Chapter 11 — Logistics, agriculture, and professional services

This chapter covers three sectors that share a structural characteristic: each is dominated by operations-and-knowledge-work that AI deployment substantially transforms but does not entirely replace. Logistics and supply chain operates a global infrastructure of physical movement that AI optimises but cannot eliminate. Agriculture combines biology, weather, and labour-intensive operations with technology integration that has accelerated through the 2010s and 2020s. Professional services — legal, accounting, consulting, tax — are the contemporary period’s most-prominent test of foundation-model capability against high-skilled, high-rents labour. The three sectors together describe much of the deployment landscape that this textbook’s analytical track has built up to.

The three sectors also share a specific deployment dynamic: in each, the AI integration has produced both substantial commercial value at the operations level and substantial cautionary cases at the organisational level. Maersk’s Tradelens platform — an industry-wide blockchain-and-data-sharing initiative that closed in 2022 after four years of operation — is the logistics sector’s clearest cautionary case. John Deere’s See & Spray and autonomous-tractor programmes are the canonical agricultural AI cases, with the company’s right-to-repair conflicts demonstrating the durable tensions of AI deployment on private platforms. The Mata v. Avianca case in June 2023 — where a New York lawyer submitted a brief citing fabricated legal cases generated by ChatGPT — is the most-publicised contemporary failure of professional-services AI deployment. The cumulative cautionary record is comparable to the Watson Health, Klarna, Boeing 737 MAX, and Cambridge Analytica patterns of preceding chapters.

This chapter develops the three sectors across fourteen sections. Sections 11.1–11.4 cover logistics and supply chain: the deployment landscape, maritime logistics with the Tradelens case, last-mile and middle-mile logistics, and Asian logistics with attention to Singapore and Malaysia. Sections 11.5–11.8 cover agriculture: the historical arc from GPS to ML, the John Deere case in detail, climate and labour drivers, and Australian agtech with the Malaysian palm-oil context. Sections 11.9–11.12 cover professional services: the deployment landscape, legal AI with the Mata v. Avianca case, accounting and consulting, and the labour-economic dynamics. Section 11.13 covers regional context for professional services. Section 11.14 sketches cross-sector convergences in 2026.

11.1 The logistics AI deployment landscape

Logistics — the movement of goods from origin to destination — is among the largest economic activities globally. Total logistics spend was approximately USD 9 trillion in 2024 (Armstrong & Associates estimate), representing 9% of global GDP. The sector encompasses ocean freight, air cargo, rail, trucking, parcel delivery, warehousing, customs clearance, and the broader supply-chain orchestration that ties these together. AI deployment in logistics has been substantial through 2010–2026, with the current period producing the most-mature deployments at scale.

The supply-chain optimisation arc. The conceptual foundations are older than the technology: linear programming for transportation problems was developed in the 1940s (Kantorovich, 1939; Dantzig’s simplex method, 1947); inventory optimisation under uncertainty was formalised in the 1950s; the bullwhip-effect literature (Lee, Padmanabhan, and Whang, 1997) established the contemporary supply-chain-coordination framework. The deployment of these methods at scale required the digital infrastructure (ERP systems, warehouse management systems, transportation management systems) that matured through the 1980s and 1990s. The 2010s ML wave layered statistical-learning approaches on top of this infrastructure; the 2020s foundation-model wave is now adding generative-AI capabilities for operational decision support.

The major operators. The deepest-deployed AI in logistics is at the major integrated operators. UPS, FedEx, DHL, and Amazon Logistics each operate at-scale in the parcel/express segment with substantial AI integration. Maersk and MSC dominate ocean container shipping; Lufthansa Cargo, FedEx Express, and Cathay Pacific Cargo dominate transpacific air cargo. The integrated platforms — Amazon’s combination of e-commerce, warehousing, and logistics — represent the contemporary frontier of vertical AI integration in the sector.

The ML methodologies in deployment. The mature applications include route optimisation (the UPS ORION case from Chapter 3 is canonical; FedEx, DHL, and others operate similar systems); demand forecasting (predicting parcel volumes, container demand, fuel needs); dynamic pricing (ocean rates, air freight, parcel surcharges adjust in near-real-time); workforce scheduling (driver and warehouse worker assignment given variable demand); fraud and damage detection (computer vision for parcel and container monitoring); and customs-and-trade-compliance automation. The methods range from classical operations research through gradient-boosted ML to (increasingly) deep-learning and foundation-model approaches.

The 2024–2026 generative-AI extension. The contemporary wave is extending logistics AI in specific directions. Generative-AI-supported customer service is being deployed at major operators (UPS, FedEx, DHL all have launched customer-facing AI assistants through 2024–2025). Foundation-model-supported supply-chain planning — where the AI helps planners reason about complex multi-objective decisions — is in early commercial deployment. AI-supported customs-classification and trade-compliance has reached substantial scale (the customs-classification problem maps well to AI capability). The deployment depth is uneven across the major operators; some are aggressive adopters, others are more cautious, but the direction is broadly consistent.

11.2 Maritime logistics and the Tradelens cautionary case

Maritime logistics — particularly container shipping — is the dominant mode of intercontinental trade by volume. Approximately 80% of world trade by volume moves by sea (UNCTAD, 2024); the container-shipping fleet of approximately 6,000 ships moves over 200 million TEU annually. The industry’s structural characteristics — global scale, capital intensity, narrow per-shipment margins, complex multi-party transactions — make it both a substantial AI deployment opportunity and a difficult deployment environment.

