Why cloud migrations fail and what AI changes about that

In the 7th episode of the AI x DevOps podcast, host Rohit Raveendran sits down with Dr Rahul Vilas Ghodke, Senior Vice President and Business Unit Leader at CGI, who leads global technology operations across Asia Pacific. Drawing on close to two decades in infrastructure delivery and a team serving 50+ Fortune 500 clients, Rahul unpacks what AI-first operations actually look like at enterprise scale.

From AI Assistants to AI Coworkers

Explore the shift from AI-assisted operations, where output depends on who is prompting, to genuinely agentic operations that act in a standardized, structured, and repeatable way. Learn why the tooling and infrastructure underneath agents matter more than the model itself.

Classical AI Is Not Dead

Understand why generative AI is not the answer to every problem. Rahul shares a practical framework for choosing between classical machine learning and generative frameworks based on economics, criticality, and the nature of the workload.

Human On the Loop, Not Out of It

Discover CGI's three-tier operations model where agents handle L1 to L2 triage and humans validate every resolution for critical applications, and why zero-touch remains a future state rather than today's reality.

Tribal Knowledge and the Cold-Start Problem

AI agents wake up with amnesia every session. Learn how CGI mines, indexes, and governs a strong knowledge repository, standardizes upstream prompt catalogs, and gamifies contribution so retrieval stays accurate and reusable.

Why Cloud Migrations Really Fail

Move past the assumption that the cloud is automatically cheaper. Rahul explains the political and architectural realities of migration at scale, the cost escalations clients face, and how a factory-based approach plus observability on the "plumbing" between environments keeps outcomes on track.

Sovereign AI, OEM Lock-In, and the Two-Year Outlook

From building GPU fabs for compliance-constrained clients to the missing interoperability standards across OEMs, explore the hard parts of sovereign AI, and why Rahul predicts AIOps consumption moving from 10-20% toward 50%+ of front-office operations within two years.

Essential listening for platform engineers, SREs, cloud architects, and engineering leaders navigating AI adoption in complex enterprise and hybrid environments.