Wednesday, July 24, 2019

Using Machine Learning to Improve Client Services



Key Challenges

   Reduce support ticket count.
   Identify customer support ticket trends.
   Reduce developer effort.

High Support Ticket Count Drives Improved Management Model

Our client is a large software company that provides monitoring, diagnostics, and analytics services. Our client’s customers faced several challenges using and understanding the services. Customer challenges resulted in numerous support tickets. Our client needed us to identify the service features that customers struggled with.

At the time, our client’s developers oversaw support ticket management. But support ticket management, along with the developers’ other work duties, required a great deal of time. To reduce the amount of time the developers used to evaluate, categorize, and allocate the support tickets, we implemented an AI-powered ticket management model.

Resolving Support Tickets Faster

To reduce the number of support tickets, we first evaluated the resolved and active tickets. We categorized the support tickets by issue type. We then refined the support ticket categories by creating subcategories based on channel source and issue frequency. We fed the resolved and active tickets, along with the ticket categories, into our machine learning model. The model identified trends in support tickets, quickly learning to recognize and categorize issues.

After two weeks of training, we implemented our support ticket management solution. We used Kusto to fetch support tickets, Microsoft Flow to automate e-mails and update incidents, and Microsoft Power BI for reporting purposes. We used the machine learning model to scan the support tickets and improve response time.

The machine learning model matched incoming support tickets to previously answered tickets with a high degree of accuracy. The model then fed data to Microsoft Flow, providing the most relevant support ticket answers in Teams. To address support ticket trends, we automated task creation. Automating task creation enabled our client's developers to understand customer challenges without directly managing support tickets. Automatic task creation also improved feature planning. Now, feature planning is directly tied to the number of support tickets. We also used our machine learning model to contribute to ongoing feature documentation. 

Improved Support Model Reduces Tickets and Improves Client Services

Our support ticket model resulted in three benefits for our client. First, our callouts in weekly meetings helped our client’s developers reduce bugs. Second, our support ticket model automatically analyzed support ticket trends, which reduced the number of support tickets. We reduced the volume of support tickets by 50% in six months. Third, our client’s developers used the support ticket model to proactively implement features of their tools. Using the support ticket model reduced client effort and cost by 30% and improved their service quality.

Our support ticket solution also increased the number of first- and third-party users utilizing our client’s service. Users enjoyed the speed at which our support ticket model brought new and improved features to their fingertips.