Enterprise adoption of agent AI is at a critical juncture, with new research from Dynatrace showing companies increasing investment but scaling cautiously until reliability is assured.
Dynatrace surveyed over 900 senior leaders responsible for agent AI implementation. Pulse of Agentic AI 2026 According to the report, companies are stalling on advanced AI projects not because they doubt their value, but because they are unable to manage, validate, and securely scale these autonomous systems.
The study found that approximately 50% of agent projects are stuck at the proof-of-concept or pilot stage, but early adopters are on the rise, with 26% of organizations having 11 or more projects.
Of these, 50% already have a limited number of agent AI projects in operation, 44% have expanded across selected departments, and 23% report mature enterprise-wide integration.
Among senior leaders’ top priorities when implementing agent AI, improving decision-making with real-time insights tops the list (51%), followed by improving system performance and reliability (50%) and reducing operational costs (50%).
The main barriers to creating agent AI are security, privacy, and compliance concerns (52%), followed by technical challenges of managing and monitoring agents at scale (51%).
These concerns mean that human guidance remains a core part of companies’ agent AI strategies, even as they build toward greater autonomy.
The report shows that leaders expect 50/50 collaboration between humans and AI for IT and day-to-day customer support applications and 60/40 collaboration between humans and AI for business applications, with 87% of organizations building or deploying agents that require human supervision.
More than half of companies (64%) said they deploy a mix of autonomous and human-supervised agents, relying on validation methods such as data quality checks (50%), human review of agent output (47%), and monitoring for drift and anomalies (41%).
Meanwhile, only 13% of organizations use fully autonomous agents, and 44% report using manual methods to review communication flows between AI agents, highlighting the need for managed monitoring mechanisms.
“While human oversight remains essential today, organizations are increasingly preparing for more autonomous, AI-driven decision-making. The focus now is on building the trust and operational reliability needed to scale agent AI responsibly,” said Alois Reitbauer, chief technology strategist at Dynatrace.
Recommended reading
According to the report, this shift means observability has become a critical intelligence layer that provides visibility at every stage of the agent AI lifecycle.
Nearly 70% of enterprises report that they already use observability to gain real-time visibility into agent behavior, highest during deployment (69%), followed by operationalization (57%), and development (54%).
“Observability is a key element for a successful agentic AI strategy,” said Reitbauer. “Observability not only helps teams understand performance and results, but also provides the transparency and confidence needed to responsibly scale agent AI with proper oversight.”
