Is Your AI Spend Actually Paying Off | Ep # 79
What happens when you put the strategic and operational sides of FinOps in the same room
This episode of FinOps in Action pulls together voices from across the industry, including Pathik Sharma of Google Cloud, Emily Cornock, Kevin Mueller, Zach Johnson, Henrique, Shailaja Beeram, Marit Hughes, and Oliver Milke, all wrestling with the same question: how do you actually manage the cost of AI. Two years ago the idea of "AI for FinOps" got brushed off as premature. Now it sits in the top three priorities in the FinOps Foundation's own survey data, and for good reason. The guests dig into why AI spend is accelerating faster than cloud ever did, why most companies still cannot prove ROI on it, and what it actually looks like to build guardrails without slowing teams down. It is a rapid fire look at where FinOps practice is headed as AI stops being a side experiment and starts becoming the biggest line item on the P&L.Here’s what we talked about:
Treat your AI investment like a science experiment. Control your variables, change one thing at a time, and always tie spend back to a measurable business outcome, not just a feeling that things are getting more productive.
Token cost is only the tip of the iceberg. Storage, the adaptive storage layer, GPU and TPU compute, and platform hosting all add up, so don’t let your team optimize the visible 20% while ignoring the rest.
Not all AI actions carry the same risk, so treat them differently. Let AI handle low stakes, non disruptive work like cost allocation labeling on its own, but keep a human in the loop for anything disruptive, like rightsizing a Kubernetes workload during Black Friday.
Don’t let analysis paralysis become your default setting. Make your best reasonable estimate on AI spend, document your assumptions, and move forward instead of letting fear triple your forecasts.
Shift FinOps further left than you think you need to. Get engineering and product involved in AI cost conversations early, because the goal is not command and control, it’s giving teams the data to make good decisions themselves.
"I think the myth is that FinOps is something that you finish, right? You don't ever arrive at FinOps." - Amanda Wagner
Connect with Amanda and Sean:
Amanda’s LinkedIn: https://www.linkedin.com/in/amanda--wagner/
Sean’s LinkedIn: https://www.linkedin.com/in/sean~kennedy/
In this episode of FinOps in Action, host Taylor Houck brings together FinOps practitioners from across the industry to tackle one of the field's toughest new challenges: managing AI spend. The conversation digs into why token costs are only part of the picture, how teams can tell if their AI investments are actually paying off, and where AI can be trusted to act on its own versus where humans need to stay in control. Along the way, the group shares real-world examples of costs brought under control without sacrificing performance and what FinOps practitioners should be doing right now to keep up with how fast AI is changing.