A growing trend in the tech world is the use of AI agents by organizations to streamline operations and increase productivity. According to a recent survey conducted by Docker with over 800 developers and decision makers, 95% of respondents view building agents as a strategic priority. This shift towards agent adoption has moved beyond experimental phases to early operational maturity, with 60% of organizations already having AI agents in production.
The main driving force behind the adoption of AI agents is the focus on productivity, efficiency, and operational transformation. Rather than solely aiming for revenue growth or cost reduction, organizations are leveraging AI agents to enhance internal processes and workflows across software, infrastructure, and operations. The rapid feedback loops and manageable risks associated with agent adoption make it an attractive option for many teams.
However, despite the growing popularity of AI agents, there are challenges hindering their widespread scaling. One of the major roadblocks identified by teams is AI agent security. In the survey, 40% of respondents highlighted security as the top blocker when it comes to building agents. The issue of security permeates through different layers of the infrastructure, making it a complex and critical constraint as deployments expand.
Infrastructure expansion requires secure sandboxing and runtime isolation to safeguard internal agents, while operational complexity poses security risks as more tools, integrations, and orchestration logic are introduced. The lack of coordination among multiple tools and the introduction of security and compliance risks through integrations further compound the security challenges faced by organizations. Ensuring that tools are secure, trusted, and enterprise-ready remains a significant challenge for 45% of organizations.
Many teams are turning to the Model Context Protocol (MCP) to enhance the connectivity and customization of their agents. While MCP offers a standardized approach to connecting agents with tools, data, and external systems, the security story has not caught up yet. Teams are adopting MCP without the security guarantees and operational controls they would expect from mature enterprise infrastructure, leading to concerns around security and compliance.
To address the security limitations of current MCP tooling and unlock the full potential of AI agents, organizations need to invest in new platforms built for enterprise scale. These platforms should have secure-by-default foundations, strong governance, and integrated policy enforcement to ensure reliable access controls and safe isolation of agents from sensitive systems.
For more insights and recommendations on scaling agents for enterprise use, you can download the full Agentic AI report from Docker’s website. Additionally, Docker will be hosting a webinar on March 25, 2026, to discuss key findings and strategies for prioritizing the next steps in AI agent adoption.
In conclusion, AI agent security plays a crucial role in determining the speed and success of agentic AI adoption in enterprises. While interest in AI agents is high, confidence in the readiness of current tooling for enterprise use remains a challenge. By investing in secure platforms and implementing strong governance practices, organizations can pave the way for scalable and efficient AI agent deployments.
To learn more about the State of Agentic AI report and Docker’s AI solutions, visit their website. You can also explore a new framework designed for agentic AI governance and security to address the evolving challenges in AI agent adoption.
For more Information, Refer to this article.
