Béranger Thomas

Immersion IA

Co-design and facilitation of AI training workshops, and production of pedagogical resources (prompting booklet, facilitator guide) for employees of the Metropolis of Lyon.

Technical sheet
Status Stable Release date 2025-06-02 Category AI training & mentoring Language Markdown · PDF · Notion Stack Notion · Canva · LaTeX · LLMs License Open Licence

Context

As part of the deployment of sovereign artificial intelligence technologies at the Metropolis of Lyon, integrating new tools based on Large Language Models (LLMs) raised significant needs for digital literacy and change management. With over 9,600 public agents managing sensitive administrative and citizen data, adopting AI required a rigorous awareness framework regarding actual capabilities, limitations, and privacy compliance (GDPR).

Challenge

The challenge was to design and facilitate a comprehensive educational path to alleviate psychological and technical barriers to AI adoption while delivering practical, actionable prompting skills. Deliverables included leading interactive “Immersion IA” workshops and producing sustainable, high-quality pedagogical materials for both end-users (metropolitan staff) and workshop facilitators.

Approach

The initiative was built around three methodological and technical pillars:

  1. Interactive Workshop Facilitation & Feedback Loop:

    • Led practical sessions at the UrbanLAB in Villeurbanne.
    • Introduced basic LLM mathematical and probabilistic foundations to demystify chatbot behavior (“An AI does not think; it predicts tokens based on context”).
    • Guided users through hand-on prompting exercises tailored to administrative use cases (document synthesis, writing assistance, search optimization).
    • Conducted structured post-workshop analysis of user feedback (collecting ideas, categorizing roadblocks, and prioritizing action items in a reporting database to guide future IT policies).
  2. Pedagogical Content Creation — “10 Keys for Effective Prompting”:

    • Authored a comprehensive prompting guidebook illustrated with realistic, professional scenarios.
    • Aligned the graphic design and layout with DatAgora’s guidelines.
    • Distributed as a physical A5 printed booklet and an interactive digital PDF hosted on a dedicated Notion workspace.
  3. Technical Companion Guide for Facilitators:

    • Authored an advanced Q&A manual to empower internal facilitators with technical knowledge.
    • Clarified generative AI concepts.
    • Provided precise technical answers to common non-technical questions (e.g., explaining why statistical models behave with politeness and courtesy).

Outcome