From Requirements to Intent: Rethinking Product Engineering for Generative AI

Abstract

Generative AI is changing product engineering by challenging what needs to be specified before a solution can be implemented. Traditional development processes progressively transform an understanding of reality into user needs, requirements, design specifications, and ultimately code. As AI becomes increasingly capable of generating implementations, detailed solution specifications become less central, while the quality of the upstream understanding becomes more important.

The first part of this trilogy examines where requirements actually begin. User needs do not originate in documents or specifications but in the context of use: in the goals people pursue, the tasks they perform, the constraints they face, and the situations in which products are used. Making this reality explicit establishes the foundation from which stakeholder requirements and, where necessary, system requirements can be derived.

The second part introduces intent as the primary engineering asset for AI-native product development. Rather than prescribing a particular solution, intent describes the problem space and the outcomes a solution is expected to achieve. Rich contextual intent gives generative AI the freedom to explore and generate different implementations while maintaining lineage to the needs and requirements that those implementations are intended to satisfy.

This shift changes the relationship between UX, Product Management, and Engineering. Upstream disciplines no longer merely provide research findings or requirements that are handed over to implementation teams. They become producers and stewards of explicit intent, while designers, engineers, and AI systems become consumers that transform this intent into solutions. The boundary between product definition and implementation therefore becomes less dependent on detailed feature specifications.

The third part proposes the Intent Lake as the infrastructure supporting this new engineering model. An Intent Lake provides a persistent foundation in which contextual understanding, needs, requirements, constraints, and other forms of intent can be maintained and consumed across the product lifecycle. Together, the three articles describe a progression from understanding reality, to making intent the primary engineering asset, to establishing the infrastructure through which humans and AI can collaboratively transform that intent into implemented solutions.

Read the full trilogy on Medium