AI product strategy for eCommerce — Google × Product School

What the video covers

Udit Agarwal, Lead Product Manager at Google where he oversees Consumer Shopping AI and API infrastructure across Search, YouTube, and Chrome, presents a framework for building AI-native e-commerce strategy. Drawing on seven years in shopping and e-commerce PM roles — from early grocery delivery teams to Google-wide anti-money laundering platforms — he moves beyond individual AI features to address the full transaction lifecycle: selection, payment, fraud checks, and returns.

Who it’s for

PMs working in e-commerce, marketplace, or retail products who need to think about AI strategy at the business model level rather than the feature level. Also relevant for PMs in adjacent domains (fintech, logistics) where multi-stakeholder optimization applies.

Key takeaways

1. Map AI to stakeholders, not features. E-commerce involves multiple stakeholders — buyers, sellers, logistics partners, payment processors — each with a unique value proposition. Agarwal argues that effective AI strategy starts by mapping which stakeholders can be AI-enabled based on their scale, cost structure, and revenue contribution, rather than picking features to add AI to.

2. A strong product strategy needs four components. Vision, stakeholder segments, prioritized initiatives, and dedicated effort per segment. The critical step is converting a broad portfolio of AI opportunities into a prioritized list of initiatives. Without this discipline, teams scatter resources across too many experiments and produce no measurable results.

3. Infrastructure is strategy, not plumbing. Successful AI transformation requires deep investment in infrastructure — GPUs, model hosting, evaluation pipelines — before the application layer can deliver consistent results. Agarwal draws from his experience building Google’s shopping platform to illustrate how infrastructure decisions shape what AI capabilities are even possible at the product level.

4. Rethink the business model from the ground up. AI does not just optimize existing processes; it enables entirely different approaches. The examples span hyper-personalized frontend experiences, autonomous backend operations, and new pricing models. Agarwal encourages PMs to think about which parts of their business have historically never been AI-enabled and to explore those as transformation opportunities.

5. PMs must be bold and transformational. Incremental improvements are not sufficient when AI can reshape every aspect of the business. Agarwal closes with a direct call: product managers need to stop treating AI as an optimization layer and start treating it as a reason to rethink how their business works.

Worth watching if…

You are a PM in e-commerce or a multi-stakeholder business and need a structured approach to prioritize where AI can create the most value across your entire value chain.