Building the capacity of National Agricultural Research Systems in applying PEPA.AI for evidence-based agrifood policy analysis: A policy co-lab workshop

PEPA.AI is an innovative digital platform that combines artificial intelligence (AI) and policy analysis to generate research-based insights for informing agrifood policy decisions. Furthermore, PEPA.AI enables policymakers and practitioners to develop evidence-based, context-specific policy recommendations and create enabling environments for scaling innovations.

The workshop allowed participants to explore how to apply AI-powered political economy and policy analysis frameworks and analytical tools to generate actionable insights and provide practical solutions for trade policy, climate adaptation practices, and food system transformation.

The PEPA.AI training workshop was centered on learning about and applying the PEPA.AI tool. The training workshop was designed with two objectives:

1. Enhancing NARES, and CGIAR participants’ digital literacy in using the PEPA.AI platform and their ability to use data and evidence in policy processes related to trade policy, climate smart practices and food systems transformation facilitated by PEPA.AI
2. Providing space for collaborative engagement on critical agrifood policy challenges and generating evidence-based to inform policy processes and policy implementation.
3. Activating a Policy Action Community with a primary focus on sharing experiences on policy analysis, engagement, and implementation. 

The collaborative effort between ECDPM through the Strengthening Evidence-Based Policy Practice for Sustainable Food Systems (StEPPFoS) project, CORAF, IFPRI and the Alliance Bioversity – CIAT created the opportunity for capacity sharing with NARES to use PEPA.AI and other AI enabled tools for policy and political economy analysis. This second Science-to-Policy Co-Lab workshop took place online on 7th May 2026. The workshop aimed to move beyond problem diagnosis and into the co-design of concrete policy options for West African agricultural integration. As part of the agenda, CORAF and ECDPM invited the Alliance of Bioversity International and CIAT to present PEPA.AI (try PEPA.AI at https://pepa.alliance.cgiar.org), an AI-powered tool for political economy and policy analysis, and guide participants through a hands-on session.

What is PEPA.AI?

PEPA.AI (Political Economy and Policy Analysis Artificial Intelligence) is a digital platform developed by the Alliance of Bioversity International and CIAT under CGIAR's Policy Innovations Science Program. It combines AI with structured political economy frameworks to generate rapid, evidence-based insights for agrifood policy decisions. The tool guides users step-by-step through policy problem formulation, stakeholder and power dynamics analysis, all in a matter of clicks.

The session was facilitated by Dr. Jonathan Mockshell, Senior Scientist at the Alliance of Bioversity International and CIAT, and Leslie Estefany Mosquera, Research Associate in the PISA4 Department.

Who Was in the Room and What Did They Need?

The workshop brought together about 20 high-level policy stakeholders from institutions including CORAF, ECDPM, ICRISAT, ICRAF, Institut National des Recherches Agricoles du Bénin, Institut d'Économie Rurale, CSIR Ghana, NGOs, and others. The group spanned diverse areas of expertise: food policy, biotechnology, agroforestry, agronomy, plant breeding, and scaling.

Before the session began, facilitators asked participants to share their starting point. While not everyone responded, half of those who did say they use AI tools regularly in their work; the rest were occasional users or new to AI altogether. When facing a policy problem, the majority turn to existing literature or colleagues with expertise in policy analysis (Figure 2).

Regarding their confidence in mapping key actors and power dynamics (Figure 3) About half of the respondents described themselves as only slightly confident.

This context and background provide a framing for the demand for PEPA.AI. The West Africa region is facing a poly-crisis of conflicts, changing climate, migration challenge, and access to affordable food.  The complexity and rapid shifts require rapid response and evidence based on supporting policy decision making. For example, there is a window of opportunity to advance regional agricultural integration, with frameworks like the CAADP Kampala Declaration, AfCFTA, and the new ECOWAP and other regional initiatives.   However, scientific evidence for informing actionable policies and programs remains a gap due to data limitations, capacity constraints, and the complexity of political economy dynamics.

Participants named familiar challenges: difficulty accessing reliable data, tight time and resource constraints, the complexity of the issues they work on daily, rapid changing policy environments, geo-politics and the persistent struggle of translating scientific results into policy briefs that reach and influence decision-makers. It was precisely this last point that shaped what they hoped PEPA.AI could offer: faster, more structured analysis, and a practical way to close the gap between research and policy.

A Practical Demo on Real Policy Challenges

The hands-on portion of the session started with a live poll: participants were asked to share the policy challenges most relevant to their work. The responses covered a range of pressing issues: 

• Trade integration and food system development.
• Food security in fragile and conflict-affected contexts.
• Sustainable food systems.
• Tree tenure in agroforestry
• Agricultural insurance and product subsidies.

Using one of these challenges as a live case, Dr. Jonathan Mockshell walked participants through a practical demonstration of PEPA.AI from formulating the policy question to generating a structured, evidence-informed output. The demonstration showed how, with just a few steps.

First impressions 

Participants responded positively to the tool. Many saw direct applications in their daily work, whether preparing policy briefs, structuring research questions, or supporting institutional advocacy. The ability to generate structured analysis quickly was particularly valued by those working across multiple policy processes simultaneously.

At the same time, participants raised a concern that arises often in discussions of AI tools: the risk of bias and hallucinations (outputs that sound credible but are inaccurate). This is a legitimate and important question. The PEPA.AI team is actively working to improve the tool and reduce these risks; However, the recommendation stands for any AI-powered tool: output should always be validated by the user. PEPA.AI is designed to accelerate and structure analysis, not to replace the judgment and local knowledge of the practitioner using it.

Staying Connected: The Policy Action Community

At the close of the session, participants were invited to join the Policy Action Community (PAC; Join the Policy Action Community: https://forms.gle/uNdu8NQQejJMoJux7), a network of researchers and policymakers working to bridge the gap between science and policy. The PAC provides a space for continued knowledge exchange, shared learning, and ongoing use of PEPA.AI across institutions and national contexts. Rather than treating the workshop as a one-off event, the PAC is designed to keep the collaboration going and build lasting capacity over time.