Science & Tech · AI · 2 hrs ago
AI agents spread faster than companies can monitor them, Splunk survey finds
A Splunk survey of 2,050 IT and engineering professionals found that AI agents are often used to diagnose system problems.
Companies commonly use them to find the cause of failures, tune alerts and draft incident guides, while many are more cautious about letting them write code.
Most respondents said AI helped them detect problems faster, but 14% said it made recovery slower.
Many companies cannot reliably catch AI hallucinations: 85% said spotting invented answers matters, but only 58% said they can do it reliably.
Responsibility is also unclear, with 38% saying their organization has not decided who should monitor AI agents.
Unclear ownership can leave security and engineering teams with gaps, even as 81% said AI monitoring has brought their teams into closer contact with security.
Monitoring can also be costly: in 57% of organizations, observability costs more than the infrastructure it monitors, and half faced an unexpected telemetry bill this year.
A Splunk survey found that AI agents are being adopted faster than many organizations can monitor them.
Among 2,050 IT and engineering respondents, 85% said visibility into AI hallucinations is important, but only 58% said their organization can reliably monitor for them.
Thirty-eight percent said their organization had not clearly assigned responsibility for monitoring AI.
Most respondents said AI improved how quickly they detect IT or security issues, while 14% said agents made issues harder to detect and resolve.
The survey also found that AI observability is adding costs, with observability costing more than the infrastructure it monitors in 57% of organizations.
- Who
- Splunk surveyed 2,050 IT and engineering professionals.
- What
- The survey found that AI agent adoption is outpacing monitoring and oversight. Respondents reported gaps in hallucination detection and unclear accountability.
- When
- The report was published on October 20, 2026.
- Where
- The survey covered respondents across ten countries.
- Why
- Not stated
This story does not have two clearly opposing sides.
Cisco published the 2026 Splunk State of Observability report.
Infosecurity Magazine reported on the survey findings.
Help Net Security reported on AI agent observability and monitoring.
- Survey respondents
- 2,050 IT and engineering professionals
- Countries covered
- 10
- Hallucination visibility considered important
- 85%
- Organizations able to reliably monitor hallucinations
- 58%
- Unassigned monitoring responsibility
- 38%
2 sources
What people are saying
When agent observability is everyone’s job, it becomes nobody’s job
Managing agents at scale requires reimagining both security and observability. Have our agents been compromised? Are they doing their job well? In the world of agentic AI, these questions are inseparable, and they're moving at machine speed
Like any new hire, they come with incredible potential as well as a period of more oversight as you integrate them in your environment
It’s like shopping for groceries where the price of a single tomato could jump from two dollars to ninety dollars overnight… and you don’t get the bill until the end of the month