AI's Decision Rights Crisis

Context engineering is an IT problem. What data your AI can access. How you curate the information environment. How you audit what it knows. Technical teams are solving that problem.

But there's a harder problem sitting one layer up that gets glossed over: Who has the authority to decide what your AI should know? And who is accountable when that turns out to be wrong? That's not an IT question. That's a decision rights question.

Context governance is an organizational design problem. Think about what "curating the information environment" actually requires someone to decide:

1. What the AI is allowed to know

   
   

2. What it needs to be protected from knowing — sensitive relationships, privileged communications, contested policy interpretations

   
   

3. How those boundaries get reviewed as the business changes

   
   

4. Who is accountable when the boundary was drawn in the wrong place

Four distinct decision rights. In most orgs, none of them have a named owner.

I've seen this play out before just not with AI, but with any major system implementation. The technical configuration gets resourced and executed well. Then something goes wrong, and the failure point wasn't the system. It was that no one had formally decided who owned the underlying logic.

Context engineering at scale is the same problem with higher stakes.

Context engineering without context governance is just sophisticated configuration.

It moves the risk downstream. It doesn't resolve it.

That's organizational design discipline. It needs organizational design ownership.