Inside the book
A story in two parts
Chapters 2–5 look back at a century of food innovation and its side effects. Chapter 6 sets out a systems framework. Chapters 7–11 look forward at five ways AI will act on the food system. Select a chapter to read its summary.
Introduction: Innovating the Agrifood SystemThe agrifood system is the largest human-engineered system on Earth – incredibly productive but also ecologically destructive.
This chapter establishes the book's central argument: agricultural innovation follows a paradoxical pattern in which technological breakthroughs solve one problem while generating new and unintended ones. Drawing on the Haber-Bosch process, the Green Revolution, and gene editing, it demonstrates that unintended consequences are not accidents but features of innovating in a complex system. The chapter challenges the Malthusian view that population growth will outpace resources, arguing instead that ingenuity is the defining human characteristic – but that managing its consequences requires systemic understanding. How do we harness AI – the next great innovation – without suffering the unintended consequences? The food system must produce 50–60 per cent more calories in the next twenty-five years from a system straining against planetary boundaries. It outlines the book's two-part structure: a historical analysis of past innovations and a forward-looking framework for understanding and anticipating AI's ability to sustainably intensify the food system.
Grow More, Grow Cheaply: The Rise of Industrial AgribusinessThis chapter analyses how the industrialisation of food production over the past 150 years transformed farming into a commodity-driven system dominated by a small number of agribusiness corporations, trapping farmers on Cochrane's Treadmill: the economic dynamic that forces continuous productivity improvement or market exit.
The chapter demonstrates how the pursuit of supply chain efficiency created global networks that are opaque, vulnerable to fraud, and structurally resistant to transparency – illustrated by the challenge of tracing a single ingredient through a complex processed food supply chain. Food waste is a systemic unintended consequence of cheap food: information asymmetries, cosmetic quality standards, and efficient production combine to waste 25–40 per cent of all food produced in the US. The chapter explores the globalisation of agricultural ingredient supply, concluding that efficiency has created abundance at the price of ecological destruction, supply chain fragility, and accountability deficits that artificial intelligence is uniquely positioned to address.
Last Longer, Travel Further: The Battle Against PerishabilityPerishability has been a powerful driver of agricultural innovation and a source of unintended consequences in the global food system.
This chapter examines the science of food spoilage – oxidation, microbial decay, ethylene production, and temperature – and analyses how innovations designed to extend shelf life and enable global supply chains impacted agricultural production, crop breeding priorities, and consumer choice. The side effects of monocultural banana production illustrate how solving perishability creates fragility. The chapter examines innovations in canning, pasteurisation, cold chain logistics, modified atmosphere packaging, and food processing, showing how each introduced new trade-offs: nutritional depletion from steel roller milling, cardiovascular risk from trans fats, and rainforest destruction from palm oil. It compares the economic logic of perishability between climacteric and non-climacteric fruit showing how shelf life determines where food can be grown, how it is priced, and what genetic traits breeders prioritise. The chapter concludes with an analysis of food miles, finding that the carbon footprint is more complex than simple distance suggests.
Evolve Faster: Unleashing the Genetic GenieThis chapter examines the consequences of genetic manipulation of food crops – from Mendelian plant breeding through transgenic genetic modification to contemporary gene editing technologies including CRISPR.
It traces the development of the first approved genetically modified food crop, through the rapid commercialisation of Monsanto's herbicide-tolerant and insect-resistant crop portfolios. GMO adoption delivered measurable benefits: reductions in chemical use, yield gains, and a lower overall toxicity for agricultural pesticides. But Monsanto's strategy – the criminalisation of seed-saving, suppression of independent research, and aggressive technology licensing – generated consumer and regulatory backlash that set biological crop innovation back by a generation. The chapter draws a crucial distinction between transgenic genetic modification and CRISPR-based gene editing, arguing that the regulatory and public conflation of the two technologies – a legacy of the GMO controversy – represents one of the most costly unintended consequences in the history of agricultural innovation and a significant barrier to deploying the AI-powered tools most needed for climate adaptation.
