Everyone talks about AI saving time. Fewer people talk about the new work it’s creating.
A recent CIO article explored a growing phenomenon being called “botsitting”—the people now responsible for monitoring AI systems, validating outputs, correcting mistakes, and making sure AI stays within organizational policies.
It’s an important observation because it highlights a challenge many organizations are beginning to experience.
AI isn’t failing. It’s creating new operational overhead. When organizations first deployed generative AI, the expectation was simple: automate repetitive work and free employees to focus on higher-value activities.
Instead, many are discovering that AI often requires continuous human supervision. Someone reviews generated content before it’s published. Someone verifies AI-generated code before it’s deployed. Someone checks that an AI agent isn’t accessing sensitive information or taking actions it shouldn’t.
Every one of those reviews takes time. Individually they’re small, but collectively, they become a significant hidden cost of enterprise AI. The root cause isn’t that AI can’t be trusted. It’s that most organizations still lack operational governance for AI.
Traditional IT governance wasn’t designed for systems that make decisions, access enterprise data, and initiate actions across multiple business applications. Today, AI capabilities are being embedded into nearly every enterprise platform. Business teams can enable new AI features with just a few clicks, developers can connect agents in hours, and new AI services appear faster than governance teams can evaluate them.
The result is a familiar pattern.
Organizations know they’re accountable for AI outcomes, but they don’t always know where AI is operating, what data it’s accessing, or what actions it’s taking. People fill that visibility gap manually.
That’s botsitting.
But it doesn’t have to be. The organizations seeing the greatest return from AI aren’t removing humans from the process entirely. They’re putting humans where they’re most valuable.
Routine AI actions should happen automatically within clearly defined policies. Higher-risk activities should trigger additional approval. Sensitive actions should require a human decision. Every action should be visible, auditable, and enforce organizational rules by default.
In other words, governance shouldn’t rely on people constantly watching AI. It should be built into how AI operates. This is why we believe the next evolution of enterprise AI isn’t simply better models. It’s better control.
As AI moves beyond answering questions and begins taking actions inside enterprise systems, organizations need more than policies sitting in a binder. They need governance that operates in real time—understanding what AI is trying to do, evaluating whether it should be allowed to do it, and involving people only when appropriate.
That’s how organizations reduce the hidden cost of AI. And that’s how they scale AI with confidence instead of caution.