India AI Impact Summit 2026

farmer-productivity-and-employment-transforming-agriculture-for-a-sustainable-future

6 min read

TRACK:Economic Growth and Social Good
SPEAKERS
  • Dr. Girum Ketema, Digital Agriculture and Finance,Agriculture Transformation Institute, Addis Ababa
  • Kirti Pandey, COSS
  • Mr. Jagadish B, EkStep Foundation
  • Mr. Kunjbihari Daga, Micro Save Consultancy
  • Mr. Parimal Singh, Nanasaheb Deshmukh Krishi Sanjeevani Project
  • Mr. Ravi R Singh, Digital Agriculture and Farmers Welfare, Govt of India
SESSION OVERVIEW

This session brings together policy leaders, technologists, and development partners to explore how AI and Digital Public Infrastructure (DPI) can strengthen agriculture across India and the Global South. It will showcase real-world implementations from Maharashtra, ATI Ethiopia, and other initiatives, highlighting improved farmer access to advisories, inputs, finance, and markets. The discussion will demonstrate how open, collaborative digital approaches can accelerate agricultural transformation, enhance food security, and build climate resilience.

VIDEO RECORDING

Farmer Productivity and Employment: Transforming Agriculture for a Sustainable Future

Detailed Summary

The moderator introduced the theme: how AI and Digital Public Infrastructure (DPI) can lift farmer productivity, create decent employment and increase climate‑resilience across India and the Global South. He noted that 3.6 million dairy producers (mostly marginal women farmers) are already linked to a national milk‑collection system, but half lack smartphones, limiting their ability to consume the digital services that have been built.

2. Amul AI – a Voice‑Enabled Advisory for Dairy Farmers

Speaker: Mr. Ravi R Singh (Government of India – Amul liaison)

  • System architecture – Amul AI sits on top of the existing Amul milk‑collection database (≈10 million cattle profiles). It ingests transactional data (milking volumes, vaccination, AI‑insemination, veterinary treatments) plus a library of documents, videos and audio that have been digitised.
  • “Sarla” assistant – A 24 × 7 IVR (Interactive Voice Response) service in Gujarati. Farmers call a toll‑free number, say their name, and the system pulls the relevant animal’s record and returns personalised guidance (e.g., stop milking a cow bitten by a dog, postpone AI‑insemination, adjust feed to raise fat percentage). The service also supports a chat‑bot in the same language for feature‑phone users.
  • Impact anecdote – A farmer whose cow had a dog bite was advised not to give milk to children for 10‑12 days, but to divert it to a market (the “carp” channel). The guidance prevented health risk and preserved market value.
  • Privacy & data‑ownership – Only minimal identifiers are exposed to the AI; the core database remains under Amul’s control. Data is anonymised before any external analytics are performed.

3. Ethiopia’s Open AgriNet – From Multiple Apps to a Unified Platform

Speaker: Dr. Girum Ketema

  • Problem of siloed apps – In Ethiopia the Ministry of Agriculture runs one app for market prices and another for advisory. Users must install, learn, and switch between them.
  • Open AgriNet solution – A single, open‑source portal that aggregates the two (and future) services. It keeps each department’s data sovereignty (datasets stay within the owning agency) but adds a layer‑above that can query all datasets to produce holistic advice.
  • AI layer – AI models (e.g., for disease detection, optimal breeding) run on the unified data, delivering context‑aware recommendations without forcing agencies to relinquish control of their raw data.

4. The Cost of Digital Siloes & The Case for Integration

Speaker: Kirti Pandey (COSS)

  • Illustration – Two separate apps: one pushing market prices, the other giving agronomic advice. Both require separate development, maintenance, and user onboarding.
  • Economic inefficiency – Duplicate investments in infrastructure and user training.
  • Integrated vision – Build a “single digital front‑door” where the farmer authenticates once and can access all services (market, weather, finance, extension) via a common API layer. The approach respects departmental data ownership while delivering a unified user experience.

5. Weather‑Centric Advisory – From Raw Forecasts to Actionable Insights

Speaker: Mr. Parimal Singh (Krishi Sanjeevani Project)

  • Evidence‑based design – Randomised impact evaluations in Odisha showed that weather‑shocks are the biggest driver of crop loss; any advisory that mitigates this yields the highest returns.
  • Message design – Translating a “70 % probability of rain” into a simple, actionable message (e.g., “Delay sowing by 5 days”) proved essential; the raw probability confused many farmers.
  • Scale – The platform has already reached 38 million Indian farmers and aims for ≥ 7 million Ethiopian smallholders.