Maersk and the digital-transformation strategy. A.P. Møller-Maersk, the Danish shipping giant founded 1904, is the world’s largest container-shipping operator measured by capacity (approximately 17% global market share). The company’s revenue exceeded USD 51 billion in 2024. Maersk’s digital-transformation strategy through the 2010s positioned the company as the most-aggressive technology adopter among major shipping lines; the strategy has produced both successes and a particularly visible failure.

The 2018 Tradelens partnership with IBM. Tradelens was launched in 2018 as a joint venture between Maersk and IBM. The vision was an industry-wide blockchain-and-data-sharing platform that would digitise the document-and-information flows underlying global trade. The premise was substantial: trade transactions involve up to 30 parties (shippers, carriers, customs authorities, port operators, terminal operators, banks, insurers, freight forwarders, brokers); the document flow is paper-heavy, error-prone, and slow; a shared platform that all parties could use would substantially reduce transaction friction. IBM’s blockchain-and-distributed-ledger expertise positioned the platform’s technical architecture; Maersk’s industry weight positioned the platform commercially.

The four-year deployment. Through 2018–2022 Tradelens reached approximately 200 organisations on the platform, including 13 of the world’s largest port operators and substantial numbers of shippers, carriers, and customs authorities. The platform processed millions of shipping events. The deployment generated peer-reviewed research, industry-conference presentations, and substantial commercial signals.

The 2022 closure. In November 2022 Maersk and IBM announced that Tradelens would be discontinued by Q1 2023. The announcement was unusual in its directness: the joint statement acknowledged that “the level of commercial viability needed to underpin this product offering would not be available at the time of the planned discontinuation.” The platform shut down in early 2023.

Structural lessons from Tradelens. The post-mortem on Tradelens has been substantial; multiple analyses (Gartner, Lloyd’s List, McKinsey, academic case studies) have converged on five lessons.

Lesson 1 — industry-wide platforms require industry-wide buy-in. Tradelens succeeded in onboarding many parties but never achieved the network-effect threshold where the platform’s value to each party justified the integration investment. Specifically, several major shipping lines (MSC, CMA CGM, ONE) did not join Tradelens; their absence reduced the platform’s usefulness for shippers who used multiple carriers. A platform that aspires to industry-wide deployment must achieve industry-wide adoption; partial adoption produces negative-feedback dynamics that erode rather than build the network.

Lesson 2 — the lead vendor’s industry position can be a liability. Maersk’s leadership of Tradelens was a structural problem: competitors were reluctant to join a platform majority-owned by their largest competitor. The same platform, structured as a neutral utility (with multiple-carrier governance from the start), might have achieved different adoption dynamics. The lesson generalises: industry-wide platforms with single-vendor leadership face adoption resistance that neutral structures can avoid.

Lesson 3 — blockchain-as-foundation introduces complexity without proportionate benefit. The Tradelens use of blockchain-and-distributed-ledger technology was originally framed as enabling cross-party trust and transparent record-keeping. Subsequent technical analysis (and IBM’s eventual de-emphasis of the blockchain framing) suggests that the same outcomes could have been achieved with a conventional cloud-database architecture; the blockchain layer added implementation complexity without proportional benefit. The lesson generalises: technology-led platform design — choosing the technology first and then identifying use cases — produces worse outcomes than problem-led design.

Lesson 4 — the existing inefficiencies’ beneficiaries resist change. Trade-document inefficiencies cost the broader trade system substantially, but specific intermediaries (freight forwarders, brokers, certain banks) profit from those inefficiencies. A platform that removes the inefficiencies threatens those intermediaries’ business models; their resistance to platform adoption is rational from their perspective. The lesson generalises: AI-and-platform interventions that disrupt established economic relationships face structural resistance from the parties whose rents the intervention reduces.

Lesson 5 — the ML-and-data deployment that survived Tradelens is operationally substantial. Maersk and the broader shipping industry continue to deploy ML extensively in operations (route optimisation, fuel management, container utilisation, predictive maintenance for ships’ engines, port-operations forecasting). The deployment that survived Tradelens is the operational ML, not the industry-wide platform. The lesson: bottom-up operational AI deployment compounds; top-down platform plays are higher-risk and often fail. The pattern is analogous to the analogous patterns in healthcare (Watson Health vs operational hospital ML), retail (general platform plays vs specific operational deployments), and finance (aspirational AI banking vs successful operational deployments).

The Tradelens case is now a standard reference in industry-platform-deployment literature. The deployment of AI in maritime logistics has continued to mature through 2023–2026 — Maersk’s Captain Peter platform for shipper-facing logistics intelligence, the Hapag-Lloyd Quick Quotes pricing system, MSC’s various digital initiatives — but as operational extensions rather than as industry-transforming platforms.

11.3 Last-mile and middle-mile logistics

Last-mile logistics — the movement from local distribution to the customer doorstep — is the most-cost-intensive segment of parcel logistics. Industry analyses (McKinsey, BCG, various) consistently estimate that last-mile costs represent 40–60% of total parcel-shipping cost, with the proportion higher for lower-value-density shipments. AI deployment in last-mile is correspondingly substantial.

UPS ORION — the canonical case. UPS’s On-Road Integrated Optimization and Navigation (ORION) system, deployed across the US through 2013–2016 and globally by 2020, is the most-cited operational-research-and-ML case in logistics. ORION combines route-optimisation methods (with operations-research and ML components) for UPS drivers’ daily routes. Documented outcomes include approximately 100 million miles per year of saved driving (UPS reports), with corresponding fuel savings of approximately 10 million gallons annually and CO2 reductions of approximately 100,000 tonnes annually. The investment was substantial — over USD 250 million in development and deployment cost over a decade — but the operational savings have justified the investment many times over.

