Agentic AI is making cloud bills unpredictable for businesses. Splunk's State of Observability 2026 report found that 55% of organizations cannot effectively control their AI costs, and 60% say unexpected billing spikes cause more financial strain than a consistently high baseline.

What You Need to Know

The shift toward consumption-based pricing for AI models means costs scale with token usage, which varies wildly depending on task complexity and model choice. Splunk's Observability report shows that 57% of organizations worry their telemetry costs now exceed the infrastructure they monitor. Teams with strong observability practices are better at attributing costs to specific AI agents and models, but most firms still lack this capability.

Why AI Billing Is Becoming Less Predictable

AI vendors are moving away from flat seat-based licenses toward consumption models. Under these models, costs fluctuate with token usage, which can spike unpredictably when agents run complex tasks or use larger models. Splunk's report found that 55% of surveyed organizations cannot effectively control these costs.

Agentic AI worsens the problem. Agents can operate independently and consume tokens at highly variable rates. A single agent might use a few thousand tokens for a simple query or millions for a multi-step reasoning task. This variance makes forecasting nearly impossible for finance teams.

Observability and Security Teams Converge

AI adoption has produced one clear benefit: better collaboration between observability and security teams. Splunk found that 81% of organizations say monitoring AI workloads has increased teamwork between these groups. Additionally, 73% can now trace problems to their underlying security causes, up from 65% last year.

"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 said in a statement.

The report also highlights that observability leaders are far more capable of attributing AI costs to specific agents, models and workloads. Among these leaders, 91% can do so, compared with only 69% of non-leaders.

  • Cost control: 55% cannot effectively manage AI spending.
  • Telemetry burden: 57% worry monitoring costs exceed infrastructure costs.
  • Collaboration boost: 81% report better observability and security teamwork.
  • Attribution gap: 91% of leaders vs. 69% of non-leaders can trace AI costs to specific models.

The Financial Strain of Surprise Bills

Three in five organizations say unexpected spikes in AI bills cause more strain than a consistently high baseline cost. This suggests many would prefer flatter, seat-based pricing over consumption models, even if the total bill is higher.

The unpredictability stems from how agents consume tokens. A single agent might call multiple models, use different context windows or retry failed tasks, all of which increase costs in ways that are hard to anticipate. Splunk's report indicates that without proper observability, finance teams are left guessing.

Why This Matters

Businesses adopting agentic AI face a budgeting crisis. Finance departments cannot forecast costs when token usage varies by task and model. This unpredictability could slow AI adoption, especially for smaller firms that lack dedicated observability teams.

Observability leaders who can attribute costs to specific agents and models will have a significant advantage. They can optimize spending, justify AI investments and avoid surprise bills. For everyone else, the risk of financial strain grows as AI usage expands.

The convergence of observability and security teams offers a silver lining, but it does not solve the core problem. Without better cost attribution, organizations will continue to struggle with unpredictable AI expenses.