AI governance is architecture
An insurance company is building two teams. One will own actions, and the other will control them.
Guardian Life is splitting its AI platform this way, calling them the execution plane and the control plane. They refer to the latter as a “governance layer as code”.
There’s a question underneath it all that organisations need to work through: where to draw the line between what any given agent, software or other system can do versus what it’s allowed to do.
It all reminds me of similar questions around delegated authority and permissions in open banking. But in this case, the ecosystem answered them.
With enterprise AI and agents, organisations have to work it out themselves.
In this video, we explore more about what this means.
Watch on YouTube