Cloud platforms beyond infrastructure and application support: Business process improvement using Cloud-based Machine Learning

Cloud platforms are well known for providing robust infrastructure, security, resilience, and application hosting services. Clouds take on managing and hosting IT infrastructure and application platforms, allowing client organizations' IT to concentrate on implementing and supporting their internal and external business processes. Beyond the basics, Cloud platforms also offer other lesser-known services that can provide functionality well beyond the basics.

One of our clients in Logistics Management engaged us to review their existing processes for Customs clearance of overseas packages to eliminate potential failure points, improve reliability, and help them transition the entire platform to the Cloud. Upon completing our review, we had built a Machine Learning solution that significantly streamlined our client's package processing and clearance system through the US Customs. The solution eliminated the bottleneck for their existing process that relied on employees with specialized knowledge to perform the tasks manually. It also reduced their package screening cost by more than 50%.

This client had processing centers worldwide with highly trained staff reviewing the descriptions for the contents of packages shipped from China to the United States. The process involved matching the content description to the rules established by US Customs to determine their importability. The entire process of scheduling batches of package content descriptions for review, assignment of these batches to various employees in processing centers, and completing the assessment, acceptance, and rejection utilized the email server. This design had made the email system a critical element of the entire process, not atypical of how email systems have become workflow engines for many organizations. The client's first order of business was to harden the email server and prevent outages as much as possible by moving the platform to the Cloud and taking advantage of Cloud-based email system resiliency.

However, as we were auditing the entire process to eliminate other points of failure, we noted that an ML-based system could determine the package importability with much more accuracy and reliability at a much higher speed. We presented a proposal to the client based on an ML discipline known as Natural Language Processing (NLP) and Text Analytics. A system built using NLP algorithms could analyze the information much faster than processing center personnel, drastically increasing the number of packages reviewed, accepted, or rejected. Service center employees also tended to make inconsistent decisions regarding an item's acceptance or rejection due to different interpretations of the item's description. Using NLP-based logic would allow the client to accept or reject packages for shipment consistently. This approach would also eliminate the need for the email-based workflow process.

We gained approval for the system to be developed and subsequently completed and deployed the solution in less than six months. As part of their deployment, NLP-based systems require training to learn the context for the words and phrases; for example, without context, content descriptions that contained words such as dagger, gun, and so on could be interpreted as weapons. These terms can also describe jewelry items such as a miniature dagger for a charm bracelet or a necklace. NLP's ability to differentiate and understand words based on context allows it to replicate human intelligence and judgment calls. The same challenge would be impossible using typical computer logic that counts on the true/false nature of data and precise matching of single words to make decisions. The existing processing center personnel were responsible for the system's ongoing training to ensure its continuous improvement.

All leading Cloud platforms offer Machine Learning Services, such as Natural Language Processing and Text Analysis, that can provide the solutions to gain insight into any process that involves unstructured data. Dispute resolution, technical support, and customer service centers are just some of the business processes that can take advantage of these services to augment or supplant customer interaction centers. These business processes can use ML-based services to turn unstructured natural-language-based communication into invaluable customer-centric insight. It allows comprehensive analysis of information hidden in texts, emails, and phone calls to determine customer sentiment in general or related to specific products and services. NLP and Text Analysis tools can also help organizations identify areas of improvement for their products and service and monitor their service center personnel's behavior to determine additional training requirements.