data processing and transformation, storage, and monitoring, leading to cost savings in labor time
- Snowflake for efficient data warehousing, creating faster data retrieval times and improved data quality
- Apache Airflow and Tableau for managing data workflows and creating visual reports for executives to more easily understand the AP process and its impact
- Machine learning models & frameworks like TensorFlow, PyTorch, and Scikit to accurately predict missing data elements and accelerate processing times. The models were trained, tested, and validated on a dataset of one million invoices.
Business Outcomes
In an ongoing partnership, these efforts are resulting in:
Improved Accuracy: The machine learning model improved its AI classification performance from 70% to 98.92%, minimizing errors in data extraction and processing.
Accelerated Processing Times: The new solution processes 100-150 invoices in the same amount of time (10-12 minutes) it used to take to process one invoice.
Cost Savings: Reduced labor costs associated with manual data entry and validation by neary $1.1 million.
Enhanced Decision-Making: Optimized operations and enhanced data quality have provided valuable insights, empowering faster, better decision making.
