Objective
Reduce the effort overheads for the Call Center Representatives by making data finding, summarization, and root cause analysis a seamless automated process.
Problem Statement
A US-based Credit Reporting agency that deals with thousands of tickets daily/weekly from various B2B customers (like Banks, Credit Agencies, and other Financial Institutions) that use their Credit Reporting platform or software. They currently use ServiceNow as their primary Incident Management platform and move the ticket and incident data to Snowflake for deeper analytics around incidents, tasks, changes & outages. Typically, for each incident, Call Center executives need to dive deep into the tickets to understand ‘what’ happened and ‘why’ it happened. They then need to inform the customers about the current state of the issue, root cause, SLA for a fix and other similar information. Most of this is currently manual and takes a significant human effort.
Solution
With the advent of Snowpark and seamless deployments of LLMs in Snowflake, it’s now easier than ever to talk to your data. LLMs can be used to convert Text-to-SQL which can help read data from the tables with high accuracy. We built this solution on containers and deployed on Snowpark Container Services (SPCS) so as to have a decoupled architecture. There’s a container for Streamlit that will act as the front-end UI for the Call Center Representatives to interact with & another container for the LLM of choice.
The Streamlit UI provides a mechanism to record feedback on the responses through a ‘Thumbs Up’ or ‘Thumbs Down’ menu option that will be recorded to improve model accuracy.
Beyond the user interaction on live data, scheduled batch jobs will join the datasets from different tables, such as Incidents, Tasks, Changes, and outages, and run LLM models on top of that data to create a well-documented Root Cause Analysis (RCA) for the IT operations teams to investigate.
Architecture Diagram
Streamlit App
Business Impact
This solution aims to save a significant amount of human effort required to dig deeper into these long trails of tickets across incidents and customers. Call center representatives will now be able to answer live queries more effectively, thereby improving the overall customer experience and satisfaction.
The IT operations team is also looking to save on a lot of time and effort required to connect the dots between Incidents, Changes & Outages and automate the creation of RCA documents.
