Blog – AI for Healthy Diets: Applying AI in agrifood systems

On 16 July 2026, the Global Donor Platform for Rural Development (GDPRD) hosted its first webinar AI for Healthy Diets – Applying AI in Agrifood Systems in its AI knowledge-exchange series. The session brought together representatives from the Swiss Agency for Development and Cooperation (SDC) and the Global Alliance for Improved Nutrition (GAIN) to explore how AI is being applied in real programmes, what enabling conditions are essential, and where donors can most effectively engage.

Michelle Tang, GDPRD Secretariat, opened the webinar by stressing the need for a shared space to navigate the fast-moving AI landscape and build common understanding. She highlighted the Platform’s role in convening “short exchanges with a focus lens, to illustrate concrete case studies as examples, either for reflection or opportunities to collaborate on.”

Alessandra Roversi, Programme Officer at SDC, framed the discussion around the foundational role of data. AI cannot be effective without strong data governance, ethical safeguards, and inclusive design:

“Without good, transparent data on the context that a project is trying to improve, there’s actually no good use of AI.”

She outlined as one example SDC’s guiding principles for responsible AI use both internally and in supported programmes, including do no harm, human oversight, participation and inclusion, localization, de‑biasing, data integrity, climate considerations, decent work, and algorithmic transparency. She stressed the importance of problem-driven approaches and the need for donor coordination:

“We need first to identify the problem you want to solve and then the AI, but not the AI per se.”

Having trouble viewing the embedded video? Watch it directly on YouTube here.

Ty Beal, Senior Technical Specialist at GAIN, shared insights from GAIN’s Strategy for Harnessing AI in Programmes and its early implementation. (Download the PPT here.)

He first described the scale of the challenge: “We have billions of people who are not able to afford healthy diets… and we have food systems that are often not transforming on pace with global development goals.”

He also highlighted the persistent data gaps: “We often hit the same challenge again and again, we don't have enough data. We have disjointed data, scarce data. It's not complete.”

Ty emphasized that AI must be integrated strategically: “We’re really thinking [about] integrating into the programmatic work area—how do we really tap into AI in a strategic way?”

And he stressed the importance of equity: “It’s very easy to reach those who have access to technology… and this becomes a bigger challenge when we think about how to reach those who are often our target population.”

Two concrete use cases helped illustrate how AI can help overcome data gaps, improve programme targeting, and support efficient programme delivery:

1. Policy Chatbot in Bangladesh

A curated, vetted set of national food system policies was integrated into an AI chatbot to help policymakers quickly access accurate, up‑to‑date information. The pilot showed high usability and accuracy, demonstrating how AI can simplify access to complex policy environments.

2. Vegetables for All – Kenya

Using satellite imagery, national statistics, and machine learning (in partnership with Atlas AI), GAIN identified low‑income urban neighbourhoods where vegetable consumption could be increased. This granular mapping enabled targeted interventions reaching more than three million people.

Mduduzi Mbuya, Director of Knowledge Leadership at GAIN, offered a clear framework for donor engagement, identifying three priority entry points:

1. Strengthen foundational data systems

“When the underlying data is solid, then AI has something to stand on.”

Mduduzi noted that many critical datasets, including demographic and health surveys, are currently under strain due to declining funding. Strong data systems, more broadly, are essential for any future AI application.

2. Focus on real bottlenecks

“It’s important to not start with, can AI do this? But instead to ask, is this an actual bottleneck that someone’s stuck on?”

If the bottleneck is unclear, the tool is likely not ready for funding.

3. Invest in governance, capacity and shared standards

“This is something that's unglamorous, but this is where value for money really shows up because these foundations get reused across future projects and not just one initiative.”

Mduduzi highlighted the importance of harmonized indicators, integration with government systems (such as DHIS2), and shared learning spaces.

A recurring theme was fragmentation. Mduduzi noted: “Multiple donors and multiple development partners, sometimes in the same country, funding tools that do similar or related things independently, not talking to each other.”

Here, the Donor Platform is uniquely positioned to address this through:

• Convening donors around shared priorities
• Clarifying what responsible AI means in practice
• Creating a community of practice to share successes and failures, including “the messy lessons and not just polished use cases”

Alessandra echoed this call, inviting participants to join the emerging community of practice under the Platform’s AI and Data for Agrifood Systems workstream.

The webinar marked the start of the Platform’s AI knowledge‑exchange series. Future sessions will explore data systems, satellite‑based analytics, and other emerging tools, continuing the Platform’s commitment to responsible, equitable, and collaborative AI.

As Alessandra concluded:

“There is a great opportunity to make good use of data and AI in transforming food systems… and it's better if we're together to navigate the pitfalls.”