Hand-off
point de passage
The moment a task, a context or a responsibility passes from one holder to the next: person to tool, tool to person, person to person, tool to tool, and the chains that combine them.
Research has shifted the focus: errors show up as failures at the information hand-off, which makes the hand-off itself the unit of analysis rather than the workflow as a whole. Six categories are commonly listed: confidence threshold, context loss, permission walls, exception sprawl, approval re-entry, and last mile.
One rule holds across the literature: errors often begin at the very first hand-off, where a human should have defined the objective, the constraints, the risk tolerance and the success condition.
What it means for an SME running AI
You do not need to examine your whole workflow. You need to examine its transitions, and there are far fewer of them than there are steps.
Sources
- Recherche 2026 sur les transferts entre agents et taxonomies d'erreur associées
Related terms
This is what the survey measures See how it is measured →
Or answer six questions on your own uses Take the check →