KT Home shopping

KT Home shopping

황현지 2026.04.27

Project Information

Company NameKT
Project NameKT GSI Feature Enhancement_AI Assistant
IndustryRetailData FormatSQL, PDF, XLS
IntegrationKT GSI(Legacy), External APICloudAzure
LLMGPT-4.1, GPT-5 miniSolution VersionCXP Agent

AI Agent for Instant Natural Language Analysis of Home Shopping Broadcast & Viewership Data

Supports cross-channel analysis across 16 home shopping channels — covering products, revenue, ratings, call volume, scheduling, and viewer flow from adjacent terrestrial broadcasters.

Challenge

Despite operating a big data analytics platform (GSI), field teams struggled to leverage it effectively — constrained by complex multi-table structures and limited accessibility to specialized analysis tools.
Complex Data Architecture
Cross-channel analysis across 16 competing home shopping channels and adjacent terrestrial broadcasters required integrating numerous tables — spanning products, revenue, ratings, call volume, scheduling, and viewer inflow/outflow data
Heavy Dependence on the IT Department
Business planners and operations teams had no direct access to data — every analysis request had to be routed through the IT/data team, creating recurring bottlenecks with an average wait time of 72 hours (3 days)
Decision-Making Bottleneck Caused by Delayed Data Analysis
"Where exactly are viewers dropping off during the current broadcast?"
"Which channels are viewers coming from — and leaving to?"
Questions like these demanded immediate answers on the floor, yet the time required to extract and process the necessary data made real-time, data-driven decision-making structurally impossible

Solution

CXP built an AI Agent environment where anyone — regardless of role or technical background — can query live broadcast, viewership, and sales data in plain language and receive instant analytical results.
LLM-Based Text-to-SQL AI Agent
Developed an AI Agent that accurately converts natural language queries into SQL statements
Applied prompt engineering to enhance query precision and the Agent's ability to interpret user intent
GSI Schema-Optimized System Configuration
Built a query generation and execution environment fully optimized for the client's GSI database schema
Automatically retrieves relevant data based on user intent and delivers SQL-powered analysis results within 20 seconds
Business-Friendly Interface
Supports multiple interaction modes — dashboard, chat-based queries, and Q&A
Empowers non-technical users to independently conduct complex data analysis, regardless of SQL proficiency
Generates standardized weekly reports in under 1 minute

Impact

Time to complete analysis reduced: 72h → 5min
Operational efficiency improved: approx. 99%
Enabled business teams to conduct data analysis independently

How we work

Requirements Defined Through End-User Interviews: Started by asking real users "what questions do you want answered?" — using their input to define the scope and priorities of analysis
Schema Design Tailored to the Client: Directly analyzed the existing database architecture to map table relationships and query patterns needed for cross-analysis across 16 channels

Subscribe to Newsletter (TX Insight)

Langcode publishes a newsletter every month on the third Wednesday of the month (TX Insight).
To stay up to date with Langcode news and the latest IT trends and insights, click the Subscribe button below.