Revolutionizing Support Case Management on AWS
A leading financial services firm implemented AWS Generative AI models for email classification and summarizations to enhance their customer experience and increase cost savings.
Our customer
Trusted by thousands of customers globally, a leading financial services firm draws upon extensive decades of industry expertise and utilizing cutting-edge fintech to help companies of all sizes navigate the future of global commerce.
- Industry: Financial services
The obstacles they faced
The customer support team at this organization faced a substantial workload, managing an overwhelming influx of approximately over 200,000 support emails every month. These emails originate from over hundreds of thousands of unique email addresses, indicating a diverse and possibly large customer base. The team's responsibilities include diligently reading and understanding each email, a process that requires careful attention to detail to accurately grasp the customers' issues and needs.
Once an email is reviewed, the team's next step is to create a support ticket. This ticketing process is crucial as it involves categorizing, prioritizing, summarizing, and assigning the issue to the appropriate department or support personnel for resolution. The average time taken from the moment an email is received to the point where a support ticket is fully created and logged is more than three hours during business hours and more than eight hours during non-business hours. This duration reflects the complexity and the thoroughness required in the process, ensuring that each customer's concern is properly understood and recorded.
How we helped
Rackspace worked with the customer on a multistage approach to streamline ticket workflows.
The first stage involved developing and putting into production an email filtering model using Amazon Comprehend's Custom Classification. The goal was to categorize emails into two classes: 'is support email' and 'is not support email'. This process is broken down into three key steps.
- Feature engineering: This step involved preparing and processing the email data to extract meaningful features that can be used by the classification model.
- Model training: In this phase, the prepared data from the feature engineering step was used to train the email classification model. This training process involved selecting an appropriate machine learning algorithm, feeding it the training data, and adjusting parameters to improve accuracy.
- Model batch inferencing on data: Once the model is trained, it's applied to a batch of data for inferencing. This means the model classified a large set of emails at once, categorizing them as either 'is support email' or 'is not support email' based on the learning it acquired during the training phase. The performance and accuracy of the model were evaluated in this step, ensuring it meets the desired criteria for effectively filtering emails.
The second stage involved building and evaluating an email classification model using Amazon Comprehend's Custom Classification to categorize emails into one of 400+ categories that includes several complex and detailed steps.
The third stage involved using Large Language Models (LLMs) in conjunction with Amazon Bedrock(Claude) for the purpose of summarizing relevant information from email communications, specifically tailored for case management.
The final stage included automation to autogenerate and assign tickets to the right categories.
Some of the primary services that Rackspace used as part of the solution included Amazon Comprehend, Amazon Textract, Amazon Bedrock, Claude, Amazon SageMaker, AWS Lambda and Amazon S3.
What we achieved together
- End-to-end automation of support case management: The customer was able to fully automate the lifecycle of support case management. The entire workflow was handled by an automated system, reducing the need for manual intervention.
- Cost savings from AI implementation: By utilizing AWS and AWS Generative AI services to automate processes traditionally performed by humans, the customer realized financial benefits due to reduced labor costs.
- Reduction in resolution times: The time taken to go from receiving a support email to resolving the issue was cut down from hours to minutes. This swift response time not only enhanced customer satisfaction but also allowed the correct team to address and rectify issues much quicker than before, leading to improved operational efficiencies.
- Improved decision making: The use of generative AI supplies insights through the summarization of ticket information and provides data-driven recommendations that inform decision making processes, thereby enhancing the business's ability to respond to challenges or opportunities in a timely manner.
- Cost savings: A significant financial impact was realized through the use of this solution for previously assigned human tasks. The customer reported cost savings that underscores the substantial reduction in personnel expenses thanks to the AWS generative AI solution that was deployed.
About Rackspace Technology
Rackspace Technology is the multicloud solutions expert. We combine our expertise with the world’s leading technologies — across applications, data and security — to deliver end-to-end solutions. We have a proven record of advising customers based on their business challenges, designing solutions that scale, building and managing those solutions, and optimizing returns into the future.