Robotic Process Automation: Case Studies

Goals

To identify use cases of companies that have automated the PO data entry process (sent via fax) using RPA software or AI.

Early Findings

Cisco

  • The issue: Manual processing of orders was found to be costly and stressful. Employees had to validate SKUs, configurations, discounts, billing addresses, etc. Manually entering a single order took hours- or several days for large purchase orders of 100 pages or more.
  • Solution: In May 2019, Cisco rolled out a fax order entry automation process. The technology used was robotic process automation with deep learning.
  • Process:
- To automate fax orders, Cisco was faced with three challenges: extracting data from unstructured document images, validating the accuracy of the data, and entering the data into the e-commerce system.
- Combining machine learning (Naïve Bayes and spaCy) with a deep learning (Faster R-CNN) algorithm proved to be a success.
- For graphics processing, the company used the Cisco UCS C480 ML server, which is optimized for machine learning and has eight NVIDIA V100 GPUs. In addition, an internal solution was built through their private cloud that could offer ‘GPU as a service,’ billing different departments according to their usage.
- The accuracy of PO data is validated through human agents and the machine learning algorithms are readjusted for continuous improvement.
  • Results:
- Productivity increased by 50% (65% of POs were automated, reducing annual support costs by 50%)
- More than 90% accuracy.
- Increased customer satisfaction as a result of 50% faster order cycles and automatic acknowledgment of order receipt.

Proposed next steps:

You need to be the project owner to select a next step.