From Monolith to Mesh: How to Model Data in the Age of Data Products and AI Agents

The primary goal of a data product is to make the data it exposes easily accessible and reusable by its consumers. For this reason, data modeling is not a mere implementation detail, but a core aspect of product design.

In recent years, the schema-on-read approach has contributed to pushing modeling practices into the background. However, with the rise of data product–oriented architectures and the growing adoption of AI Agents designed to consume them, the need for explicit, robust, and reuse-oriented modeling has once again become central.

Fortunately, we’re not starting from scratch. Well-established techniques such as dimensional modeling, data vault, and the unified star schema provide a solid foundation. The challenge is that most of these were originally designed for monolithic, centralized environments like traditional data warehouses or data lakes.

In this talk, we’ll examine the main data modeling techniques and discuss how to adapt them to distributed and modular architectures, with the goal of designing AI-ready data products that are easy to use, reuse, and compose.