Cisco finds 55% of IT professionals say they cannot control AI costs
The sixth Splunk observability report finds AI spending hard to predict and agent oversight lagging, while a group of ‘leader’ organizations is more likely to trace spend to specific agents and models
- 85% say visibility into AI hallucinations is important; 58% say they handle it effectively
- 67% of respondents use a single observability platform
- 81% say monitoring AI workloads has increased collaboration between observability and security teams
- 91% of leaders accurately attribute AI costs, against 69% of non-leaders
- 54% of leaders say observability has yielded at least $5m (£3.7m) in annual business revenue, against 28% of non-leaders
For organizations putting agents into live workflows, monitoring becomes a budget line in its own right, because watching what agents do generates data that is billed on top of the models they run on. Without measures of output quality, an agent’s wrong action in software development or customer service can go unnoticed until it becomes the kind of business failure Cisco’s release describes. The findings come from a vendor that sells the monitoring it presents as the remedy, and the release gives shares of respondents rather than the size of the bills behind them.
“For today’s enterprises, the excitement of AI is clashing with the reality of the bill.”Kamal Hathi, senior vice-president and general manager, Splunk, Cisco
Cisco released the Splunk State of Observability 2026: Restoring Predictability in the AI Era report on 20 October, in which 55% of respondents say they cannot control their AI costs. Oxford Economics carried out the research for Cisco, which sells Splunk observability software, surveying 2,050 IT operations and engineering professionals in 10 countries between February and March 2026.
Cisco links the cost problem to AI use that is growing in volume while staying hard to predict. Unexpected billing spikes cause more operational strain than high baseline costs, according to 60% of respondents, and 57% say their telemetry costs have eclipsed the cost of the infrastructure being monitored.
Oversight trails deployment, according to the survey. While 87% call monitoring AI a priority, 54% have no clear business outcomes tied to it and 56% cannot measure AI workload metrics such as large language model (LLM) response quality. Concern about the unintended consequences of AI agents reaches 83%.
The release describes a group of ‘leader’ organizations that, Cisco says, use agents often to fix and prevent issues, closely monitor model quality, cost, usage, security and privacy, and treat telemetry as a factor in strategic planning.
Sources
- Cisco, “Splunk State of Observability 2026: Avoiding the Hidden Costs of AI Agents”, press release, 20 October 2026. https://newsroom.cisco.com/c/r/newsroom/en/us/a/y2026/m10/splunk-state-of-observability-2026-avoiding-the-hidden-costs-of-ai-agents.html
- IT Brief UK, “Cisco says UK observability costs outstrip IT budgets”, 21 October 2026. https://itbrief.co.uk/story/cisco-says-uk-observability-costs-outstrip-it-budgets
