From Job Postings to Work Tasks: Building a Task-Level AI Exposure Data Pipeline with n8n and LLM

GenAI is changing work, but the key question is not only which occupations may be replaced. It is also which tasks AI can assist with, and which still depend on physical work, human interaction, professional judgment, or legal responsibility.

This session presents a reproducible task-level workflow built with n8n, Supabase, and an open-source LLM. Using job postings from five categories on TaiwanJobs(台灣就業通), the project extracts evidence-based work tasks and classifies them into four AI Exposure levels, from E0 to E3.

The pipeline also includes automatic validation, rule-based warnings, and human review. The results are summarized at the task, job, and occupational levels and presented through a public dashboard.