AI debt
dette d'IA
Verbatim definition: the accumulated obligation to remediate AI-generated artefacts that were produced faster than they were validated. Like technical debt, it has a principal, an interest rate, debtors, and those who carry the burden, who are rarely the same people.
Three mechanisms are named: verification burden, rework externalities, and asymmetric adoption. The central argument is that responsible adoption requires measuring impact at workflow and system level rather than at the level of isolated user productivity.
What this source does not establish
A preprint, qualitative, with no quantified findings. It establishes the frame; the measurements do not exist yet.
What it means for an SME running AI
Speed gained at the point of production is borrowed, not earned. The repayment shows up downstream, in a different team, in a different month.
Sources
- John Cheung, « AI Debt: Verification Burdens, Rework Externalities, and the Hidden Costs of Asymmetric Generative-AI Adoption », Cambridge Open Engage, 23 mai 2026
This is what the survey measures See how it is measured →
Or answer six questions on your own uses Take the check →