California Management Review: when AI joins the product team

What the article is about

Published in the California Management Review in February 2026, this piece by Shivam Srivastava examines what happens to innovation capacity when AI becomes part of product team workflows. The central finding is not that AI enables innovation autonomously — the data suggests most organizations that adopt AI do not reach a point where the impact is measurable.

Context and companies

The article draws on a range of company examples. P&G’s internal AI hackathon showed AI-augmented teams generated higher-quality outputs roughly 15% faster, with individuals using AI performing as well as two-person teams working without it. Nestlé and IBM developed an AI system for exploring sustainable packaging options by mining scientific literature and patents. Mondelēz deployed generative AI for marketing content, with projected cost reductions of 30 to 50 percent. Barry Callebaut, the largest chocolate maker in the world, partnered with NotCo AI to embed machine learning into recipe development, with AI generating and evaluating novel formulations to accelerate R&D.

Against these examples sits the central statistic: 65% of leading organizations now use generative AI, but only 10 to 15% achieve measurable business impact, and 85% of AI initiatives never reach full production.

Key takeaway

Srivastava’s argument is that the gap between adoption and impact is a leadership and governance problem, not a technology problem. What the successful cases share is staged investment with defined go/no-go criteria, designated executive accountability across functions, embedded human review at output boundaries, and investment in what he calls decision literacy — the capacity of managers to evaluate AI outputs critically before acting on them.

Who it is useful for

Product executives, heads of innovation, and senior PMs responsible for managing AI adoption across teams or business units, particularly those looking for evidence-based frameworks rather than prescriptive playbooks.