제주특별자치도 공공정책연수원
Master leader program — AI by skill level
About the program
Train everyone at once and it stays shallow; train a few deeply and it never spreads. This program took the second route and designed the spread into it — trained master leaders return to their own departments and pass it on to colleagues. Even among public servants, AI skill varies widely. So a survey before the start measured current usage and demand to set the difficulty, and the course split into intermediate and advanced tracks with different assignments. Practice used the data and documents the provincial administration actually handles. The assignments ran from organizing and visualizing data to drafting reports and building a chatbot for recurring citizen inquiries.
What we ran
A tiered leadership program: a pre-survey set the level, then intermediate and advanced tracks ran through to finished work assignments
How the course was designed
Designed around graduates returning to their departments to pass it on, not around completion certificates
A pre-survey on current usage and training demand set the difficulty for each group
Practice built on the province's own administrative data and documents, so it transfers straight out of the classroom
What it covered
Assessment and group assignment
The pre-survey mapped awareness and skill gaps, and groups were formed by level
[Intermediate] Data analysis and workflow automation
Assignments to organize, analyze, and visualize administrative data and automate repetitive work
[Advanced] AI agents and harder assignments
Working with agents on more demanding real-world assignments
Applying it to documents and reporting
Drafting reports, producing public-facing material such as card news, and organizing and visualizing tabular data
Building a chatbot for routine work
Hands-on: a chatbot for recurring citizen inquiries and guidance requests
How it ran
Intermediate and advanced rounds run separately, across multiple sessions
An assistant instructor on site to help anyone stuck during practice
The venue network checked in advance, with a backup line ready in case practice slowed down
A satisfaction survey after each round, fed into the next
Ground rules we covered
Never entering personal data or non-public administrative records
Facts and legal citations in AI output must be verified by the staff member before use
Gallery
