Improving accuracy and efficiency across Finance & Accounting

EXL helped the client implement a modular, AWS-based response management system, powered by our proprietary analytics & AI solution


Challenge

Our client, a leading healthcare services provider, owns and manages surgical facilities, physician-partnered hospitals, and diagnostic laboratories, as well as offering anesthesia, pharmacy, ancillary, and optical services throughout the United States.

The client had more than 50 email inboxes across its Finance & Accounting (F&A) functions, with 150k+ inbound emails every year. This required considerable effort for the F&A teams to effectively assign queries to the right resources and manually classify invoices. Accounts Payable (AP) processing, communication and the overall customer experience was hampered by the complexity of this process.

Solution

EXL helped the client implement a modular, AWS-based response management system, powered by our proprietary analytics & AI solution, to automate the time-consuming task of monitoring emails across F&A.

With more than 3,500 emails processed via EXL’s response management solution during the first 4 months of deployment, the goal of 85%+ accuracy in machine learning classification was achieved.

Key features of the AWS-based response management solution included:

  • Real-time email integration with Office 365 cloud service
  • Python-based AI/ML intent recognition and classification built on EC2
  • NLP-based sentiment analysis and prioritization
  • AWS-native solution providing resilience, scalability & efficiency
  • Intelligent auto assignment of requests to SMEs along with smart workflow management

The EXL solution leverages AWS VPC for security, EC2, Microsoft SQL Server on EC2, S3, Route 53, CloudWatch and integrates with AWS KMS, Secrets Manager and provides a single pane management experience. The design is scalable to support multi-region availability as needed.

With more than 3,500 emails processed via EXL’s response management solution during the first 4 months of deployment, the goal of 85%+ accuracy in machine learning classification was achieved.