Agentic AI Adoption Grows in Organizations Amid Security Concerns

Agentic AI is becoming more prevalent in large organizations, as highlighted by a recent Dynatrace research report. The report indicates that 70% of respondents utilize AI agents in IT operations and system monitoring, with nearly half applying agentic AI across both internal and external use cases. Many organizations are currently investing between $2 million and $5 million annually in these technologies, demonstrating a commitment to integrating AI into core operational functions.

Despite this momentum, the adoption of agentic AI remains uneven. Half of the surveyed organizations report that their agentic AI projects are running in production for limited use cases, while 44% have projects broadly adopted within specific departments. Most teams manage between two and ten active agentic AI projects, primarily in IT operations, cybersecurity, and data processing.

Technical performance is crucial for advancing these projects, with security and data privacy being the top concerns. Organizations prioritize the accuracy and reliability of AI outputs, and many teams consider observability essential for a successful rollout. Nearly 70% of respondents already employ observability tools during agentic AI implementation, focusing on monitoring training data quality and validating outputs.

However, significant technical barriers persist. Many organizations cite security, privacy, or compliance issues as obstacles, along with challenges in managing and monitoring agents at scale. Limited visibility into agent behavior complicates the ability to trace the effects of autonomous actions, particularly as systems become more interconnected.

Human oversight remains a critical component of agentic AI decision-making. Over two-thirds of decisions made by agentic AI are verified by humans, with data quality checks and monitoring for drift being common validation methods. Most organizations develop a mix of autonomous and human-supervised agents, depending on the task and associated risks.

When evaluating the success of agentic AI, reliability and resilience are key metrics. About 60% of respondents identify technical performance as their primary success indicator, alongside operational efficiency and customer satisfaction. Early detection of errors and rapid response are vital for maintaining performance under stress.

The next phase of agentic AI adoption focuses on governance and control. Teams emphasize the need for shared factual signals and standardized metrics to guide autonomous actions. Observability is seen as a foundational control layer that connects these elements throughout the AI lifecycle.

As organizations expand their use of agentic AI, they must balance the benefits of automation with the need for robust monitoring and oversight to ensure systems behave reliably in real-world conditions.