Splunk State of Observability 2026 report: Avoiding the hidden costs of AI agents

Cisco recently released the Splunk State of Observability 2026: Restoring Predictability in the AI Era report, marking its sixth analysis of the observability industry.

The new research, which surveyed 2,050 ITOps and engineering respondents across 10 countries, exposes a significant disconnect: the urgency to deploy AI agents is rapidly outpacing the ability to monitor them.

With AI costs becoming more volatile, organisations are struggling to maintain both agent quality and fiscal predictability. This year’s report highlights how observability is helping to restore stability and financial predictability of agentic AI systems, providing the governance to build trust with agents, and ensure AI spend is tethered to business value.

In addition, with 83% of respondents saying they’re concerned about the unintended consequences of AI agents, there’s an urgent need for “agent observability,” or the ability to monitor AI agents and put guardrails in place. This need is driving ITOps and engineering teams to collaborate more closely with security teams, as agents gain access to the entire tech stack.

“As organisations increasingly integrate AI agents into their operations, managing performance, costs and potential risks is becoming more complex”, said Ahmed el Saadi, Group Vice President, MEA & TRC, Splunk. “Observability plays a critical role in restoring predictability and building confidence in the agentic AI era. By providing greater visibility into AI performance and spending, organisations can better manage costs, strengthen oversight and ensure their AI investments deliver measurable business value”.

Agent observability emerges as a business priority and a gap

While many see the value of agent observability technology, respondents still struggle with how to execute or measure the tools effectively.

• 87% confirm monitoring AI is a priority for their organisation; however 54% don’t have clear business outcomes tied to it.
• 56% reveal they’re unable to measure AI workload metrics like LLM response quality, a core metric to tracking the success of an AI agent.
• While 85% say it’s important to have visibility into AI hallucinations, only 58% claim they’re effective at doing so.

AI is driving unpredictable costs

AI costs and usage can become convoluted. In addition, AI is exacerbating a data volume problem for many enterprises, making sifting through mountains of data almost impossible.

• 55% say they can’t control AI costs, as AI use has become more complex with both consumption increasing and usage remaining hard to predict.
• 60% say unexpected billing spikes cause more operational strain than high baseline costs.

The convergence of observability and security ensures AI trust

As AI agents gain more access to operations in every layer of a business, monitoring them requires collaboration between ITOps, engineering, and security teams, breaking down walls that have historically kept them apart.

• 81% say monitoring AI workloads has increased the collaboration between observability and security teams, and 73% can accurately trace performance issues to security root causes, up from 65% last year.
• More observability teams are troubleshooting (77%) with security teams, up from 62% in 2025. Data sharing between teams has climbed from 74% to 84%.

The leader’s edge: restoring predictability in the agentic era

In the research, an advanced group of organisations emerged as leaders who are achieving concrete business outcomes with their observability practices. These leaders follow three best practices:

1. Rely on AI agents to fix and prevent issues frequently. Leaders use agents to perform and execute key remediation tasks often.
2. Closely monitor the factors that keep AI in check. Leaders focus closely on monitoring model quality, cost, usage, security, and privacy.
3. Connect signals to business impact. These advanced respondents agree that telemetry is a key factor in their planning process and helping to shape strategic decisions.

These leaders are seeing transformative results – they are far more likely to accurately attribute AI costs (91% of leaders vs. 69% of non-leaders), by pinpointing which agents, models, and workloads drive spend. They are also 2.4 times more likely than non-leaders to say the ROI of their observability platforms significantly exceeds expectations.

To learn more, explore the blogs and resources below:

Methodology

Oxford Economics researchers surveyed 2,050 ITOps and engineering professionals from practitioners to executives (including developers, SREs, systems engineers, infrastructure operations professionals, CTOs, and CIOs) from February through March 2026. Respondents resided in Australia, France, Germany, India, Japan, New Zealand, Singapore, Saudi Arabia, the United Kingdom, and the United States. Respondents represented 16 industries: business services, construction and engineering, consumer packaged goods, education, financial services, government (federal/national, state, and local), healthcare, life sciences, manufacturing, technology, media, oil/gas, retail/wholesale, telecom, transportation/logistics, and utilities.

Image Credit: Cisco