AI Eco-Design
The choices that actually matter I Hélène LEVRAULT
Webinar
AI Eco-Design
The choices that actually matter
“It’s technical.” “It’s just about generative AI.” “It’s cheap, so we barely use it, so the impact is small.” Three assumptions that quietly stall action… and all three are wrong.
In this 35-minute webinar, Hélène LEVRAULT, co-founder and CPTO at Verdikt, breaks down AI’s environmental impact into 5 concrete criteria you can apply at the framing stage of any project: need, technology choice, model and parameters, data location, and anticipated usage.
Backed by real numbers — from a 1000x gap between an LLM call and a SQL query, to a runaway $500M AI spend in a single month — this session gives you a governance framework you can put to work immediately, not another abstract sustainability pitch.
WHAT YOU'LL FIND IN THIS WEBINAR
- Why AI’s environmental impact isn’t just a generative AI story — legacy AI (machine learning, predictive maintenance, OCR) is still where most of your deployed footprint actually sits
- The 5 criteria for eco-designing AI: legitimacy of the need, technology choice, model selection and parameters, data location, and anticipated usage
- Real numbers from the field: a 1000x gap between an LLM call and a SQL query, a 30x gap between an LLM and a small language model, and how Covéa, Suez and others turned these into concrete arbitrations
- What happens without guardrails — from a $500M runaway AI spend to internal cultures that reward consumption over sobriety
- A reusable governance loop — measure, model, frame, decommission — to keep AI accountable over time, not just at launch
Watch the replay
Verdikt webinar
Request the resources
using the form below
Stay tuned!
Subscribe to our newsletter
newsletter
Meet Verdikt at Produrable 2026
2 days, 2 chances to talk about Sustainable IT with Verdikt!
EVENT