6. Digital Public Infrastructure (DPI) Principles in System Design

Speaker: Mr. Kunjbihari Daga (Micro Save Consultancy)

  • Four DPI pillars – (1) Interoperability, (2) Data minimisation, (3) User‑centric design, (4) Governance & accountability.
  • Implementation – “Bihar Krishi Sanjeevani” kept legacy data sources (state‑level farmer registries, scheme portals) intact, built state‑highways of data that route requests to the appropriate source, and offered voice‑command entry in local dialects.
  • Impact on financial inclusion – By consolidating a farmer’s digital identity and transaction history, banks can assess creditworthiness more accurately, unlocking rural liquidity.

7. Employment Effects of AI in Agriculture

Speaker: Mr. Jagadish B (EkStep Foundation)

  • Common fear – AI will replace agricultural labour.
  • Counter‑argument – In dairy farming, labour is intensive (milking, animal health checks). AI augments workers by delivering precision advice, reducing disease, and improving milk quality, increasing the need for skilled operators (e.g., AI‑assisted veterinary technicians).
  • Projected growth – Amul aims to serve 1 lakh cattle‑owners this year and scale to 2 lakh in the next 3‑4 years, implying a net rise in demand for AI‑savvy extension workers.

8. Data Sovereignty, Ethical AI & Trust

  • Privacy model – Data stays with the originating agency; AI accesses it through policy‑driven APIs that return only the minimal inference needed.
  • Ethical guardrails – All models are audited for bias, especially in language‑translation and gender‑sensitive advice, to avoid inadvertent discrimination.

9. Language Diversity & Localization

  • Challenge – India alone has ≈ 22 official languages and hundreds of dialects; Ethiopia adds multiple Afro‑asiatic languages.
  • Solution path – Build local glossaries and train LLMs on region‑specific corpora; continuously monitor “fallback” failures and feed corrections back into the model.
  • Current status – The voice‑assistant recognises 10‑15 Gujarati accents; ongoing work to cover Marathi, Odia, Amharic, Afaan Oromo etc.

10. Audience Q & A – Rapid‑Fire Highlights

Question (topic)Main responder(s)Key points answered
Krishi Call Centre vs AIRavi R Singh & Kirti PandeyAI can triage at scale, but human call‑centres retain empathy for complex cases; hybrid model recommended.
Financial inclusion through AIKunjbihari DagaUnified digital identity improves bankability, enabling direct‑benefit transfers and easier loan access.
Difference between Open AgriNet and existing push‑servicesDr. Girum KetemaOpen AgriNet is open‑source and interoperable, while older push services are proprietary, silo‑bound.
Scale of dialect coverageJagadish BSystem currently supports ≈ 12 major dialects; expansion driven by crowdsourced recordings.
Co‑creation with farmersParimal Singh & Kirti PandeyAdvisory content is co‑designed with farmer focus groups; iterative field pilots ensure relevance.

Key Takeaways

  • Unified data layer: Across India, Ethiopia and other partner countries, a single API‑driven “data highway” that respects departmental sovereignty is the cornerstone for scalable AI services.
  • Voice‑first, local‑language interfaces (e.g., Amul’s “Sarla”) dramatically increase adoption among feature‑phone farmers, especially women.
  • Privacy‑by‑design: Farmers’ personal and livestock data remain under the control of the originating institution; AI receives only anonymised, need‑to‑know inferences.
  • Weather advisory is the highest‑impact vertical; translating raw forecasts into simple, actionable messages yields measurable reductions in crop loss (≈ 25 % in Odisha).
  • Financial inclusion improves when a farmer’s digital identity aggregates transaction, input‑purchase and advisory data, enabling banks to assess credit risk more accurately.
  • AI does not replace labour; it creates new skilled roles (e.g., AI‑assisted veterinary technicians, data‑curators) and can increase labour productivity in dairy and mixed‑farm settings.
  • Language diversity remains a technical hurdle; continuous collection of regional speech samples and iterative model fine‑tuning are essential.
  • Ethical AI governance (bias audits, transparency, accountability) is treated as a non‑negotiable prerequisite for any large‑scale rollout.
  • Co‑design with farmers is vital—solutions must be built around real pain points, not as top‑down tech experiments.
  • Open‑source platforms like Open AgriNet and Digital Public Infrastructure frameworks provide a replicable blueprint for other Global‑South contexts.

See Also:

Graph View