Korea Industrial Complex Corporation
| Company Name | Korea Industrial Complex Corporation |
| Project Name | An AI Chatbot Platform for Tailored Support Program Recommendations |
| Industry | Public | Data Format | HTML, PDF |
| Integration | Support Program Guide Webpage | Cloud | Azure |
| LLM | GPT-4 | Solution Version | CXP Agent |
An AI Chatbot Platform That Unifies Thousands of Support Programs into Tailored Recommendations
An AI chatbot platform that automatically collects and standardizes support program announcements scattered across multiple sites, recommending the right programs through a single natural-language question
We built an environment where companies can explore support programs matched to their industry and conditions through conversation alone — and drill into the details of any specific program among thousands of announcements via multi-turn Q&A.
The support program information companies need was scattered across multiple institutional sites, forcing them to visit each site in person to find announcements, compare information in differing formats, and struggle to discover the programs they were eligible for.
Scattered announcements and inconsistent information structures
Support program information was spread across many sites, forcing companies to visit each URL one by one to check it. Announcement formats and data fields (schemas) differed from platform to platform, leaving no unified foundation for searching and comparing multiple announcements in one place. Difficulty selecting programs that fit a company
The right program varies by a company's industry, business type, and funding purpose — yet eligibility, application periods, and funding amounts all had to be compared across individual announcements by hand, an ever-growing burden. We built a data pipeline that automatically collects and standardizes fragmented announcements, paired with an AI chatbot that reads a company's requirements from natural language and recommends matching programs.
An automated pipeline for collecting and standardizing announcements
Built an automated pipeline that crawls and preprocesses support program announcements from each institutional site in real time. Standardizes key fields — program name, eligible recipients, application period, funding amount — into a common schema and automatically loads them into a NoSQL database. Tailored recommendations through natural-language condition parsing
Automatically extracts key conditions — industry, funding requirements, timeline, amount — from natural-language questions and converts them into search queries. Automatically routes between a simple single-program lookup and a multi-program comparison-and-recommendation path, depending on the intent behind the question. Matches a company's conditions against each announcement's eligibility requirements and presents an optimal shortlist of candidate programs, ranked by priority. Context-retaining follow-up Q&A
Even as follow-up questions about a recommended program continue — detailed eligibility, preferential conditions, how to apply — the chatbot keeps answering while retaining the prior conversation context. Provides links to jump straight to the original announcement and application page when needed. Unified support program information previously scattered across multiple institutional sites into a single conversational channel. Reduced condition-based search and comparison across thousands of announcements to a single natural-language question. End-to-end automation from collection to conversation. Crawling → preprocessing → loading → condition parsing → recommendation → multi-turn Q&A, composed as one automated pipeline. Experience designed around the information seeker. Conversation flows and recommendation logic were designed from the perspective of companies searching for programs, not the institutions publishing them.