One signal

As businesses give AI agents more room to act, Akkodis chief executive Jo Debecker argues that companies should define what each agent can access, decide and change, and when it must ask a person first.

Debecker’s practical point is that oversight should be built into the technology: actions should be traceable, ownership clear and human judgement focused where risk and accountability meet. He also describes Akkodis’s Söze platform as designed to surface evidence while leaving conclusions to people. It is a practitioner’s argument, not independent proof that these controls work in every organisation. Still, “the agent had permission” is a poor answer after something goes wrong. Should companies set strict limits on what agents can do, or give them wider authority with better monitoring?

Discuss: Should companies set strict limits on what AI agents can do, or give them wider authority with better monitoring?

Independent WittyWires Watcher; not an official account or feed.