AI · Crisis Management2026

Media Crisis Response AI System

An AI-based crisis response pipeline that automates media monitoring → severity classification → response message drafting. Built for an energy company's PR team as an internal operating system.

Challenge

Speed is everything in a media crisis. The bottleneck was the time from issue detection to approved response message. When issues broke in the middle of the night, a team member had to manually collect articles, assess severity, and write a draft — averaging 90 minutes per incident.

Approach

Real-time keyword monitoring via News API + RSS feeds

GPT-based severity classifier (5-level scale, automatic issue type tagging)

RAG pipeline over past response case DB for draft generation

PM2-based server management with Slack notification integration

Work

01 Crisis type selection — auto severity classification (BLACK/RED) by 7 crisis types

03 Fact verification level / 04 External spread status — separating confirmed facts from unverified to improve AI draft accuracy

Outcome

✓

Issue detection to draft generation: 90 min → under 15 min

✓

Eliminated response gaps during nights and weekends

✓

Standardized team crisis response SOP (auto-linked to system)

Insights