Yes, I think Shadow AI is already happening in many project environments, and often much faster than formal governance structures can adapt. In most cases, people are not using AI tools with malicious intent.
They are using them because:
• Delivery pressure is high,
• Coordination overhead is increasing,
• Information volume is exploding,
•Teams are trying to reduce operational friction.
That is why I believe Shadow AI is less a technology problem and more an organizational adaptation signal.
In many organizations, workflows are already changing before policies, governance models and leadership structures fully catch up.
At the same time, excessive control can create another risk:
Teams simply stop discussing AI usage openly.
And once AI adoption becomes invisible, organizations lose:
• Visibility,
• Auditability,
• Governance,
• The ability to learn collectively from emerging practices.
So, to me, the goal should probably not be eliminating all Shadow AI.
The real challenge is creating governance models that are adaptive enough to:
• Encourage transparency,
• Define safe boundaries,
• Protect sensitive information,
• Preserve accountability,
• Still allow experimentation and operational learning.
Otherwise, organizations risk creating a paradox where governance intended to reduce risk actually pushes AI usage further underground.
I also think one of the biggest leadership shifts ahead is that project managers will increasingly need to govern AI-enabled work ecosystems, not just human workflows.
And that requires a different balance between flexibility, oversight, trust and responsible autonomy.