Your AI costs are muddy waters.
You can see your cloud bill climbing, tokens spilling - but you can't actually tell where the money's going. That chatbot? That multi-agent workflow? The experiment nobody turned off?
We can't treat AI infrastructure like VMs or containers - tag your resources, track your spend, done. All that breaks down when tokens are flying between agents, models, and workflows.
This talk covers the shift from infrastructure-level tagging to application-level token tracing. I'll show you how OpenTelemetry lets you see where every token goes, so you can finally allocate costs accurately across complex multi-agent systems.
If you're figuring out FinOps for AI (most of us are), this session gives you practical approaches that work today.