Automated Email Replies for a FinTech Firm Enhancing Customer Engagement and Resolution Time

Solutions

• The raw client data undergoes preprocessing to ensure quality, addressing inconsistencies, missing values, and irrelevant information. 
• Preprocessing involves setting up an Amazon SageMaker instance, understanding the data, and scripting data acquisition, cleaning, transformation, and validation. 
• Processed data is stored in Amazon DynamoDB, serving as the customer data source for the model. 
• Two automated flows are created using Power Automate to retrieve newly arrived emails and reply to them. 
• Amazon API Gateway integrates with AWS Lambda to trigger the new mail arrival flow in Power Automate through an API. 
• A validation layer in AWS Lambda verifies customer emails using PAN Number or FOLIO number against stored customer data in Amazon DynamoDB. 
• The “Anthropic Claude v2” model is chosen for generating human-like text responses to customer emails. 
• Parameters are set to optimize the model’s performance and reduce randomization and probability for text generation. 
• An Amazon DynamoDB Table records previous conversations using the conversation ID of the email chain, enabling the model to find continuity and respond to continuous queries. 
• AWS Lambda functions are configured to make API calls to the model for email response generation and processing before transmitting responses to the Power Automate reply flow through REST API.