70% of Enterprises Are Running AI Agents, but Only 18% Have Inventoried Them

My take: A WitnessAI report published this week has one number that frames where we are: 70% of surveyed organizations are already using or piloting autonomous AI agents. Yet only 18% have all those agents formally inventoried and approved by their security team. That is not a small oversight: it means deploying tools with access to sensitive data without knowing exactly how many exist or what they are doing.

The financial consequences are already real. 43% of executives said AI-related security incidents cost their organization $2 million or more in the past year. And 68% reported that at least some AI projects ran over budget. AI still pays off: 64% say the value outweighs the risks. But that balance does not sustain itself; it requires active management.

If your organization is deploying AI agents, the question to ask is not just how many you have launched, but how many you can name, track, and audit today.