Control is fragmenting, but accountability isn’t

The major problems in the world are the result of the difference between the way nature works and the way people think.

Gregory Bateson, social scientist

We talk about governance as though it were primarily a problem of ownership. Who owns the system? Who approves the change? Who is responsible when something goes wrong? These questions made sense when technologies were relatively limited and systems largely existed within the organization’s own walls.

I don’t believe that assumption really holds anymore. What we casually call "a system" is now better described as an ecosystem: cloud infrastructure, third-party vendors, APIs, software platforms, AI models, and human decisions. Each of these different components functions according to its own logic, incentives, and constraints. This changes what it means to govern.

Traditional governance assumes that understanding follows ownership. If responsibility is clearly defined, then it invariably translates to accountability. Complex systems, however, rarely perform according to organizational charts. Their behavior emerges through relationships: software x people, automation x judgment, vendors x internal teams, policies x everyday work. Individual components can function exactly as intended while the system as a whole produces unexpected outcomes.

But hasn’t governance always depended on a certain level of legibility? Effective governance requires understanding every bit of the process, including how those components interact, where dependencies accumulate, and how small changes propagate across the whole. As systems become more interconnected, that understanding is progressively harder to reach. We gain visibility into more individual pieces yet lose sight of the relationships that shape system behavior.

Generative AI has only accelerated this shift. Every model introduces another layer of dependency: training data, foundation models, orchestration frameworks, cloud providers, human oversight, and so on. Questions that once seemed straightforward (who made this decision? Who owns this process?) are now harder to answer because no single component produces the end result. It’s now the system.

Perhaps this is where governance now finds itself, in this growing distance between control and understanding. That may require rethinking one of governance's oldest assumptions: can anyone fully understand the system they think they’re governing? Maybe that is the question we need to start asking now.