AI agritech can substantially boost smallholder outcomes—raising yields by double‑digit percentages and cutting input costs—but gains are uneven and hinge on connectivity, local skills and supportive regulation.
A systematic review of the economic impact of artificial intelligence on agricultural productivity, sustainability, and rural livelihoods
A structured review of 60+ recent studies finds AI-powered agritech in developing contexts can raise yields by roughly 12–45%, cut input costs up to ~25%, and improve supply‑chain efficiency, but scalable and equitable adoption is constrained by infrastructure, skills, and policy gaps.
As agriculture confronts climate change, land degradation, labour scarcity, and food insecurity, the application of Artificial Intelligence (AI) has emerged as a critical strategy to enhance productivity, sustainability, and rural resilience. This study explores how AI-powered digital agriculture transforms agricultural economics in developing contexts, with a particular focus on Sub-Saharan Africa. Using a structured literature review and thematic synthesis, we analyzed over 60 peer-reviewed articles and institutional reports published between 2020 and 2025. The analysis was organized around three key outcome areas: productivity, sustainability, and rural livelihoods. AI technologies, ranging from predictive analytics and advisory systems to smart irrigation, pest/disease detection, and precision fertilization, demonstrate a consistent pattern of impact. Studies reveal AI-driven yield increases between 12% and 45%, input cost reductions of up to 25%, and measurable improvements in supply chain efficiency. Furthermore, AI contributes to labour reallocation and market integration in rural areas. However, barriers such as infrastructure gaps, digital literacy, and policy vacuum limit scalable adoption. This review bridges technological applications with economic development frameworks, offering an interdisciplinary lens on AI’s role in agriculture. It articulates pathways through which AI enhances technical efficiency, environmental resilience, and rural economic transformation. This review is novel in its structured synthesis of economic outcomes using a triadic framework. It introduces comparative effect-size analysis to quantify AI’s impact in the study. The paper recommends investments in localized AI solutions, rural digital infrastructure, and policy environments that foster inclusive technology diffusion. The study contributes to the growing field of digital agricultural economics, providing evidence-based guidance for policymakers, agritech investors, and development stakeholders.
Summary
Main Finding
AI-powered digital agriculture substantially improves farm-level productivity, environmental performance, and some measures of rural welfare, but benefits are uneven and constrained by infrastructure, capacity, and governance gaps. The reviewed evidence (2020–2025) reports typical yield gains in the mid-teens to low‑40s percent (commonly 10–30%; review summary range 12–45%), input-cost reductions up to ~25%, and notable efficiency gains in water and nutrient use under AI-enabled systems. However, most evidence comes from pilots and case studies, limiting generalizability and raising questions about scale, distributional effects, and long-run impacts.
Key Points
- Productivity
- AI tools (predictive analytics, advisory systems, precision fertilization, pest/disease detection, smart irrigation) consistently improve technical efficiency and yields. Reported increases clustered around 10–30% in many studies; the review’s comparative effect-size synthesis cites a 12–45% range across contexts.
- Input-use efficiency improves (fertilizer, water, labour) with cost savings reported up to ~25%.
- Sustainability
- AI-backed nutrient and irrigation management lowers environmental externalities (reduced nitrogen leaching, lower water use). Examples include closed-environment agriculture (CEA) cases with ~60% water savings and ~40% lower energy use while maintaining yields.
- Net sustainability gains depend on system design; AI/data infrastructure can impose additional energy/resource footprints.
- Rural livelihoods & labour markets
- AI can displace routine, low-skill tasks but also creates new roles (data analysts, drone operators, precision-tech technicians). Some pilot evidence indicates improved incomes and market integration when AI is bundled with finance or cooperative models.
- Persistent digital divides (gender, literacy, connectivity) and algorithmic bias risk excluding marginalized farmers.
- Heterogeneity & gaps
- Adoption and impacts differ between Global North (capital-intensive automation) and Global South (mobile/advisory platforms). Most studies are pilots, short-term, or localized; few large-scale, longitudinal or cost–benefit analyses across farm sizes and cropping systems exist.
- Governance & policy
- Key barriers: rural connectivity, digital literacy, tailored/localized AI models, data governance, transparency of algorithms, and supportive public–private coordination.
Data & Methods
- Study design: Mixed-methods systematic review, case synthesis, and comparative analysis focused on three outcome domains—productivity, sustainability, and livelihoods.
- Literature base: Sources published 2020–2025. From an initial 142 records the authors screened to a final set of 60 sources (43 peer‑reviewed, 17 grey literature) for coding and synthesis. Full texts were coded in NVivo.
