CGIAR Centers unite to forge FAIR and AI-Ready data ecosystem for global agriculture

In an effort to build a more connected and intelligent global agricultural research ecosystem, CGIAR convened a Data Harmonization Workshop from June 17-20, 2025, at CIMMYT headquarters in Texcoco, Mexico. The primary goal of the workshop was to develop a practical framework for making data across all CGIAR Centers FAIR-Findable, Accessible, Interoperable, and Reusable. This effort is driven by increased donor pressure for transparent data sharing and the need to prepare vast datasets for the demands of Artificial Intelligence, which can uncover new insights in agricultural science. By harmonizing data, CGIAR aims to amplify its research impact, create efficiencies, and enhance partnership opportunities.

The hybrid event brought together participants from 12 CGIAR Centers was a DTA initiative, specifically under its Area of Work 1 (AoW1), which focuses on building a robust data ecosystem for CGIAR.

Building consensus across centers

Unlike past efforts that struggled to gain traction, this four-day working session emphasized co-creation and consensus, a narrative shift from rigid "standardization" to a more collaborative "harmonization". The focus is on common guidelines, not imposed standards, emphasizing a participatory process that builds on prior efforts and respects the diverse data practices across Centers. The process involved a mix of in-person collaboration-with representatives from Alliance of Bioversity & CIAT, CIMMYT, IFPRI, and IITA-and virtual contributions from AfricaRice, CIFOR-ICRAF, CIP, ICARDA, ICRISAT, ILRI, IRRI, IWMI, and WorldFish.

Participants agreed that the first priority would be harmonizing data publishing rather than data collection. Two initial domains-Agronomy and Socioeconomics & Gender-were used to test-drive the process of defining "Core Variables," creating harmonization guidelines, and designing pathways for adoption. These pathways include both mandate-driven approaches (backed by management and linked to funding requirements) and motivation-driven approaches (encouraging voluntary uptake by researchers through clear incentives).

Key outcomes and outputs

The workshop produced three major outputs:

1. Draft Harmonization Guidelines (v0.1) - A practical document to guide future data publishing, including naming conventions, metadata requirements, and attribution standards.
2. Core Variable Sets - Baseline lists of variables for agronomy and socioeconomics data, designed to improve comparability and reusability.
3. Two-Pathway Adoption Model - Combining top-down mandates with bottom-up motivation to encourage compliance without overburdening researchers.

Core variables are not a ceiling but a starting point. "Centers will have the flexibility to add Extended Variables that meet specific research needs while still aligning with the baseline for interoperability.

Why it matters

Donor agencies are increasingly demanding data sharing and reusability as a condition for funding, while AI-driven research tools require clean, standardized data to function effectively. Workshop participants reflected on why previous harmonization efforts faltered-citing lack of incentives, insufficient technical readiness, and a perception among scientists that data sharing adds to their workload without offering clear benefits.

This initiative aims to change that by building a shared sense of ownership and embedding harmonization into workflows, proposal processes, and monitoring frameworks such as CGIAR's PRMS. The ultimate vision is that all public CGIAR datasets will meet a harmonization baseline, boosting impact, reducing duplication, and enabling cutting-edge AI applications.

Next steps

Immediate priorities include finalizing the v0.1 Harmonization Guidelines, completing core variables for additional domains such as crop breeding, livestock (including aquatics), and environment & climate, and rolling out a communications and engagement strategy.

A call to action

As CGIAR steps into a new era of data-driven science, this workshop marks a critical milestone-one that could transform not just how data is handled within the system, but how agricultural research is conducted and shared worldwide.