AI Agents Heading For A Cloud Migration Moment

Prashant Jalan is the founder of Guickly and previously led applied AI teams in the mapping technology industry.

​Before Google, I worked on the order management and infrastructure systems behind Oracle’s public cloud. It was the backstage, unfiltered view of the cloud boom unfolding. Not the architecture diagrams, but how enterprises actually bought cloud, consumed it and got billed for it. What I remember most is how fast the buying outran the managing.

An entire industry grew in that gap. One company tracked cloud sprawl. Another hunted orphaned instances nobody shut down. Another parsed AWS invoices. Another caught shadow IT, the departments quietly buying their own cloud accounts. Each looked like a separate product category. They were symptoms of one condition. Companies adopted cloud faster than they built the infrastructure to manage it.

The market eventually agreed. Microsoft bought Cloudyn in 2017, VMware bought CloudHealth in 2018, Apptio bought Cloudability and IBM later bought Apptio. Those separate products consolidated into one discipline named Cloud FinOps. Then a foundation under the Linux Foundation standardized it. It took roughly a decade.​

The Same Symptoms, Ten Years Later​​

Now look at the list of AI problems enterprises are trying to solve. Shadow AI. Runaway token bills. Tokenmaxxing. Licenses nobody uses. AI ROI nobody can prove. Costs nobody can attribute to a team or user. It feels like six problems and six categories of tooling.

It is one condition again. Companies adopted AI faster than they built the infrastructure to manage it. You cannot optimize what you cannot see. You cannot control what you have not discovered. You cannot prove return if you do not know the cost. I have written before that AI is the first trillion-dollar spend without a mature measurement layer. The cloud era is the reason I am confident that layer will get built. The open question is who will build it and who will control it.​