Agentic FinOps and the AI Cost Explosion ft. Pathik Sharma | Ep #76

What does agentic FinOps actually look like in practice, and how should practitioners start thinking about it?

Taylor Houck sits down with Pathik Sharma, Cloud Cost Optimization Lead at Google Cloud and Cofounder of their Cloud FinOps practice, to talk about how AI is closing the gap between knowing and doing in FinOps. Pathik shares how teams can hand off low risk tasks like tagging and labeling to AI agents while keeping humans in the loop on anything production related, and offers a four bucket framework for evaluating AI ROI: cost efficiency, productivity, differentiation, and revenue. His take for practitioners: embrace the change, learn the tooling, and let AI handle the friction so you can focus on business value.

Here’s what we talked about:

• Don't silo your FinOps practice. Cost optimization doesn't exist in a vacuum. Factor in performance, security, scalability, and capacity from the start, and build tight relationships with platform, SRE, and app teams to get anything done.

• Use AI agents for the low risk wins first. Start with non-disruptive tasks like tagging and labeling before giving AI autonomy over anything that touches production. Build confidence incrementally.

• Keep humans in the loop on consequential actions. Have your AI agent create a pull request and route it to the application owner for approval rather than pushing changes directly. The app team still owns uptime.

• Don't pick your AI model on instinct. Build a golden dataset, define what good looks like, and test models against it. One retail company cut their AI costs from $340K to $17K a month by switching models after running the data.

• Start from the problem, not the solution. Identify the real friction points your FinOps and engineering teams face, then figure out where AI reduces that friction. Chasing AI for its own sake is how you burn the budget without value.

Chapters:

00:46 Meet Pathik Sharma

01:58 AI Makes FinOps Urgent

03:09 From Tinkering To Priority

06:22 Defining Agentic FinOps

06:33 Culture And The Knowing Doing Gap

08:56 Kubernetes Agent Example

10:58 Is It Still FinOps

12:37 Humans In The Loop

13:23 Safe Automation With Tagging

15:07 PR Based Remediation Workflow

18:14 Trustworthy Recommendations First

20:58 Build Trust Incrementally

23:35 Managing AI Spend Beyond Tokens

27:27 FinOps For AI Framework

28:32 Retail Case Study Huge Savings

32:27 ROI Buckets And Closing Advice

🎧 Subscribe for more FinOps leadership conversations → https://www.finopsinaction.com/

Quote of the Show:

• "Embrace the change that is happening and learn it. FinOps holds keys to the kingdom" - Pathik Sharma

Links:

• LinkedIn: https://www.linkedin.com/in/pathik-sharma/
• Website: https://pathiksharma.com/

Ways to Tune In:

• Amazon Music: https://music.amazon.com/podcasts/f25a9d18-c12f-4ee4-93f5-2aa96e509b55
• Apple Podcasts: https://podcasts.apple.com/us/podcast/finops-in-action/id1790497808
• iHeart Radio: https://iheart.com/podcast/268443483/
• Podchaser: https://www.podchaser.com/podcasts/finops-in-action-5958339
• Spotify: https://open.spotify.com/show/3IpjMc3qxDXZAjic5Zq21t
• Substack: https://finopsinaction.substack.com/
• Transistor: finopsinaction.com
• YouTube: https://www.youtube.com/@PointFive_Inc