APG shares how we moved from scattered highlights to production-grade chatbots and large language models. You’ll see why we centralized on an enterprise chatbot, a candid lessons learned from our RAG use cases, our reference architecture, how we trained our staff, and how we keep things compliant with the EU AI Act without slowing down delivery. We share concrete lessons learned, some principle we adhear to, and our roadmap for 2026.

Target Audience: Data/AI engineers, solution/enterprise architects, platform owners, product owners, BI leaders, analytics managers, CDAOs, CISOs, governance and compliance leads, IT operations leads, transformation managers
Prerequisites: Basic familiarity with large language models
Level: Basic

Extended Abstract:
Austrian Power Grid operates Austria’s transmission system in cooperation with national and European partners, ensuring safe, secure operations while enabling the transition to renewable, affordable energy.

In this session Pascal Plank explains how APG turned GenAI from isolated experiments into dependable capability that supports back-office and process work such as request handling and communications.

We walk through our initial approaches, our lessons learned, and our chosen path forward to reduce fragmentation, improve quality,  simplify governance, and compliance with the European AI Act.

This talk also describes how we enabled people to effectively use chatbot-systems and how we track adoption and added value. Pascal closes with current outcomes, pitfalls we’d avoid next time, and a pragmatic 2026 roadmap for APG's journey with Large Language Models.