FedEx and DHL parallel deployments. FedEx operates similar route-optimisation systems through its FedEx Express network, with its own internal-development-plus-acquisition history (the 2015 acquisition of GENCO Distribution Solutions added substantial logistics-optimisation capabilities). DHL operates similar systems globally; the company’s Resilience360 platform (now sold to Everstream Analytics) addressed broader supply-chain visibility with ML-based components. The major parcel operators have converged on broadly similar AI-deployment patterns: route optimisation, demand forecasting, capacity planning, customer-service automation, fraud detection.

Amazon Logistics — the integrated approach. Amazon Logistics represents a structurally different model. Where UPS, FedEx, and DHL are pure-play logistics operators, Amazon integrates logistics with the e-commerce platform and the AWS infrastructure. The integration produces specific advantages: Amazon’s demand forecasts are based on actual purchase data rather than aggregated shipper signals; the inventory placement decisions can be optimised against fulfilment-cost and customer-experience tradeoffs simultaneously; the logistics network can be designed for Amazon’s specific shipment mix rather than the general parcel volume. By 2024, Amazon Logistics handled approximately 5 billion packages annually, surpassing both UPS and FedEx in US ground delivery volume. The integration with Amazon Robotics’ warehouse automation (Chapter 8 covers in detail), Amazon’s own air cargo operations (Amazon Air, launched 2016), and the broader e-commerce demand-and-supply system produces a deployment depth that the pure-play operators cannot match without similar vertical integration.

The autonomous-vehicle logistics dynamic. The 2010s expectation that autonomous trucks and last-mile vehicles would substantially transform logistics by mid-2020s has not materialised. The 2024–2025 closures of major autonomous-trucking efforts (Aurora Innovation has continued; the broader category has consolidated substantially) and the limited deployment of autonomous last-mile delivery (Cruise’s exit from the autonomous-taxi market in late 2024 was indicative of broader pressure on the AV-deployment thesis) reflect the harder-than-anticipated technical and operational challenges. The 2026 state is operational autonomy at narrow scales (Waymo and a small number of others operating commercial robotaxis in specific cities; some autonomous-trucking operations on specific highway segments) rather than the broad autonomous-vehicle deployment that earlier projections suggested.

Drone delivery. Wing (Alphabet subsidiary, founded 2012), Zipline (founded 2014), Amazon Prime Air (announced 2013, launched commercial operations 2022), and various other drone-delivery operators have produced operational drone delivery at narrow scales. Zipline’s deployment in Rwanda for medical-supply delivery (operational since 2016) has been the most-mature commercial drone-delivery deployment globally; the company has expanded to other African countries, the United States, and Japan. The general parcel-delivery use case has been more constrained by regulatory and operational considerations; drone delivery in 2026 is operational but represents a very small fraction of total parcel volume.

11.4 Asian logistics — Singapore, Malaysia, regional

The Asian logistics context — particularly the East-and-Southeast Asian shipping-and-logistics infrastructure — is central to global trade. The Asia-Pacific region accounts for approximately 60% of global container throughput; the major Asian ports dominate the world’s busiest-port rankings (Shanghai, Singapore, Ningbo-Zhoushan, Shenzhen, Guangzhou, Qingdao, Hong Kong, Busan, Tianjin, and Klang are typically in the top 15).

Singapore as logistics hub. Singapore’s positioning as a global logistics hub combines several structural advantages: geographic positioning at the Strait of Malacca (the world’s busiest shipping lane); deep harbour infrastructure; comprehensive air-cargo connectivity via Changi; political stability and rule-of-law; English-language-and-multinational-business environment. The Port of Singapore Authority (PSA International) operates one of the world’s most-sophisticated container terminals, with substantial automation including the Tuas Mega Port (under progressive activation through 2024–2030 to be the world’s largest fully-automated container terminal at full build-out). The Maritime and Port Authority of Singapore (MPA) operates extensive AI-deployment programmes; Singapore’s “TradeNet” digital-customs platform (operational since 1989, continuously upgraded) is among the world’s longest-running digital-trade-facilitation systems. Specific Singapore-based logistics-AI firms include Stowmaster (vessel stowage planning), Portcast (predictive container shipping ETAs), and various others.

Port Klang Malaysia. Port Klang (officially Pelabuhan Klang), located approximately 40 km south of Kuala Lumpur, is Malaysia’s primary container port and one of the world’s busiest. The port handled approximately 14 million TEU in 2024, ranking 12th globally by container throughput. The port’s two operators — Westports Holdings (Bursa Malaysia-listed) and Northport (a Malaysia Marine and Heavy Engineering subsidiary) — operate substantial automation programmes. Westports’ digital-transformation programme through 2018–2025 has been particularly substantial; the company’s operations-management infrastructure incorporates ML-based vessel scheduling, equipment optimisation, and predictive maintenance for terminal equipment.

Malaysian logistics cluster. Beyond Port Klang, Malaysia hosts substantial logistics infrastructure: Penang Port (sea); Johor Port (sea, with the strategic position adjacent to Singapore); KLIA’s air-cargo facilities; the network of inland container depots and logistics parks (notably Westport’s distribution-and-logistics complexes and the Kuala Lumpur South distribution centre cluster). Malaysian logistics firms include POS Malaysia (the national postal service, with substantial logistics-and-parcel diversification), GDEX (regional courier), Tasco (broader logistics), and many others. AI deployment in Malaysian logistics is at mid-scale relative to global frontiers — operationally substantial but not at the deployment depth of Singapore or the global leaders.

