The Trust Gap: Why Only 5% of Enterprises Trust AI Agents Enough to Ship

Cisco President and Chief Product Officer Jeetu Patel recently shared insights at the RSA Conference 2026 regarding the challenges faced by enterprises in transitioning AI agent pilots to production. While 85% of enterprises are currently running AI agent pilots, only 5% have successfully moved these agents into production. Patel highlighted the critical factor behind this gap as trust, emphasizing that establishing trust is essential for market dominance and preventing bankruptcy.

According to a recent Cisco survey, the primary issue hindering the transition of AI agents to production is the lack of a trust architecture. Patel compared AI agents to teenagers, noting their intelligence but also their potential for immature decision-making without consequences. He stressed the importance of implementing guardrails and proper oversight to ensure the trustworthy delegation of tasks to agents.

Cisco’s response to the trust deficit at RSA Conference 2026 included announcing new security measures to protect agents and enhance detection and response capabilities. One notable announcement was the launch of Defense Claw, an open-source framework designed to enhance security for AI agents. This framework integrates various security services to provide automated security enforcement for agents running in Open Shell containers.

Patel highlighted Cisco’s commitment to advancing AI technology within the company, aiming to have 70% of its products built entirely by AI by the end of calendar year 2027. This shift towards zero-human-code engineering signifies a significant transformation in the way products are developed and underscores Cisco’s dedication to innovation and automation.

In addition to advancements in AI technology, Cisco also extended its zero trust approach to the agentic workforce through new Duo IAM and Secure Access capabilities. These enhancements aim to provide time-bound, task-specific permissions to every agent, bolstering security measures within enterprises.

Patel outlined five strategic advantages that winning enterprises can leverage to succeed in the agentic era. These advantages include sustained speed, trust and delegation, token efficiency, human judgment, and AI dexterity. Security teams can begin verifying these advantages by measuring deployment velocity, auditing delegation chains, monitoring token consumption, tracking decision points, and measuring AI tool adoption rates across security engineering teams.

As the industry continues to evolve, the importance of telemetry in security operations becomes increasingly crucial. Distinguishing between actions initiated by agents and humans is essential for maintaining security and preventing unauthorized actions. By closing the telemetry gap in logging configurations, security teams can enhance their ability to detect and respond to security incidents effectively.

In conclusion, Cisco’s innovative approach to AI technology and security underscores the company’s commitment to driving advancements in the industry. By focusing on trust, automation, and security measures, Cisco aims to empower enterprises to navigate the challenges of the agentic era successfully.