Splunk’s State of Observability 2026 report has claimed AI costs are actually becoming less predictable for businesses as vendors shift toward consumption-based models.

All of this comes against a backdrop of rising agentic AI adoption, where agents can work independently and consume huge variances in tokens depending on the task, complexity and model choice.

More than half (55%) of the organisations surveyed say they can’t effectively control their AI costs, with as many (57%) now worried their telemetry costs are higher than the actual infrastructure they’re monitoring.

But while there are clear challenges surrounding cost observability, artificial intelligence has led to one clear benefit – four in five (81%) now say monitoring AI workloads has increased collaboration between observability and security teams, and 73% can accurately trace problems to their underlying security causes (up from 65% last year).

Companies are also seeing more data sharing between the two teams. “This report shows that observability and security data are converging, and that teams need to collaborate more than ever,” Cisco Chief Product Officer Jeetu Patel explained.

The report also highlights the relevance of observability teams within the broader AI strategy – an overwhelming 91% of observability leaders say they can accurately attribute AI costs to particular agents, models and workloads (compared with 69% of non-leaders), making this an important consideration for AI ROI concerns.

Still, shifts to pricing models and in how agents consume tokens continue to cause a headache for the finance department, with three in five saying that unexpected spikes in their AI bills cause more strain than a consistently high baseline cost, implying they’d much prefer to budget for flatter seat-based models over consumption models.