Cross-border and ASEAN logistics. The ASEAN region’s economic integration through the ASEAN Economic Community framework (operational since 2015) has produced increasing cross-border logistics complexity that AI deployment helps address. Specific applications include cross-border customs facilitation (the ASEAN Single Window connects member-state customs systems), regional logistics-network optimisation (where shippers move goods across multiple ASEAN countries with varying infrastructure), and the regional e-commerce logistics infrastructure that supports Shopee, Lazada, and Tokopedia (Chapter 8). The 2024–2026 trajectory has produced substantial new investment in ASEAN logistics infrastructure, with Chinese-and-other operators (J&T Express, Cainiao, SF Express) entering the region competitively.

The Indonesian and Vietnamese context. Indonesia, the region’s largest economy by GDP, hosts substantial logistics activity dominated by domestic firms (J&T Express, JNE, Pos Indonesia, Tiki) operating across the country’s archipelagic geography (the geographic complexity is itself a structural challenge that AI helps address). Vietnam’s logistics sector has grown rapidly through 2018–2024 with the country’s broader economic expansion; Saigon Newport, Vinalines, and the major Vietnamese express operators have all been substantial AI adopters. The region’s logistics-AI deployment depth is uneven across countries; Singapore and Malaysia lead, with Vietnam and Indonesia rapidly catching up.

11.5 The agricultural AI arc — from GPS to ML

Agriculture’s technology integration has been continuous over a longer period than the contemporary AI literature emphasises. The first GPS-supported precision-agriculture systems were deployed in commercial farming in the early 1990s. Variable-rate application (VRA) of fertilisers and pesticides — based on within-field variation in soil and crop conditions — became widespread through the 2000s. Sensor-driven monitoring (soil moisture, weather, crop development indices via satellite) accelerated through the 2010s. The 2020s ML wave is layering deep-learning-based computer vision, robotics, and increasingly foundation-model-based agronomic decision support on top of this foundation.

The major equipment OEMs. The agricultural-equipment industry is concentrated among three major players: Deere & Company (with the John Deere brand), AGCO Corporation, and CNH Industrial (the parent of Case IH and New Holland). All three have been substantial AI investors through 2010–2026. The OEMs’ positioning is structural: agricultural equipment costs have been rising over decades; the business model has shifted from one-time equipment sales toward equipment-plus-service-and-software bundles where the AI-and-data services are increasingly central to the value proposition.

The agtech startup ecosystem. Beyond the equipment OEMs, a substantial agtech startup ecosystem has emerged through the 2010s and 2020s. Specific company categories include: precision agriculture (Climate FieldView, originally a Climate Corporation product, now a Bayer brand following the 2018 Monsanto acquisition); farm-management software (Granular, FarmLogs, Agworld); robotics and autonomy (Iron Ox, Plenty, AeroFarms in the indoor-vertical space; multiple open-field robotics startups); satellite and sensor-based monitoring (Planet, Indigo Agriculture, Taranis); livestock-management technology (various); and the broader AI-supported agronomy platforms.

The specific ML applications. The mature applications include yield prediction (combining weather, soil, and crop-development sensor data); pest-and-disease detection (from satellite, drone, or ground-based imaging); precision-application recommendations (where to apply fertiliser, pesticide, irrigation); equipment-routing optimisation (analogous to logistics route optimisation but adapted to field operations); harvest-quality prediction; and supply-chain integration with food-processing-and-distribution operations. The methodologies range from classical regression and decision trees through gradient-boosted ML to deep-learning-based computer vision.

The deployment pattern. Agricultural AI deployment is characterised by uneven adoption across farms. The largest and most-technologically-sophisticated farms (typically Midwestern US row-crop operations of 1,000+ hectares; large Australian wheat-and-livestock operations; Brazilian soybean operations; certain European and East-Asian operations) have adopted the technology aggressively. Smaller and less-sophisticated farms have adopted less; the technology adoption gap is one of the structural features of contemporary agriculture, with implications for productivity divergence within the sector.

11.6 John Deere — the canonical agricultural AI case

John Deere (Deere & Company; founded 1837 in Grand Detour, Illinois; Moline, Illinois headquarters since 1848) is the world’s largest agricultural-equipment manufacturer and the agricultural sector’s most-publicised AI case. The company’s revenue exceeded USD 51 billion in fiscal 2024; agricultural equipment accounts for approximately 60% of revenue, with construction-and-forestry and financial services contributing the balance. Deere’s AI-and-precision-agriculture strategy has been substantial through 2010–2026 and is the canonical reference for understanding agricultural AI deployment at scale.

See & Spray — the Blue River Technology acquisition. Deere acquired Blue River Technology in September 2017 for USD 305 million. Blue River, founded 2011 in Sunnyvale, had developed computer-vision-driven targeted herbicide application technology — the See & Spray system. The system uses cameras mounted on a sprayer to identify weeds in real time as the sprayer moves through a field, applying herbicide only to the weeds rather than broadcast-spraying the entire field. The herbicide reduction is substantial (90%+ in published trials); the cost savings and environmental benefits have been documented across multiple deployments. See & Spray Ultimate, the production version released in 2022, has been integrated with Deere’s broader sprayer line; cumulative deployments exceeded several thousand units by 2024.

