GenAI + Product Development

Upret-AI-N

GenAI Self-Serve Analytics Tool

RoleIC Owner & Developer
TeamIN CX Concessions Analytics
Timeline4 Weeks
StakeholdersIN CS Leadership, CX Analytics Consumers, Cedric Platform Team
Tools
Claude LLMPrompt EngineeringPythonSQLCedric Platform
240 hrs/yearProductivity Savings
32Users (4 weeks)
80+Prompts Answered
CX-firstFirst of its kind

The analytics team spent an average of 4 hours per week building ad hoc deep dives for stakeholders or helping consumers debug self-serve processes. Additionally, 3 hours weekly went to repetitive manual consolidation of WBR inputs across program owners. Teams were bottlenecked on analyst availability for even standard queries.

I built a repository of standard code snippets and data sources used within CX analytics — covering contacts, concessions, and customer signals. Using this repository, I trained a Cedric-based AI assistant with prompt engineering practices to make it context-aware for any variety of CX analytical use cases. I introduced standard code writing practices and Amazon-specific SQL awareness to improve the usability of outputs.

The bot amassed 32 unique users and answered over 80 unique prompts in its first 4 weeks. The analytics team now has an annualized productivity savings potential of 240 hours. It became the first self-serve query builder in IN CS.

Why It Mattered

This fundamentally changed how teams interact with data — from 'ask the analyst' to 'ask the bot.' It removed the analytics team as a bottleneck for standard queries whilst maintaining domain-specific accuracy. The tool is now being explored for expansion to other IN CS analytics teams.