AI-ready Data Products

The AI Continent Action Plan, published by the European Commission in April 2025 highlights under its “Data for AI” pillar that “access to reliable and well-organised data is essential if the EU is to unlock the full potential of AI”. The European Data Union Strategy in November 2025, (with the subtitle “Unlocking data for AI”) details the need of “scaling up access to quality data for AI and innovation”.

“Data for AI” has long been a key area of focus for BDVA community, and therefore was central to the BDVA responses and feedback to these EC initiatives, published as “Towards a European AI-Data Value Ecosystem”, and “Data at the core of Europe’s Digital Strategy”.

To make these plans a reality, the BDVA community has identified an urgent need for the traditional concept of data product (focused on packaging and sharing datasets for general use) to evolve in order to meet the specialised requirements of AI, a new paradigm that in BDVA we refer to as “AI-ready Data Products”.

The present paper builds on past and ongoing discussions and activities within the BDVA community around the paradigm of “Data for AI” and how this can be embedded in the data product approach. The paper provides the path to follow to redefine the paradigm of data products in a way that AI practitioners can fully harness the power and value of data within evolving data ecosystems. In doing so, it aims to position these data ecosystems as strong catalysts for AI innovation by delivering industry-ready solutions tailored to the complexities and unique requirements of advanced AI applications.

To the best of our knowledge, this is the first attempt to systematically evolve the concept of a data product so that it accommodates the needs, constraints, and practices of AI systems and AI practitioners. The work of the community has produced the following key needs: Lifecycle: The stages of the lifecycle need to be revisited to incorporate AIready elements with a particular focus on extended metadata requirements.

Technical: The quality of data is central to its use by AI and techniques for preparing data for AI applications, and metadata models for describing data suited for AI consumption. There is a clean need to support both static data products to dynamic AI-ready Data Products, incorporating information from different stages of the data pipeline and explaining how these can be aligned with AI workflows.

Governance and Compliance: Need to be enhanced to include the ethical dimension of Data Products, existing regulations and standards, and specific licensing and contractual considerations for AI models and AIready Data Products. Readiness framework: To drive adoption a readiness framework is needed to capture the broader requirements identified for AI-ready Data Products.

The convergence of data management best practices, modern data product principles, and AI-specific requirements is highly promising to enable "AIready" Data Products. The concept requires further exploration, knowledge exchange, co-creation, and validation across disciplines, industrial players, and other relevant stakeholders. Therefore, we plan to continue this work within BDVA community (including a new version of this document in 2026 that collects all these new developments), but also to extend the dialogue to external stakeholders, other associations, standardisation bodies, policymakers, industry actors and AI communities.

The BDVA community aspires to transform the initial ideas of this paper into concrete standards, tools, and methodologies that make AI-ready Data Products a reality in various industries and ecosystems, ultimately contributing to the successful implementation of the European Data Union Strategy through truly integrated AI–data value ecosystem that link infrastructure, data, AI, talent, regulation and innovation.