Autonomous tractor programme. Deere’s autonomous-tractor programme reached commercial deployment in 2022 with the launch of the autonomous 8R tractor with TruSet technology. The system enables farmers to deploy tractors in tilling and other field operations without an operator in the cab; the farmer monitors operations via a mobile interface. The 2024 expansion to additional field operations (planting in the autonomous 8R; tillage in the autonomous 9RX) extended the deployment scope. Deere’s stated ambition is fully-autonomous farm operation by 2030; the practical state of deployment in 2026 is autonomous tillage and certain other operations at substantial scale, with autonomous planting and harvesting still at early commercial deployment.

The data-asset question. A particularly contested dimension of Deere’s AI strategy concerns data ownership. Deere’s John Deere Operations Center platform (launched 2014) collects extensive operational data from farms using Deere equipment — yield maps, application records, equipment usage, soil and weather observations. The data collection has produced one of the largest farm-operations datasets globally. The contested question is who owns and controls this data: the farmer (whose operations generated the data), Deere (whose platform aggregates the data), or some combination. Deere’s terms of service have evolved through 2014–2024 to specify rights more clearly; the framework currently positions Deere as the data custodian with specific use rights, while farmers retain ownership of their underlying operational data. The question remains contested by farm-and-agriculture advocacy groups; the resolution has implications for the broader AI-platform-data-rights landscape.

The right-to-repair conflict. A specific durable conflict between Deere and farm operators concerns the right to repair Deere equipment. Deere’s policy through the 2010s and 2020s has been to require that certain maintenance and repair activities use Deere-authorised dealers and Deere-authorised parts; software-locks on equipment electronics prevent third-party servicing of certain components. Farm advocates and right-to-repair supporters have argued that this restriction reduces farmer agency, increases repair costs, and slows agricultural operations during time-sensitive periods (harvest, planting). The conflict has produced state-level legislation in several US states (Massachusetts, Minnesota, New York have enacted right-to-repair laws affecting agricultural equipment), federal regulatory attention (the FTC’s 2021 right-to-repair report; subsequent enforcement actions), and direct industry agreements (Deere and the American Farm Bureau Federation signed a 2023 Memorandum of Understanding committing to expanded farmer-and-third-party repair access). The conflict’s underlying tension — between the equipment manufacturer’s commercial-and-IP interests and the equipment owner’s autonomy interest — generalises to many AI-platform contexts where the manufacturer retains substantial control over deployed systems.

Climate FieldView — the Bayer competitor. The major competitor to Deere’s data-and-platform positioning is Climate FieldView, originally a Climate Corporation product. Climate Corporation was founded in 2006, acquired by Monsanto in 2013 for USD 930 million, and became a Bayer asset following Bayer’s 2018 Monsanto acquisition. FieldView is positioned as an equipment-agnostic farm-management platform — it integrates data from John Deere, AGCO, CNH Industrial, and other equipment, providing the farmer a unified data interface. The strategic positioning is structurally different from Deere’s: FieldView aspires to be the neutral data layer; Deere’s equivalent platforms aspire to be Deere-equipment-aligned. The competitive dynamics have shaped both companies’ strategies through 2018–2026; Deere has progressively opened its platform to non-Deere equipment in response to competitive pressure.

Structural lessons from the Deere case. Three lessons generalise from the Deere case to broader AI deployment.

Lesson 1 — operational AI compounds, but platform AI is contested. Deere’s See & Spray, autonomous tractors, and operational ML deployments have produced documented commercial value with relatively modest controversy. The Operations Center platform — which aspires to be the data-and-platform foundation — has produced more substantial commercial-and-policy controversy. The pattern echoes the Tradelens lesson (Section 11.2): operational AI works; industry-platform AI is harder.

Lesson 2 — IP and control questions accumulate over time. The right-to-repair conflict reflects a long-running accumulation of control choices that, individually, were defensible from Deere’s perspective but cumulatively produced substantial customer-and-regulatory backlash. The pattern generalises to AI platforms: control choices that seem reasonable in isolation can accumulate into problematic positions; periodic strategic review of control structures is necessary.

Lesson 3 — the data-asset question is durably contested. The question of who owns the data generated by AI-equipped equipment is unsettled across multiple sectors (Deere in agriculture; Tesla in automobiles; various equipment manufacturers in industrial settings). The resolution will substantially shape how AI deployment economics work over the next decade; it is not obviously a question with a single right answer.

11.7 Climate, sustainability, and the labour question in agriculture

Three external drivers shape contemporary agricultural AI deployment beyond the technology dynamics: climate volatility, labour availability, and sustainability requirements. Each interacts with AI deployment in specific ways.

Climate volatility. The 2014–2024 period produced increasing weather volatility globally, with documented impacts on agricultural productivity. The 2018 European heat-and-drought, the 2021 Pacific Northwest heat dome, the 2022 Western US megadrought, the 2023 South Asian monsoon disruption, and many other specific events produced agricultural disruptions of substantial magnitude. The volatility increases the value of AI-supported agronomic decision-making — predictive weather analysis; adaptive crop selection; precision irrigation and input management; forecasting that supports forward-pricing and insurance decisions. The deployment trajectory through 2026–2030 is therefore likely to be steeper than the underlying technology dynamics alone would suggest.

The labour-shortage drivers. Agricultural labour availability has been declining in advanced economies for decades. In the United States, the H-2A visa programme (which authorises temporary agricultural foreign workers) has expanded from approximately 50,000 visas annually in the early 2000s to approximately 380,000 in 2023; the demand for additional H-2A workers consistently exceeds availability. In Australia, the seasonal-worker programme and Pacific Australia Labour Mobility (PALM) scheme have similarly expanded, with persistent labour-shortage reports during peak agricultural seasons. The labour shortage is not a temporary fluctuation; it reflects structural demographic-and-economic dynamics (urbanisation; reduced agricultural-labour preference; alternative employment opportunities) that are unlikely to reverse. AI-and-automation deployment in agriculture is partly a response to the labour-availability constraint; investment in autonomy increases as labour becomes more expensive and less reliable.

