I'm Dinesh DM. I work as a Product Manager at Mavvrik, an early-stage company building cost intelligence for AI and agent systems.
Before this, sixteen years in IT and infrastructure. Hardware repair and small-business networks. Research computing at a Government of India institution. An independent consulting practice. Along the way, I watched a few technology cycles play out from close range.
I started this blog because there is a question I keep coming back to that does not have settled answers yet:
What do AI agents actually cost, and what should they cost?
Not the API bill. The deeper economics. Why two agents doing similar work have wildly different cost profiles. Where money goes inside multi-step workflows. What is worth optimizing and what is not. How non-deterministic systems break the cost frameworks that worked for traditional infrastructure.
FinOps gave enterprises a language for cloud cost. AI agents need something adjacent but different. They retry. They reason in loops. They call other agents. The patterns do not map cleanly onto what came before.
I do not have a finished theory. I have working ideas, half-built frameworks, and a lot of questions. This is where I will think about them out loud.
What you'll find here over time:
How agent systems generate cost, layer by layer
Cost teardowns of public agent architectures
The gap between what cost dashboards show and what finance teams actually need
If any of this sounds useful, subscribe. If you disagree with something I write, ping me on linkedin. I would rather be corrected than be confidently wrong in public.
Dinesh
