Unlocking AI’s Potential for Agricultural Innovation and DPI-Enabled Economic Growth

Abstract

The session examined how Digital Public Infrastructure (DPI)—the foundational, reusable digital building blocks of identity, payments, and data sharing—can unlock Artificial Intelligence (AI)‑driven advisory services for small‑holder farmers. Through a Gates‑Foundation‑led keynote and a multi‑national panel, participants highlighted concrete Indian initiatives (AgriStack, Vistar, Krishi Samridhi), Ethiopia’s ID‑enabled advisory model, and UAE’s AI‑for‑agriculture ecosystem. Key themes included sustainable financing models, the importance of governance and trust, benchmarking of AI solutions, and inclusive design that reaches the most marginalised farmers.

Detailed Summary

• Nidhi Bhasin opened the session, introducing the theme: AI powered by Digital Public Infrastructure (DPI) can provide trusted, data‑driven advisory services to small‑holder farmers, improve risk management, and open market opportunities.
• She announced the upcoming keynote by Sanjay Jain and noted recent Indian budget allocations for AgriStack and Bharat Vistar (the national AI advisory platform).

2. Keynote – Sanjay Jain

2.1 The Global Context

• Poverty trends: Extreme‑poverty rates have fallen by two‑thirds, yet $30 bn in aid has been cut in the last year.
• Resource constraints: With tighter budgets, new technologies—DPI and AI—must enable “more with less.”

2.2 DPI as a Foundation

• Definition (World Bank): DPI = reusable digital building blocks (digital identity, payments, data‑sharing) built for the public good.
• Gates Foundation investment: Supporting open‑source DPI platforms like MOSIP that can be customized by countries.

2.3 Indian Case Studies

Initiative	What it does	Impact
Krishi Samridhi (Odisha)	Digital services for 7 m farmers (cost < $0.18 / farmer / yr)	Boosts income, especially during climate shocks
AgriStack	National registries for farmers, crops, plots	Provides the data backbone for AI advisory
Vistar platform	Synthesises multisource content into farmer‑friendly advice (built on AgriStack)	Announced in the 2026 Indian budget; resources allocated

• Key Insight: When DPI is government‑backed and open, AI solutions can be layered on top (e.g., AI‑driven advisory, credit, insurance).

2.4 Ethiopia Example

• FIDA digital ID: Registers farmers, geotags land, delivers AI‑tailored advice in local language, and connects to FIDA Pass e‑payment for market sales.
• Lesson: Effective DPI + AI can improve yields, diversify crops, and increase farmer incomes.

2.5 Sustainability & Business Models

• Common failure: Fragmented, siloed digital systems (e.g., Nigeria’s 13 separate ID schemes) raise costs and impede integration.
• Success factor: Building single, interoperable DPI (as in India) creates a shared infrastructure that reduces duplication and sustains services.

2.6 Trust, Governance, and Scale

• Trust pillars: Transparency, consistency, accountability, and technical performance.
• Governance: DPI must be institutionally owned (mandated ministries, budget lines) to survive political changes.

2.7 Closing of the Keynote

• Emphasised that AI + DPI = inclusive agriculture; invited the panel to explore concrete examples and challenges.

3. Panel Discussion

3.1 Panel Structure

• Moderator: Nidhi Bhasin
• Panelists: Sanjay Jain (Gates), Fatema Al Mulla (U.S. Presidential Courts), Jagadish Babu (EkStep), Niriksha Shetty (PxD), plus brief interjections from a moderator named “Nadeem.”

3.2 Sustainable Business & Partnership Models (Directed to Sanjay)

• Key Points from Sanjay:
  • Sustainable services arise when countries adopt a common DPI layer (single ID, payments, data exchange).
  • Public‑private partnerships (PPPs) succeed when governments provide the horizontal DPI and private actors build vertical AI solutions.
  • Failure often stems from department‑level duplication and lack of a shared data architecture.

3.3 UAE‑India Collaboration (Question to Fatema)

• Fatema’s Overview:

  • UAE launched the AI ecosystem for global agricultural development (Dec 2023).
  • Two flagship initiatives:
    1. CGIAR AI Hub – translates AI data into public goods.
    2. Institute for Agriculture & AI (within Mohammed bin Zayed University of AI) – mobilises data, compute, technical assistance, and training.
  • Goal: Deliver AI tools as public goods to partner countries lacking AI capacity.

• Follow‑up: Fatema highlighted concrete steps for governments:

  • Institutional champion within ministries.
  • Data pipelines (collect, curate, share).
  • Capacity‑building (train local AI talent).

3.4 AI‑Enabled Advisory Systems – From Pilot to Scale (Question to Fatema)

• Key Actions Identified:

  1. Government ownership of the DPI stack.
  2. Multi‑channel delivery (SMS, smartphones, extension workers) to reach low‑connectivity areas.
  3. Co‑creation with farmers – design for, not for, the farmer.
  4. Iterative benchmarking – continuously evaluate contextual relevance, safety, and trust.

