Human-on-the-loop

Definition

An oversight model in which AI performs work without approval at every step while a person supervises performance and can intervene.

Put simply

The AI operates while a person watches the system rather than approving everything it does.

What boards and leaders should understand

This can support higher levels of autonomy while retaining human intervention.

But it only works if the person receives useful information, can recognise when something is going wrong and has the practical ability to stop or change the system.

Questions to consider

  • What information would the supervising person need in order to recognise that intervention is required?