Regulate, Educate: The Visible Hand of the StateThis chapter examines state interventions in food and nutrition – regulation, education, and fiscal policy – and analyses their mixed record in improving diets.
It traces 50 years of US nutrition education policy, from the 1977 McGovern dietary guidelines to the 2026 dietary guidelines revision. The chapter uses the nutrition facts panel as a case study in the gap between regulatory intent and consumer behaviour: examining the methodology of calorie calculation, the manipulation of serving sizes through the RACC, and the exploitation of health claim regulations through marketing. The ingredient list is examined as both a regulatory achievement and a consumer education failure – illustrated with an analysis of processed food ingredient complexity. Drawing on the concept of embodied, intuitive food knowledge – the chapter argues that information alone is insufficient to change dietary behaviour. It concludes with a comparative analysis of sugar taxation evidence from the United Kingdom, the US and Mexico, finding fiscal intervention the most effective state tool examined – and asking what this implies for the future of regulatory policy.
Traps and LeversThis chapter applies systems thinking to explain why well-intentioned agrifood innovations consistently produce unintended consequences – and introduces the framework for Part II.
Drawing on Donella Meadows' Thinking in Systems, it identifies the recurring systemic traps of the global agrifood system: the Tragedy of the Commons; Policy Resistance generated by incumbent industry interests; Escalation dynamics between agribusinesses and pests; and Drift to Low Performance in nutrition standards. The chapter argues that these traps are features of a complex system's response to interventions targeted at symptoms. It then uses Meadows' hierarchy of leverage points – from parameters and feedback loops to the rules and goals of the system – to identify 5 levers by which artificial intelligence will shape the global food system: data creation, knowledge generation, decision-making, action-taking, and systemic understanding. These levers form the organising architecture for Chapters 7 to 11. The chapter argues that AI's transformative potential lies in its capacity to re-engineer the information flows, feedback mechanisms, and decision architectures that determine how the whole system behaves.
The Data ExplosionAI is transforming the collection, analysis, and economic value of agricultural data.
This chapter traces the computer vision revolution from the ImageNet breakthrough through the development of foundation models and analyses their deployment in agriculture: crop disease detection, precision spraying, phenotyping, yield estimation and satellite remote sensing. It introduces a new framework for categorising AI's operational impact on farm tasks. The chapter challenges the 'data is the new oil' metaphor, arguing that agricultural data is non-rivalrous, non-fungible, and subject to network effects that make it behave more like a chemical catalyst. It examines the emerging economics of ecosystem services and discusses the structural problems that have undermined agricultural carbon markets: permanence, measurement, additionality, leakage, and double-counting. The chapter argues that AI-powered data collection is a precondition for solving all five. It concludes by identifying unintended consequences of the agricultural data explosion: data privacy, sovereignty risks, a new concentration of power in agtech, and the Tragedy of the Commons of shared data.
The Commoditization of KnowledgeKnowledge is among the most asymmetrical resources in the food system – and AI will commoditise much of it, with profound consequences for competitive advantage, agronomy, and the balance of power.
This chapter introduces a taxonomy of knowledge types – explicit vs tacit, simulatable vs non-simulatable – and analyses which types AI can already replicate and which remain beyond reach. Building on Michael Polanyi's concept of tacit knowledge, it argues that the most economically valuable agricultural knowledge – the accumulated expertise of crop breeders, farmers, and agronomists – is precisely the kind that AI finds hardest to replicate. AI will simultaneously develop ineffable capabilities that exceed human expert performance. The chapter argues that as AI commoditises agronomic expertise, competitive advantage will shift from access to knowledge to the quality of judgment applied to AI-generated insights. It argues that the democratisation of agricultural knowledge through AI may increase resource consumption rather than reduce it. The chapter concludes by analysing the implications of knowledge commoditisation for education, extension services, and the future of farm advisory.
