On 2 June 2026, IDED led the AI track of the CFIT workshop in Ljubjana, bringing together public procurement practitioners from the Netherlands, Denmark, Ireland, the United Kingdom and Germany to work through what AI is, what it actually costs, and what should be required of it when bought with public money. We have now published both a full write-up of the session and a segmented series of short educational video clips.

The article, Local First: What Public Buyers Need to Understand About AI Before They Procure It, covers the full range of topics discussed in the session — from how AI models are made and where the energy goes, to a concrete procurement recommendation: local-first systems that solve tasks on hardware organisations already own, escalating to cloud only when genuinely needed. It is written for procurement practitioners and policymakers who want to understand the technology well enough to specify it clearly.

The session has also been cut into thirteen standalone educational clips, each covering one theme from the workshop:

• Do We Need It — The Hidden Cost of AI (0:50)

• The Digital Economy — Where AI Fits In (3:01)

• How an AI Model Is Made (5:17)

• AI Is Not Intelligent — and Why It Is Still Scary (2:13)

• Training vs Inference — Scaling Laws, Energy and Prompts (5:18)

• Model Types — LLM, Multimodal and Generative (1:12)

• Agentic AI — What Tool Calling Changes (2:51)

• Knowledge Grounding and Shared Memory (2:27)

• How It Works End to End — and Tokens as a Commodity (1:40)

• Local First — and the Scale of Frontier Models (3:17)

• Measuring Environmental Impact Per Token (2:00)

• Q&A — Old Hardware and the Need for Speed (2:58)

• Q&A — Is Agentic AI a Security Threat? (1:24)