POCInternal ToolsAgentic AI
KATS — KT AI Enhanced Ticket Support System
AI troubleshooting built on Kepner-Tregoe problem analysis
Personal project
Proof of concept. The built-in browser agent is a mock; an optional local RAG backend answers from the real ticket base, and nothing leaves the machine.
A support-ticket workflow built on a 10-step troubleshooting action plan rooted in Kepner-Tregoe problem analysis, with an AI agent that finds the shortest path to resolution from the knowledge base, customer history and problem register.
Problem & business context
Ticket tools record what happened to a ticket, but rarely what the problem was — so the same issue is solved again and again, repeat customer issues go unnoticed and engineers duplicate work on the same outage.
My role & contribution
Designed and built the POC.
What I built or designed
- Universal Troubleshooting Action Plan: a 10-step method based on Kepner-Tregoe problem analysis
- AI agent that reads the knowledge base, customer history and problem register — including telling you when not to troubleshoot at all
- Runs offline in the browser from a single file
- Optional RAG backend: PostgreSQL/pgvector, a retrieval API and a local LLM