AI demos only need one good run. But production systems have to hold up across thousands or millions of interactions. Sounds like quite a conundrum, doesn’t it? And so, Kailin Noivo sits down with Sowmya Podila, Senior Data Scientist of Generative AI at Target, to explore why so many enterprise AI use cases stall between proof of concept and production, and what separates experiments from systems that can truly scale.

Sowmya explains why successful deployments require far more than a capable model. Enterprises need evaluation frameworks, observability, security, governance, clear ownership, responsible AI practices, and infrastructure that can perform reliably at scale. She also shares why narrowly scoped domain agents tend to outperform broad generalist systems, especially when teams already have established processes, historical data, and a clear definition of what the business needs.

This conversation also looks at lower-risk opportunities, AI-assisted incident response, and the growing importance of making retail catalogs discoverable to AI agents as product discovery starts moving beyond traditional search.


Sowmya Podila is a Senior Data Scientist of Generative AI at Target, where she works on enterprise-scale agentic AI, evaluation, retail intelligence, and AI-enabled engineering systems. Her work at Target has included building evaluation frameworks for AI shopping experiences, multi-agent systems for retail planning, GenAI tools for trend discovery, natural-language data access, and AI-assisted SRE workflows. Before Target, she worked across AI, machine learning, cloud, and analytics roles at AWS, Gartner, and Tata Consultancy Services. Her focus sits at the intersection of AI engineering and business value, helping enterprises deploy AI in ways that are scalable, measurable, and practical. 

Want to uncover hidden friction and convert more revenue? Click https://go.fame.so/3KVTMDtK to book a demo of Noibu — the ecommerce monitoring and experience analytics platform.