Dynatrace has released its State of Observability 2026 report, based on a survey of 1,202 technology leaders globally, including Australia, pointing to growing use of AI agents alongside gaps in oversight and operational validation.
The report’s headline findings indicate that enterprises are increasing reliance on AI agents in observability and IT operations, but many lack insight into how those systems reach decisions. Dynatrace reported that 87% of organisations use AI agents “moderately or extensively” in observability and IT operations, while 44% aim to reach fully autonomous operations within five years, compared with 12% today. Despite that ambition, humans still verify an average of 66% of AI-powered decisions.
On decision transparency, Dynatrace said 43% of respondents consider it important to understand how AI agents reach decisions, but only 27% of organisations currently provide visibility into AI agent decision-making.
The report also links expanding AI workloads to growing complexity and cost pressures. Dynatrace said 49% of organisations now run between six and 25 AI projects, up from 42% a year ago, while 43% cited AI, large language model and agentic workload expansion as a leading source of technology complexity. It also reported that 81% saw increased compute and infrastructure spending from AI workloads, and 77% reported a greater need for cost-allocation visibility.
The email sharing the report pointed to recent commentary from Deloitte on “frontier AI incidents” in which autonomous systems bypassed intended controls, arguing this underscores the risks of deploying AI without adequate oversight.
Australia-specific data in the Dynatrace report highlighted relatively higher reported complexity linked to AI adoption. According to the summary provided, 55% of Australian respondents said AI has increased the complexity of their technology environment, compared with 47% in the Americas and 46% in EMEA. Multi-cloud and hybrid management was identified as the top source of complexity at 48%.
On outcomes and perceived value, the Australian results cited faster software delivery and faster issue resolution as leading outcomes (58% each), while 41% cited automated root cause analysis as a core value. The summary also noted challenges with data, with 39% pointing to log volume and 37% to correlating data across conventional and AI workloads, while governing sensitive data was cited at 19%.
The report summary also flagged remediation performance in Australia. It stated that Australia showed the lowest remediation among surveyed countries, with 32% taking one to four weeks to remediate vulnerabilities and 20% fixing issues in under 24 hours. Predictive detection was described as weak at 28%.
You can read the full report here.

