IT Pros Rely on AI Agents Despite Monitoring Gaps
A new industry report finds most organizations lack standardized methods to measure the accuracy of AI tools.
Updated on Oct. 22, 2026 in Artificial Intelligence
Live Poll
Do you trust AI agents to manage critical technical infrastructure without human oversight?
Sixty-two percent of IT and engineering professionals currently use AI agents to diagnose system issues, according to a 2026 report from Splunk. However, more than half of these experts lack a reliable method to measure agent accuracy.
Why it matters
Organizations are struggling to govern AI deployment, with 38% lacking defined ownership for monitoring agents and 85% prioritizing the detection of model hallucinations. This uncertainty persists despite 81% of teams reporting increased cross-departmental collaboration.
Of the 2,050 professionals surveyed, 58% report an ability to reliably detect AI hallucinations. Meanwhile, 25% of respondents never assign new code writing tasks to AI agents to mitigate performance risks.
The players
Splunk
Splunk is a global software company that specializes in data analysis, log management, and security solutions for IT professionals.
The details
Teams are currently integrating AI agents into workflows for root cause analysis, alert tuning, and runbook drafting. To maintain oversight, engineers perform manual double-checking of agent reasoning to verify results before implementation.
Timeline
Splunk conducted and published the observability survey in 2026.
The Tech Race
This data marks a pivot from the race to adopt AI toward the struggle to govern it. As companies move beyond simple automation, they face the replacement of legacy monitoring stacks with new AI-specific observability frameworks.
IT professionals may face increased scrutiny as organizations attempt to curb high telemetry costs and improve agent oversight. Engineers can expect stricter manual validation requirements for AI-generated code until automated monitoring tools become more accessible.
The takeaway
Organizations should establish clear departmental ownership for AI monitoring to prevent security gaps. Implementing a rigorous manual double-check process for agent reasoning is a necessary safeguard while automated hallucination detection tools remain inconsistent.
Further reading
Learn more about the latest innovations and challenges in Artificial Intelligence.
Live Poll
Do you trust AI agents to manage critical technical infrastructure without human oversight?
