Improved client service based on smart and fast insights mining.
The client is a US medium sized FMCG company aimed at increasing the bottom line through enhancing their customer lifetime value. They needed a bespoke customer service analytics AI. So the company was looking to implement a Natural Language Processing (Text Mining) solution to compile and analyze various sources of data from their Client Experience team and use the data to create insights to improve service to clients.
They needed to take multiple sources of unstructured data (text and voice recordings) received from different channels and establish a pipeline for data analysis (using NLP), text mining for getting business insights.
Client needed our team to :
AWS ARCHITECTURE DEVELOPED
Stage 1. We focused on building the Audio to Text pipeline and other data landing mechanisms. Multiple diverse data sources were expected so the infrastructure was supposed to be prepared to receive all of them.
Then, we set up the data quality check for all the incoming data separately (based on the data type).
The main task for our team was to get all the incoming information to the single point of analysis for customer service analytics AI. To do so, we designed the pipeline according to the client’s business requirements. The complete POC for the data landing and audio-to-text pipeline was ready in the first 2 weeks with infrastructure being deployed into the client’s AWS account.
Stage 2. We moved on to Ingestion and Processing. After getting all sources to a single point of analysis, it was being pushed down through the pipeline to be analyzed with Amazon Comprehend. Multiple languages were supported and specific models were being used due to the client’s requirements. This was one of the customer service analytics benefits of the solution.
Our team implemented fault detection and retry mechanisms on every step of the pipeline. The output data was stored on S3 and shared with the client. This task was completed by us in 4 weeks.
Stage 3. During the Data Analysis and Visualization stage, we stored data on S3 as an AWS Glue Data Catalog. All information stored on S3 was available immediately to be queried through Amazon Athena. We implemented Machine learning models through the AWS SageMaker.
The processed data and analysis was delivered to AWS Redshift for future visualization. The ETL part of the process we implemented through AWS Glue. Data in Redshift we shared with the client as well as PowerBI dashboards built on top of this data. This scope of work our team completed in 3 weeks.
As a result, our team of engineers has provided the client with a robust customer service text analytics for fast text mining and data analysis and analytics. Now, the solution mines the email and audio data and extracts valuable business insights.
The customer service analytics AI solution now helps the client continuously track customers’ sentiment, adjust services accordingly and promptly respond to challenges, increasing customer satisfaction.