AI for Good: From Evidence to Scaled Development Solutions

Contents

Executive Summary

This panel discussion at the AI India Summit examined how AI can be operationalized at scale for humanitarian and development outcomes, specifically focusing on food security, poverty reduction, and agricultural transformation. Featuring senior leaders from India's government, FAO, IFAD, WFP, and independent evaluation bodies, the session emphasized that AI success requires not just technology but integrated policy frameworks, dedicated financing mechanisms, human-centered design, and grounded evaluation practices—moving from pilot projects to sustainable, inclusive implementation.

Key Takeaways

AI for development is not about technology first—it's about systems, financing, and human design. The conversation moved decisively away from "what's the AI solution?" to "what does the ecosystem (governance, money, data, people, evaluation) need to look like?"
Scale requires two parallel tracks: strong digital public infrastructure (government responsibility) plus private sector innovation (market dynamics). Mixing them up leads to failure; separating them enables both to thrive.
Field experience reveals the real barriers are not technical but institutional and human: cost of subscriptions, literacy gaps, structural barriers to market access for women, and information overload. Technology is one lever among many.
Evaluation and evidence must evolve with AI. Evaluation bodies can use AI to sample broader, deeper data and produce customized recommendations—but only if human judgment, independence, and bias-checking remain central.
Inclusive, grounded AI requires asking uncomfortable questions at the highest levels. Sessions like this matter because senior policy and finance leaders are being held accountable for whether AI actually reaches the poorest, smallest, and most marginalized farmers—not just early adopters.

Summary & Analysis

Key Topics Covered

AI Policy & Regulation: Risk-based, phased approaches; multi-ministerial governance; labor market sensitivity
AI Infrastructure & Financing: Role of government in building digital public infrastructure (roads) vs. private sector innovation (cars); blended finance models; non-sovereign instruments
Innovation & Scaling: Portfolio management of technologies; accessibility and affordability as prerequisites for scale; knowledge democratization
Field Operations: Real-world AI applications in food security, supply chain optimization, emergency response, and resilience-building
Evaluation & Evidence: Using AI to improve evaluation rigor; sampling broader populations; reducing bias; customized recommendations
Inclusive Design: Women farmers and smallholders; digital divide; literacy, access, and supply barriers; backward compatibility (feature phones vs. smartphones)
Data Governance & Advisory Services: Curated, localized AI models vs. generic tools; avoiding misinformation; trusted agricultural advisories
Interdependencies: Connecting AI systems across organizations; interoperability; cross-functional integration

Key Points & Insights

Sequencing Matters More Than Speed: Dr. Nagiswaran emphasized phased, risk-based regulation over omnibus rules. Governments should understand regulation as a democratic obligation to prevent market concentration and protect public interests—not just move fast.
"Roads Not Cars": IFAD's Brenda Mulele Gund articulated a critical distinction: government must build digital public infrastructure (farmer registries, interoperability, data governance, digital payments, extension systems), while private sector innovates services. Public money unlocks private investment; grants enable pilots, not scale.
Speed, Precision, and Foresight Are AI's Unique Contributions: FAO's Wanso Merto noted AI enables capabilities impossible without it—new vaccine discovery, early warning systems, advanced scientific research. AI is not optional for addressing exponential challenges; it's commensurate to the magnitude of problems.
Quality Data and Human Expertise Are Non-Negotiable: Megan Naidu's "myth-busting" highlighted the persistent false belief that AI is magic. Reality: AI requires good, reliable data, human expertise, and significant investment in training. Garbage in, garbage out applies directly.
Localized, Curated AI Beats Generic Models: Field experience shows farmers should not rely on generic ChatGPT-style tools for agricultural advice. Advisory systems must use localized datasets, vetted by researchers and extension agents, to avoid biases and misinformation.
AI Amplifies Both Solutions and Biases: Independent evaluation requires careful guardrails. If poorly designed, AI scales solutions faster—but also scales inequities and biases faster. Human judgment cannot be removed; it must remain central.
Women and Smallholders Are Early Adopters, Not Laggards: Brenda Mulele Gund countered the narrative that women are late-technology adopters. They adopt readily but face structural barriers: cost, subscription fatigue, literacy, market access, and financing. Solutions require full-circle approaches—literacy + access + supply + market linkage.
Inclusivity Requires Intentional Design, Not Afterthought: FAO emphasized the need to be "intentional" in designing solutions around people who most need them. A vaccine delivery accelerator or smallholder productivity tool differs fundamentally from a generic AI system. Public-led innovation is critical.
AI Evaluation Criteria Must Include Scale and Customization: Indran Naidu (IFAD evaluation) outlined assessment criteria: relevance, efficiency, impact, transparency, sustainability—with emphasis on scale (is it reaching who it should?) and customization (are recommendations generic or tailored to context?). Results are still emerging; systematic assessment frameworks are needed.
The Cross-Cutting Governance Gap: Dr. Nagiswaran flagged that technologists alone cannot determine AI policy. Countries need cross-cutting, interministerial AI Economic Councils to address socioeconomic implications (labor, power balances, inequality). Foundational skills in education matter as much as technical training.