• Illustrative Example:

  • AI weather forecasting (co‑funded by Gates + UAE). Delivered to 38 m Indian farmers via SMS; feedback loops highlighted mismatches with local religious calendars, underscoring the need for culturally aware advice.

3.5 Benchmarking & Evaluation (Question to Sanjay)

• Sanjay’s View:
  • Benchmarks (e.g., AgriBench) are essential for public accountability and competition among solution providers.
  • Benchmarks must be dynamic, evolving as models improve and user needs shift.
  • Joint effort needed: academia, governments, NGOs, private sector to co‑design and maintain benchmarks.

3.6 Reusable Templates & Blueprints (Question to Niriksha)

• Niriksha’s Recommendations for Governments:
  1. Institutional mandate & long‑term budgeting – ensure DPI programmes are not “one‑off pilots.”
  2. Invest in robust data and content pipelines – keep AI models fed with fresh, high‑quality data.
  3. Feedback loops from farmers & frontline workers – measure real impact, not just usage metrics.
  4. Avoid locking into a single technology – design for interoperability so newer models can replace older ones without breaking the stack.

3.7 DPI vs. AI Solutions – Clarifying the Relationship (Question to Jagadish)

• Jagadish’s Explanation (summarised from his brief interjections):
  • DPI = horizontal layer (identity, payments, data rails).
  • AI solutions = vertical applications that sit on top of DPI (e.g., advisory, insurance).
  • Both must be co‑designed; DPI provides the personalisation capabilities essential for inclusive AI.

3.8 Inclusion – Language, Literacy, Connectivity (Open Discussion)

• Consensus Points:

  • Multi‑language support is critical; example of Bili language model built in Maharashtra within five weeks.
  • Channel diversity: SMS, voice, USSD, smartphone apps, and offline “paper‑plus‑radio” approaches for low‑connectivity zones.
  • Co‑design: Solutions should be built with farmers, incorporating their constraints, cultural practices, and trust signals.

• Illustrative Stories:

  • A farmer receiving AI weather forecasts via SMS rejected the advice because it conflicted with a religious calendar—demonstrating that cultural relevance trumps technical accuracy.
  • Trust building requires consistent, accurate advice and transparent communication of uncertainties.

3.9 Closing Remarks (Moderator & Panel)

• Panelists emphasised:

  • Trust and governance are as important as the technology itself.
  • Public‑good orientation: DPI should be built first; AI solutions can then be rapidly deployed at scale.
  • Evidence‑based scaling: Benchmarking and rigorous impact evaluation must guide expansion.

• Moderator (Nidhi) summed up: AI and DPI together can act as a “system of records” that unlocks personalized, inclusive advisory services for every farmer, but only if built on strong governance, sustainable financing, and inclusive design.

Key Takeaways

• DPI is the essential horizontal foundation (identity, payments, data‑sharing) that enables personalised, scalable AI advisory services for agriculture.
• Single, interoperable DPI systems (e.g., India’s AgriStack) are far more sustainable than fragmented, siloed approaches.
• Public‑private partnerships work best when governments provide the DPI layer and private actors develop vertical AI solutions.
• Trust, transparency, and accountable governance are non‑negotiable; without them, farmer adoption collapses.
• Benchmarking (e.g., AgriBench) must be continuous, providing a shared yardstick for safety, relevance, and performance across solutions.
• Inclusion demands multi‑channel, multilingual delivery and co‑design with farmers, respecting local customs, literacy levels, and connectivity constraints.
• Evidence‑based scaling—using impact metrics, feedback loops, and rigorous evaluation—ensures that pilots become lasting, large‑scale programs.
• Case studies illustrate success:
  • India: Krishi Samridhi, AgriStack, Vistar (budget‑backed, multi‑state rollout).
  • Ethiopia: FIDA ID + AI advisory + e‑payment (localized, language‑specific).
  • UAE: AI ecosystem for global agriculture, CGIAR AI Hub, Institute for Agriculture & AI (public‑good model).
• Future agenda: Build strong DPI registries first, then layer AI on top; keep governance central, allow modular innovation, and continuously benchmark to maintain trust and effectiveness.

---

See Also:

• ai-for-the-global-south-from-governance-to-inclusion
• from-evidence-to-scale-testing-financing-and-operationalizing-technology-and-ai-for-development-and-humanitarian-action
• democratizing-ai-resources-in-india
• ai-innovators-exchange-accelerating-innovation-through-startup-and-industry-synergy
• ai-for-everyone-empowering-people-businesses-and-society
• ai-for-inclusive-and-resilient-agricultural-food-systems
• ai-for-economic-growth-and-social-good-ai-for-all-driving-economic-advancement-and-societal-well-being
• flipping-the-script-how-the-global-majority-can-recode-the-ai-economy