On 22 September 2026 I ran the fifth workshop of the Next-Gen Mentoring series, organised by the NextAIRE project (Horizon Europe) with FERIT in Osijek.
Content
The starting point: a sensor reading is not yet information. The session covered how AI can turn indoor readings, outdoor context, and health guidelines into something a non-expert can act on.
A CO₂ or PM₂.₅ value is a measurement. A decision needs the room, the people in it, what is happening outside, and a next step someone can actually take. For an assistant to get from one to the other without inventing anything, the numbers come from tools, not from the model’s memory: outdoor air quality through a weather and air quality tool, plots from a chart tool, not from HTML the model made up.
I built the exercises so anyone can repeat them after the session: MCP server setups, n8n workflow examples, and a drop-in skill for Claude, Codex, or Cursor. The skill takes an environmental signal, such as local weather or an air quality reading, and explains what it means and what to do next, using a small knowledge base drawn from WHO, WELL, and GO IAQS.
To try it, ask one plain question: Should I ventilate this room this afternoon?
Publication
No separate publication; see Materials.