Notable Quotes or Statements

"Garbage in, garbage out. The kind of output and involvement you get from AI will also depend on how rich and how deep your prompts are. So, I think the more you read, the better you will be in using AI." — Dr. V. Anantaeshwarin, Chief Economic Advisor, Government of India

"Without AI today, we would lose speed, we would lose precision, and we would lose foresight—three key essentials for advanced scientific research. Without AI, we would miss some major discoveries." — Dr. Wanso Merto, Director of Innovation, FAO

"The biggest challenge for scaling is not a lack of solutions. It's a lack of opportunity for connecting people and stakeholders together to access information. Solutions must be accessible and affordable." — Dr. Wanso Merto, FAO

"AI is real, but it's not magic. It needs good, reliable data, it needs human expertise, and it needs investment in time and training." — Megan Naidu, Chief Data Officer, WFP

"Public money builds roads. Private sector brings cars. Grants have taken us to pilots, but we cannot use grants for scale. That's why we shift financing logic." — Brenda Mulele Gund, Global Lead ICT4D, IFAD

"Regulation cannot be omnibus—it has to be risk-based or proportionate. Regulators have a democratic obligation to the public to prevent market concentration and balance of power." — Dr. V. Anantaeshwarin, Government of India

"Women are early adopters of technology. The challenge isn't adoption; it's inequalities in access, literacy, and market linkages." — Brenda Mulele Gund, IFAD

"AI is an assisted technology. It cannot exclude human beings. It's a tool that makes us more productive—but the human dimension is absolutely essential." — Dr. Indran A. Naidu, Director, Independent Office of Evaluation, IFAD

"If you create an AI solution that doesn't help deliver vaccines faster or produce something for small farmers, we're not interested." — Dr. Wanso Merto, FAO

Speakers & Organizations Mentioned

Name	Title / Role	Organization
Dr. V. Anantaeshwarin	Chief Economic Advisor	Government of India
Dr. Wanso Merto	Director of Innovation	FAO (Food & Agriculture Organization, UN)
Brenda Mulele Gund	Global Lead, ICT4D	IFAD (International Fund for Agricultural Development)
Megan Naidu	Chief Data Officer	WFP (World Food Program, UN)
Dr. Indran A. Naidu	Director, Independent Office of Evaluation	IFAD
Vashnavi Pavitran (Moderator)	Global Technology Partnerships Manager	WFP
Dr. Anub Sharma	(Organizing partner, mentioned at close)	IFAD

Organizations Referenced:

UN World Food Program (WFP)
FAO (Food & Agriculture Organization)
IFAD (International Fund for Agricultural Development)
Government of India (Ministry of Electronics & IT, RBI, SEBI mentioned)
UNDP
UNICEF
World Vision
International Development institutions (World Bank referenced)

Technical Concepts & Resources

AI Tools & Models Referenced

ChatGPT — Generic LLM cited as insufficient for agricultural advisory (requires localized, curated models instead)
Farmer Chat — AI-powered tool enabling farmers to access weather-specific agricultural advice
Drone and satellite imagery — Used in emergency response, search and rescue, crop assessment
Portfolio sense-making and management — FAO's use of AI to analyze massive datasets of project reports to identify most impactful innovations

Datasets & Infrastructure

Food System Technology and Innovation Outlook (IBI initiative) — FAO's first global database on agri-food system technologies, accessible via the Science, Technology & Innovation Portal
Farmer registries — Foundational digital public infrastructure: farmer identity, location, crops, production, sales data
Localized datasets — Country-specific data used to train agricultural advisory models (not global/general datasets)
Visa Project — India government initiative for democratizing data

Conceptual Frameworks & Governance Models

Risk-based/Proportionate Regulation — Phased approach to AI governance (vs. omnibus or unregulated)
AI Economic Council — Cross-cutting, interministerial body proposed in India's National Economic Survey
Digital Public Infrastructure — Government-built foundation (registries, interoperability, data governance, digital payments, extension delivery)
Blended Catalytic Investment — Private + public sector combination to move from pilots to scale
Backward Compatibility — Designing AI services to work on feature phones, not just smartphones
Evaluation Criteria for AI Interventions — Relevance, efficiency, impact, transparency, sustainability, and scale (breadth of reach) + customization (tailored recommendations)

Methodologies & Practices

Farmer Field Schools — Community-based learning model for women farmers (mentioned in IFAD context)
Traceability Systems — Using AI to verify organic/premium produce quality and support market access for smallholders
Early Warning Systems — AI-powered predictive models for droughts, floods, and emergency response
Supply Chain Route Optimization — AI used by WFP to deliver food faster and more efficiently
Curated Agricultural Advisory Services — Vetted information from researchers and extension agents, not generic AI responses

Policy & Strategy Documents

India National Economic Survey (2024) — First dedicated chapter on AI governance; cited throughout
WFP Global Data Strategy — Published under Megan Naidu's leadership
WFP Artificial Intelligence Strategy — Launched as first-ever formal AI strategy for WFP
India's Five Centers of AI Excellence — Announced in recent government budget; dispersed geographically across the country

Implicit Assumptions & Gaps

Results are still early: Panelists acknowledged AI impact evaluation remains nascent; systematic assessment of scale/customization outcomes will take years
Political will assumed: The discussion assumes governments will prioritize inclusive, localized AI—not merely maximize efficiency or profit
Market linkages not fully detailed: While women farmers' market access was flagged as critical, concrete solutions linking AI advisory to premium market prices were mentioned but not deeply explored
Cost of curation unclear: The necessity of localized, curated AI models raises questions about sustainability—who pays for ongoing data governance and model refinement in low-income countries?

End of Summary