Vertical and indoor farming. Vertical farming — growing crops in stacked indoor environments with fully-controlled lighting, temperature, water, and nutrients — has been a high-profile AI-and-agriculture frontier. Major operators have included AeroFarms (founded 2004, Newark NJ; emerged from bankruptcy in 2024 under new ownership); Plenty (founded 2014, San Francisco; substantial Series E funding through 2024); Bowery Farming (founded 2015, New York; closed November 2024 after multiple operational challenges); Infarm (Berlin-based; substantial 2022–2023 retrenchment). The category has produced substantial venture investment but inconsistent commercial outcomes. The operational economics — high capital costs for facility construction; substantial energy costs for lighting (as the AI-energy intersection of Chapter 10 implies); narrow margins on the produce categories that vertical farming targets — have proven harder than the early-2020s projections suggested. The 2026 state of vertical farming is operational at narrow scale, contested at broader scale: certain produce categories (leafy greens; herbs; some berries) have viable indoor operations; the broader claim that vertical farming would substantially displace conventional outdoor agriculture has not materialised.

The economic-realism check. A specific concern with agricultural AI is that the deployment economics, while favourable for the largest farms, do not extend to the broader agricultural sector. Median farm size globally is small; smallholder agriculture (farms under 5 hectares) accounts for approximately 80% of global farm units. The capital costs of AI-and-precision-agriculture infrastructure are largely incompatible with smallholder economics; the deployment is therefore concentrated in the largest farms, with productivity divergence within the agricultural sector as a structural consequence. The 2026 question is whether AI deployment can scale down to smaller-farm contexts (mobile-app-based decision support; shared-services-and-rental models for AI-equipped equipment; satellite-based monitoring that does not require farm-level infrastructure investment) or whether the technology’s deployment will remain concentrated at the largest-farm level. The answer matters substantially for the broader economic and food-security implications.

11.8 Australian agtech and the Malaysian palm-oil context

The Australian and Malaysian agricultural contexts are substantial, with distinct AI-deployment patterns shaped by their different agricultural structures.

Australian agriculture. Australia is among the world’s most-significant agricultural exporters: wheat, beef, lamb, wool, dairy, grains, sugar, wine, horticulture. The 2024 agricultural production value exceeded AUD 90 billion. The sector’s structure differs from the US Midwestern row-crop pattern that dominates the agricultural-AI literature: Australian farms are generally larger (median wheat-farm size is 1,500+ hectares); operations are more dispersed across the continent; climate variability is structurally larger than in most other major agricultural producers; and the sector has substantial export-orientation (approximately 70% of agricultural production is exported).

The Australian agtech ecosystem. Australia has produced one of the more substantial agtech startup ecosystems globally. Specific firms include AgriDigital (post-trade-and-supply-chain platform for grain), AGUR (livestock-management software), Agworld (farm-management platform with substantial international expansion), and Maia Grazing (livestock-grazing-management software). The Australian agtech ecosystem benefits from substantial government and CSIRO research support; the CSIRO Agriculture and Food research division has been a substantial contributor to applied agricultural AI research. University-based agtech research is concentrated at the University of Queensland (the Queensland Alliance for Agriculture and Food Innovation), the University of Sydney, the University of New England, and Monash University.

The Australian agtech challenge. Despite the substantial ecosystem, Australian agtech firms have faced specific commercial challenges. The domestic market is small (relative to the US or EU); international expansion is necessary for scale, with the substantial costs and complexity that international expansion entails. The 2022–2024 venture-funding compression has been particularly difficult for agtech firms with longer paths to scale. Specific firm failures and consolidations through 2023–2025 (specific notable cases include the 2023 closure of CropLogic, the 2024 wind-down of Smart-Apply) have shaped the sector’s contemporary scale. The 2025–2026 trajectory has shown some recovery, particularly for firms with demonstrated international traction.

Malaysian palm-oil and AI deployment. Malaysia is the world’s second-largest palm-oil producer (after Indonesia), with approximately 30% of global palm-oil production. The sector’s structure is dominated by large plantation operators (Sime Darby Plantations, FGV Holdings, IOI Corporation, Kuala Lumpur Kepong, United Plantations) operating extensive plantation areas. Sime Darby Plantations alone manages approximately 600,000 hectares globally, with substantial Malaysian and Indonesian operations.

Sime Darby Plantations and the AI-deployment context. Sime Darby Plantations’ AI-deployment programme through 2018–2025 has been substantial. Specific applications include drone-and-satellite-based monitoring of plantation health, AI-driven yield prediction, harvest-scheduling optimisation, and computer-vision-based fruit-quality assessment at processing facilities. The company’s 2023 announcement of a substantial multi-year AI-investment programme (with stated commitments to digital transformation including AI capability) has been one of the more-prominent corporate AI commitments in Malaysian agriculture. The sustainability dimension is particularly important for Malaysian palm-oil — the sector faces persistent international scrutiny on deforestation, labour practices, and biodiversity impacts; AI-supported monitoring of plantation operations is partly a response to the certification-and-traceability requirements imposed by major buyers (the Roundtable on Sustainable Palm Oil; the EU Deforestation Regulation effective 2024).

