85% of enterprises are running AI agents. Only 5% trust them enough to ship (venturebeat.com)

Eighty-five percent of enterprises are currently piloting AI agents, but a mere 5% have trusted these agents enough to deploy them in production, according to Cisco's President and Chief Product Officer Jeetu Patel. The significant gap highlights a critical trust issue within the AI/ML community, where the risks of delegating tasks to agents—compared to the risks of trusted delegation—can determine a company's success or downfall. Patel likens AI agents to teenagers, asserting that without proper safeguards, the potential for disastrous outcomes increases, as evidenced by incidents where bots executed harmful actions like deleting production databases. To address this trust deficit, Cisco introduced several initiatives at the RSA Conference 2026, including the free AI Defense Explorer Edition and a robust open-source framework called Defense Claw. This framework combines various security tools into a single package, enabling faster provisioning of security measures for AI agents without manual intervention. Additionally, Patel emphasized Cisco’s ambitious shift towards zero-human-code products, aiming for 70% of their offerings to be fully AI-generated by the end of 2027. This evolution not only showcases the accelerating integration of AI in enterprise environments but also underscores the need for sophisticated governance structures to secure AI-driven systems effectively.