The Delegation of Decision-MakingAI is moving beyond information provision toward autonomous decision-making – a shift with profound implications for market power, industry structure, and governance.
This chapter examines AI Agents: autonomous systems that can negotiate, transact, and act without human intervention. It analyses how AI-powered personal nutrition platforms will leverage advances in nutrigenomics, microbiome science, and personalised medicine to deliver dietary guidance calibrated to individual profiles and health conditions – transforming food from a commodity into a differentiated product. It examines how AI advisory agents will democratise access to expert agronomic knowledge, reversing decades of knowledge concentration in large agribusiness corporations. The chapter analyses the systemic risks of AI Agent proliferation: the emergence of hidden collusion; the rise of price discovery; and the concentration of market influence among a small number of providers. It argues that the delegation of decision-making to AI represents the most significant and least-governed shift in the agrifood system's power architecture, and that the window for establishing appropriate regulatory frameworks is closing fast.
The Automation of PrecisionAI is transforming physical work in agriculture through autonomy, precision, and a new paradigm of Continuous Management that challenges the trend toward ever-larger equipment.
This chapter analyses the state of ag robotics – precision spraying, mechanical weeding, and fruit harvesting – and applies Moravec's Paradox to explain why tasks trivial for humans remain computationally harder than tasks that appear complex. Continuous Management is a structural shift: replacing episodic interventions by large, soil-compacting machinery with continuous, fine monitoring and treatment by lightweight autonomous robots – with implications for soil health, input efficiency, and farm size. The chapter examines vertical farming, concluding that despite substantial investment, indoor growing cannot become a meaningful source of calories at scale. It applies disruptive innovation theory to ag robotics, arguing that automation will achieve scale by enabling new business models that the economics of large machinery rendered unviable. The chapter concludes by examining unintended consequences: right-to-repair, new platform dependencies, and the risk that automation accelerates farm consolidation.
The Acceleration of DiscoveryAI is collapsing the timescales of discovery in agriculture – transforming crop breeding, protein science, and plant genomics in ways that will reshape the food system's biological foundations over the coming decades.
This chapter examines the shift from marker-assisted selection to AI-powered genomic prediction, analysing the challenges of applying AI to plant genetics: sparse phenotypic data, chromosomal polyploidy, the GxExM interaction problem, and the concentration of elite germplasm ownership. It examines the future of alternative protein – cultured meat, precision fermentation, plant-based formulations – and analyses which are scalable and which will likely remain marginal. The chapter introduces biological world models: AI systems trained to predict field performance from genetics – and argues these represent the most transformative use of AI in agriculture, analogous to AlphaFold's impact on protein structure. The chapter concludes with an analysis of unintended consequences: the risk of AI-optimised breeding convergence; the exploitation of IP; and the reallocation of value across the chain as seed companies capture margin previously earned by processors and chemical companies.
Conclusion: Forewarned is ForearmedThis chapter offers actionable recommendations for policymakers and industry leaders on the integration of AI into the global agrifood system.
Drawing on the history of electrification, it argues that AI will transform the food system through three meta-effects: improving decision confidence; redistributing tasks from centralised institutions to decentralised agents; and reallocating value between actors. The chapter identifies risks requiring governance: the concentration of AI capability in a small number of platform companies; collusion between AI Agents; erosion of farmer data sovereignty; and the risk that AI cements the existing extractive system rather than enables its transformation. Recommendations for policymakers include data sovereignty standards, outcome-based certification for agricultural AI systems, and frameworks to prevent the concentration of platform power. For industry leaders, it argues that the window to shape AI adoption is narrow – decisions made in the next five years will determine the landscape for a generation. Unlike Goethe's Sorcerer's Apprentice, we have the knowledge to anticipate the consequences. The question is whether we will act on it.