The labour and rights questions in Malaysian palm oil. A specific concern in Malaysian palm-oil is labour conditions. The sector has historically depended on Indonesian and Bangladeshi migrant labour with documented labour-rights concerns. The US Customs and Border Protection’s 2020 Withhold Release Order against several Malaysian palm-oil operators (Sime Darby Plantations and FGV Holdings, both of which were subsequently lifted following remediation) reflected this dynamic. AI deployment in Malaysian palm-oil has not directly addressed the labour-conditions question; the deployment is concentrated in operations and yield-management rather than in labour-monitoring or worker-welfare. The structural tension — between AI deployment for productivity and broader sustainability/labour requirements — is a feature of the Malaysian agricultural-AI landscape that broader literature has not extensively addressed.

The other Malaysian agricultural sectors. Beyond palm oil, Malaysian agriculture includes substantial natural-rubber production, paddy/rice cultivation, fruit horticulture (durian, banana, papaya, and increasingly the export-oriented stone-fruit and citrus sectors), and aquaculture. AI deployment in these sectors is generally less advanced than in palm oil; the smaller-scale producer base does not support the same deployment depth. The Department of Agriculture Malaysia and the Malaysian Agricultural Research and Development Institute (MARDI) operate research programmes that include AI-and-precision-agriculture components, with deployment focus on smallholder applications.

11.9 The professional services AI landscape

Professional services — legal, accounting, consulting, tax, and adjacent knowledge-work sectors — have been the contemporary period’s most-prominent test of AI’s capability to handle high-skilled labour. The deployment landscape is substantial; the structural questions about labour displacement, value capture, and the future of professional work are unresolved.

Why professional services is structurally amenable to AI. Three properties of professional services produce particular AI-deployment salience. First, the work is largely text-based; foundation models excel at text. Second, the work involves substantial template-application — much legal, accounting, and consulting work follows established patterns that AI can learn and apply. Third, the labour costs are high (professional services labour is among the highest-paid categories of work globally), which makes the cost-savings from AI substitution particularly large in absolute terms. The combination produces deployment economics that are favourable to AI adoption.

The “knowledge work” displacement question. A specific question that has dominated discussions of professional-services AI is whether the deployment will substantially displace knowledge workers, substantially augment them, or fall in some intermediate position. The empirical evidence through 2024–2026 suggests substantial augmentation rather than substantial displacement: professional-services employment has not declined in aggregate, but the work-mix within firms has shifted (more time on judgment-heavy and client-relationship work; less time on routine document review, basic research, simple analytical tasks). The pattern is consistent with the broader Acemoglu and Restrepo (2020) framework: AI augments labour where labour can adapt, displaces labour where adaptation is harder, and the net effect is heterogeneous across roles.

The major professional-services segments and their AI dynamics. Legal services has the most-detailed deployment narrative (Section 11.10) including specific cautionary cases. Accounting and audit (Section 11.11) has been a steady but lower-profile deployment domain. Management consulting (also Section 11.11) has produced some of the most-aggressive AI-investment trajectories. Tax preparation has matured AI deployment for over a decade (Intuit’s TurboTax, H&R Block’s Tax-Pro Review) with the 2024–2026 generative-AI extension producing further capability increases. Real-estate services (residential and commercial transactions; valuation; portfolio management) have substantial but uneven deployment. Insurance underwriting and claims processing (covered partly in Chapter 7 for healthcare insurance and Chapter 6 for property-and-casualty) have been long-running AI-deployment domains.

11.11 Accounting, audit, and consulting AI

Accounting and consulting are the other major professional-services segments where AI deployment has been substantial.

The Big Four — PwC, Deloitte, EY, KPMG. The Big Four accounting-and-consulting firms collectively employ over 1.5 million professionals globally and operate in essentially every major economy. AI deployment at the Big Four has been substantial through 2018–2026. PwC announced a USD 1 billion AI investment in 2023 with subsequent expansions; Deloitte’s parallel investments have included specific partnerships with NVIDIA and Anthropic; EY’s wavespace innovation-platform extends AI capabilities; KPMG’s parallel programmes have been similarly substantial. The deployment focus has been on three categories: audit-and-assurance (where AI supports the audit process by analysing transactions, identifying anomalies, and supporting audit-evidence review); tax services (where AI supports tax-return preparation, planning, and advisory); and consulting (where AI supports specific advisory engagements and the broader firm-knowledge-management infrastructure).

Audit and assurance applications. AI in audit has been deployed for over a decade, with major capabilities including transaction-anomaly detection (analysing large transaction populations for unusual patterns); compliance-and-controls evaluation (assessing whether processes operate as documented); analytical procedures (statistical analysis of financial data for audit-relevance signals); and increasingly generative-AI-supported drafting of audit working papers. The 2023–2025 generative-AI extensions have produced specific tools — KPMG’s KPMG Clara, PwC’s GenAI Audit, EY’s Helix, Deloitte’s Omnia — that integrate AI capabilities with the underlying audit infrastructure. The deployment is not without controversy; the Public Company Accounting Oversight Board (PCAOB) has issued specific guidance on AI use in audits, with continuing attention to the audit-quality implications.

The McKinsey, BCG, Bain consulting transformation. The major management-consulting firms (McKinsey & Company, Boston Consulting Group, Bain & Company, plus the consulting arms of the Big Four) have substantially integrated AI into their service delivery through 2022–2026. McKinsey’s QuantumBlack division (originally an analytics firm, acquired in 2015) operates as the firm’s AI-and-analytics centre of excellence. BCG’s Gamma division and Bain’s Vector division provide similar functions. The consulting firms’ AI deployment is double-edged: the firms use AI to deliver client engagements more efficiently; they also advise client companies on AI strategy and deployment. The combination produces a substantial flywheel — the firms learn from client deployments and apply learning to their own practice.

