By Thaddée Leblond · Published · Updated: July 2026
Rejecting AI in the name of sobriety would be as simplistic as adopting it without limit. It consumes resources — computing, energy, hardware — but it can also serve the transition by optimising processes and reducing waste. The right question is not "AI or not", but "which use, for what value": useful uses, deliberate sobriety, clear governance.
A two-sided subject
Rejecting AI in the name of sobriety would be as simplistic as adopting it without limit. AI has a real environmental cost (model training, infrastructure, hardware), but it can also generate gains: fewer trips, less waste, better planning. It all depends on the balance between benefit and cost.
The principle: useful use
Sobriety applied to AI means reserving it for cases where it brings real value, and avoiding systematic, excessive recourse. Not every task warrants a powerful model; often, a simpler tool suffices. Choosing the right tool for the right need is already an act of sobriety.
AI in service of the transition
In some contexts, AI directly helps reduce impacts: energy optimisation of buildings or processes, predictive maintenance that avoids replacements, planning that reduces scrap and trips. The criterion stays the same: the gain must exceed the tool's cost.
Govern to stay responsible
As with the other issues of responsible digital technology, the key is governance. An internal AI policy specifies relevant uses, protects sensitive data, requires human validation and places AI within a coherent approach — both environmental and ethical. Sobriety here meets responsibility in the broad sense.
The balanced stance
Responsible digital technology does not say "no to AI": it says "a chosen, measured and governed AI". For an SME, it is a tenable and differentiating posture — drawing on the tool without making it an ecological or ethical blind spot.
Knowledge about AI's impact and its framework evolves quickly. Rely on up-to-date data for your decisions.
Frequently asked questions
No, but it requires discernment. AI consumes resources (computing, energy, hardware), while also being able to optimise processes and reduce waste. The question is not 'AI or not', but 'which use, for what value'.
By reserving it for genuinely valuable uses, avoiding systematic and excessive recourse, favouring models suited to the need rather than the heaviest, and measuring the real benefit against the cost.
Yes, in some cases: energy optimisation, waste reduction, predictive maintenance, better planning. The point is that the gain exceeds the tool's own environmental cost.
Yes. An internal AI policy sets useful uses, protects data, requires human validation and places AI within a responsible logic — environmental as well as ethical.
Sources
- Public observatories — contribution of AI to the transition
- Responsible AI framework (governance, data, human validation)