85% of Enterprises Run AI Agents, But Only 5% Trust Them

At the RSA Conference 2026, a startling statistic emerged from a Cisco executive: 85% of enterprises are currently running pilot programs for AI agents, yet only 5% have moved those agents into full production. The massive gap between experimentation and deployment reveals a single, critical barrier: trust. Cisco's President Jeetu Patel identified trust as the 'key barrier to adoption' during his keynote address. While companies are eager to explore the potential of autonomous AI agents—tools that can handle tasks, make decisions, and interact with other systems—they remain deeply hesitant to let them operate without human oversight. The fear is not unfounded. AI agents can hallucinate, misinterpret instructions, or take actions that have unintended consequences, especially in complex enterprise environments. The 5% production rate suggests that current validation, security, and governance frameworks are insufficient. Enterprises need more than just a functional model; they need guarantees of reliability, explainability, and safety. They need to know that an AI agent handling customer data or financial transactions will not go rogue. Patel argued that the industry must shift its focus from pure capability to trustworthiness. This means investing in robust testing protocols, continuous monitoring, and transparent logging of agent decisions. It also means building 'guardrails' that prevent agents from overstepping their bounds. For now, the vast majority of enterprises are stuck in the pilot phase, watching cautiously from the sidelines. The companies that solve the trust equation first will unlock the next wave of productivity gains. The rest will remain in a state of perpetual experimentation, wondering if the risk is worth the reward.