- Search & inclusion: Academic databases (SpringerLink, ScienceDirect, IEEE Xplore, MDPI, SSRN, IGI Global) and grey repositories (FAO, IFPRI, World Bank) using targeted Boolean queries (e.g., “artificial intelligence” AND agriculture AND productivity). Inclusion required empirical/case evidence and clear metrics on the triadic outcomes.
- Quality appraisal: Articles assessed on clarity of outcome metrics, AI application description, methodological transparency, and context specificity; low-quality items were excluded if they failed two or more criteria.
- Analytical framework: A triadic framework (productivity, sustainability, livelihoods). To compare heterogeneous outcomes, effects were standardized via relative percentage-change effect sizes: Effect Size (%) = (Outcome_AI − Outcome_Baseline) / Outcome_Baseline × 100 Midpoints used where ranges reported; emphasis on percentage changes rather than parametric meta‑analysis due to reporting heterogeneity.
- Exclusions: Pure theory without impact data, studies solely on robotics/IoT without AI, non-English texts.
Implications for AI Economics
- Mechanisms for economic change
- AI raises total factor productivity primarily through better allocation of inputs and reducing informational frictions, consistent with neoclassical and induced-innovation frameworks.
- AI also drives structural change (labour substitution and new task creation) consistent with Schumpeterian “creative destruction”; this necessitates attention to skill formation and rural labor markets.
- Policy and investment priorities
- Invest in rural digital infrastructure (connectivity, sensors, scalable data services) and in human capital (digital literacy, extension services) to enable inclusive diffusion.
- Support localized AI models and datasets to reduce algorithmic bias and improve relevance to smallholders and diverse cropping systems.
- Design public policies for data governance, algorithmic transparency, and public–private R&D collaboration to balance innovation with equity and accountability.
- Consider blended interventions: AI tools coupled with finance (microcredit/insurance), cooperatives, and tailored extension to maximize welfare impacts.
- Research priorities for AI economics
- Conduct large-scale, longitudinal impact evaluations and cost–benefit analyses across farm sizes and cropping systems to assess scalability and distributional effects.
- Quantify spillovers (market integration, supply-chain efficiency), net environmental lifecycle impacts (including AI infrastructure energy use), and labor-market transitions.
- Evaluate policy instruments (subsidies, data trusts, training programs) to identify approaches that democratize AI benefits and mitigate risks to vulnerable groups.
- Implications for stakeholders
- Policymakers: prioritize enabling infrastructure, inclusive governance, and targeted programs for smallholders.
- Agritech investors: focus on localized solutions, affordability, and partnerships that lower adoption barriers.
- Development agencies: fund capacity building, rigorous impact evaluation, and models that integrate AI with finance and extension.
Short takeaway: AI shows meaningful promise to boost agricultural productivity and sustainability, but the economic gains will be realized only if investments and policies address connectivity, capacity, localization, and governance—otherwise benefits risk being concentrated and short‑lived.
Assessment
Paper Typereview_meta
Evidence Strengthmedium — The paper is a structured review synthesizing 60+ studies that include some causal designs (RCTs and quasi-experimental studies) alongside observational case studies and pilots; reported effect-size ranges are informative but heterogeneous, often short-term, and subject to contextual and publication biases, limiting causal certainty at scale.
Methods Rigormedium — Uses a structured literature review and comparative effect-size synthesis across recent (2020–2025) peer‑reviewed articles and reports, but does not appear to be a formal, pre-registered meta-analysis with uniform inclusion/exclusion criteria and heterogeneity/quality weighting, and the underlying studies vary in methodological quality.
SampleEvidence base of over 60 peer‑reviewed articles and institutional reports from 2020–2025, focused primarily on Sub‑Saharan Africa with some comparative developing‑country cases; underlying studies include RCTs, quasi‑experimental evaluations, observational case studies, and pilot implementations of AI applications (predictive analytics, digital advisory systems, smart irrigation, pest/disease detection, precision fertilization).
Themesproductivity adoption governance innovation labor_markets inequality skills_training
GeneralizabilityPrimarily focused on Sub‑Saharan Africa and similar developing‑country contexts; findings may not generalize to high‑income agricultural systems., Heterogeneous effects by crop type, farm size, value‑chain position and institutional context limit broad extrapolation., Many underlying studies are pilots or short‑run evaluations, restricting inference about long‑term and system‑level impacts., Results depend heavily on local infrastructure (connectivity, power) and digital capacity, constraining transferability to poorly connected areas., Potential publication and selection biases (projects led by NGOs/companies, limited null results) may inflate reported effect ranges., Technology- and implementation-specific findings (different AI tools produce different outcomes), reducing cross-application generalizability.
