(Photo by Sarah Kilian on Unsplash)
Every AI-based system will goof now and then. Sometimes they'll goof a lot.
The more sensitive the work being automated by that system, then, the more human oversight the system requires.
Otherwise the system's actions are a black box to affected parties.
The takeaway lesson for CxOs and product managers:
To you, that automation boils down to predictions and records in a database. But there are always people on the receiving end. Please build accordingly.
Anniversary of the CrowdStrike incident
One year on, how are the lessons holding up?
A very human problem
Periodic reminder: do not let the machines run unattended.