The AI invoice nobody planned for — and how Cisco is solving it from the inside

Cisco shares how it tackled runaway AI costs by building an internal tokenomics discipline. The core challenge: disconnected dashboards and no unified view of which teams consume tokens and why. Their solution centers on Splunk Agent Observability (enhanced by the Galileo acquisition) to monitor agent behavior, detect hallucinations, enforce guardrails, and attribute token spend to specific teams, projects, and business outcomes. Running across 2,000+ live AI agents and 24,000 active users, Cisco maps token usage to code commits and ROI metrics, enabling finance and engineering leaders to forecast and justify AI budgets rather than react to unexplained spikes.