Claims (21)
| Claim | Direction | Confidence | Outcome | Details |
|---|---|---|---|---|
| AI-powered digital agriculture in developing contexts—especially Sub-Saharan Africa—can materially improve productivity, sustainability, and rural livelihoods. Firm Productivity | positive | medium | aggregate outcomes: productivity, sustainability, rural livelihoods |
n=60
0.14
|
| Across reviewed studies (2020–2025), AI interventions are associated with yield gains of roughly 12–45%. Firm Productivity | positive | medium | crop yield (% change) |
n=60
12-45%
0.14
|
| AI interventions are associated with input cost reductions up to ~25%. Firm Productivity | positive | medium | input costs (% reduction) |
n=60
up to ~25%
0.14
|
| AI deployment has produced measurable supply-chain efficiency improvements and better market integration in reviewed cases. Organizational Efficiency | positive | medium | supply-chain efficiency and market integration (e.g., logistics time, transaction costs, market access) |
n=60
0.14
|
| Evidence of labour reallocation within rural economies following AI-driven productivity changes was observed in the reviewed literature. Employment | mixed | medium | labour allocation / employment composition in rural economies |
n=60
0.14
|
| Technologies assessed in the review include predictive analytics, digital advisory systems, smart irrigation, pest/disease detection, and precision fertilization. Other | null_result | high | types of AI/digital agriculture technologies studied |
n=60
0.24
|
| AI-enabled interventions produced technical efficiency gains through better input targeting and reduced waste. Firm Productivity | positive | medium | technical efficiency (input targeting accuracy, quantity of inputs used, waste reduction) |
n=60
0.14
|
| AI applications contributed to environmental resilience via water and fertiliser savings and earlier pest detection in some studies. Consumer Welfare | positive | medium | water use, fertiliser use, pest detection timeliness |
n=60
0.14
|
| AI interventions supported economic transformation in some contexts by improving market access and enabling reallocation toward higher-value tasks. Firm Revenue | positive | medium | market access indicators, income sources, task composition |
n=60
0.14
|
| Scalable adoption of AI in developing-country agriculture is constrained by infrastructure gaps (connectivity, power, data platforms). Adoption Rate | negative | medium | adoption rates / scalability mediated by connectivity, power, platform availability |
n=60
0.14
|
| Limited digital literacy and human capacity among smallholders is a key barrier to adoption and effective use of AI solutions. Adoption Rate | negative | medium | adoption and effective use of AI tools; digital literacy metrics |
n=60
0.14
|
| Policy and regulatory vacuum (data governance, interoperability, safeguards) limits scale and inclusive diffusion of AI in agriculture. Governance And Regulation | negative | medium | policy/regulatory environment effects on adoption and inclusivity |
n=60
0.14
|
| Effect sizes and impacts vary substantially across contexts—by crop, farm size, and institutional setting. Other | null_result | high | heterogeneity of effect sizes by crop type, farm size, institutional context |
n=60
0.24
|
| The evidence base reviewed comprises more than 60 peer-reviewed articles and institutional reports from 2020–2025, primarily focusing on Sub-Saharan Africa. Other | null_result | high | number and regional focus of studies in the review |
n=60
0.24
|
| The review used a structured literature review with thematic synthesis and a comparative effect-size analysis to quantify ranges for yield, cost, and efficiency outcomes. Other | null_result | high | review methodology and analytical approach |
n=60
0.24
|
| There is variation in study design and quality in the evidence base (RCTs, quasi-experimental studies, observational case studies, pilots). Other | null_result | high | study design types and quality variation |
n=60
0.24
|
| There is limited long-term impact evidence and few system-level assessments of AI in developing-country agriculture. Research Productivity | negative | high | presence/absence of long-term impact evaluations and system-level assessments |
n=60
0.24
|
| Policy implication: prioritize funding for localized AI solutions (context-specific models, language/extension support) and rural digital infrastructure (connectivity, data platforms, stable electricity). Governance And Regulation | positive | low | investment priorities to improve adoption and impact |
0.07
|
| Policy implication: develop data governance, interoperability, and safeguards to encourage public–private collaboration while protecting smallholders. Governance And Regulation | positive | low | policy and regulatory framework quality |
0.07
|
| Recommendation: support capacity building—digital literacy, agronomic knowledge, and extension systems—to increase adoption and equitable benefits. Skill Acquisition | positive | medium | digital literacy, extension capacity, equitable adoption |
0.14
|
| Research recommendation: invest in longer-run, rigorous impact evaluations (RCTs, panel studies) and system-level assessments to capture spillovers and sustainability outcomes. Research Productivity | null_result | high | need for longer-run rigorous evaluations and system-level studies |
0.24
|