The internal-vs-external use distinction. A specific tension in professional-services AI deployment is between internal use (the firm uses AI to improve its own operations) and external use (the firm uses AI to deliver client services more efficiently). The two have different implications. Internal use captures cost savings within the firm; external use either captures additional client value or compresses client billing. The contemporary pattern (visible across legal, accounting, and consulting) is mostly internal-cost-savings without proportional billing reduction; the firms capture the AI-driven productivity gains as profit rather than passing them through to clients in the form of lower fees. The structural question is how durable this pattern is; client pressure for AI-driven fee compression has been growing through 2024–2026, with implications for the firms’ economics.

Tax preparation AI. Intuit’s TurboTax has been one of the longest-running consumer-AI deployments in professional services, with capabilities developed continuously since the early 2000s. The 2023–2024 generative-AI extensions to TurboTax — including the “Intuit Assist” conversational interface — have substantially extended the product’s capability. H&R Block has launched comparable capabilities. The competitive dynamic has been substantial: Intuit’s market position remains strong but contested; the foundation-model capability has lowered the barriers to entry for newer entrants. The 2024–2025 venture-funded entrants in tax-preparation AI (specific firms include several stealth-stage ones whose visibility is still emerging) reflect the category’s continued attractiveness.

11.12 The professional-services labour and economic question

The structural question of professional-services AI is not the technical-deployment question (the technology works) but the labour-and-economic question: how does the value created by AI deployment distribute among professionals, firms, and clients?

The displacement projection vs reality. Early-2020s projections from various consulting firms (notably the Goldman Sachs 2023 economic analysis projecting 300 million job losses globally from generative AI deployment, with substantial concentration in white-collar professional services) suggested that professional-services employment could decline substantially through the 2020s. The actual employment trajectory through 2024–2026 has been different. Professional-services employment in the United States grew through 2023 and 2024; legal-services employment, accounting-services employment, and consulting-services employment have all increased modestly rather than decreased. The pattern is consistent with augmentation rather than displacement: AI is adopted; productivity per professional increases; the productivity gain is partially absorbed by service-volume growth rather than entirely by employment reduction.

The augmentation pattern. The deployment pattern observable in 2026 is augmentation with shifted task mix. Junior professionals (associate-level lawyers, junior accountants, entry-level consultants) face the most direct AI substitution for the routine work that previously dominated their early-career experience. Senior professionals (partners, directors, principals) capture more of the value as their judgment-and-relationship-driven work is amplified by AI capability rather than substituted. The career-progression implications are substantial: traditional pathways (junior tasks → mid-career synthesis → senior judgment) are disrupted; the question of how junior professionals develop the experience that senior judgment requires is unresolved. The pattern has been a particular topic of internal-firm debate at Big Law and the Big Four through 2024–2026.

The economic-rents question. A specific concern is who captures the AI-driven productivity gains. The current pattern (firms capture gains as profit; client fees do not decline proportionally) sustains the firms’ economics in the short term but creates structural tension. Clients increasingly perceive that AI-driven productivity gains should be shared; some clients (particularly large corporate clients with internal AI capability) are pushing aggressively for fee compression. The 2025–2026 negotiation between firms and major clients on the appropriate share of AI gains is one of the contemporary period’s most-significant professional-services dynamics.

The legal-services-specific case. Legal services has the most-detailed economic analysis of these dynamics. The Stanford CodeX “AI in Law” research programme and various academic-and-industry studies have produced specific estimates: AI-augmented lawyers can complete certain tasks (contract review; due-diligence analysis; legal research) 30–80% faster than non-augmented lawyers, with comparable or better accuracy. The implication for hourly billing — if AI-augmented lawyers spend less time on each task — would be substantial fee compression. The actual pattern has been more nuanced: hourly billing for specific tasks has compressed somewhat, but firms have offset this by shifting some work to alternative-fee structures (fixed-fee engagements; subscription-based legal services; outcome-contingent fees) that align firm revenue with client outcomes rather than with time spent. The transition is uneven; some firms have moved aggressively to alternative-fee structures, others have resisted; the long-run equilibrium is unsettled.

11.13 Regional context — Australia and Malaysia in professional services

The Australian and Malaysian professional-services contexts have specific characteristics that shape AI deployment.

Australian professional services. The Australian legal profession is structured around six major commercial firms (Allens, King & Wood Mallesons, Herbert Smith Freehills, Clayton Utz, Ashurst, MinterEllison) plus specific other practices. The accounting profession is dominated by the Big Four with substantial domestic firms (Pitcher Partners, BDO, Grant Thornton). Management consulting includes the Big Three (McKinsey, BCG, Bain) plus the Big Four consulting arms and substantial domestic firms. AI deployment in Australian professional services has followed the global pattern; the Australian-domiciled firms have generally adopted Harvey AI, Casetext, or comparable tools through 2023–2025. The Australian legal-professional-conduct framework has issued AI-use guidance broadly aligned with the American Bar Association’s approach; specific state law-society guidance has continued to develop through 2024–2026.

The Australian-specific dynamics. Australia’s smaller legal-services market and distinctive regulatory context (the Australian Solicitors’ Conduct Rules; specific state-bar variations; the broader Australian competition-and-consumer law framework) produce specific deployment patterns. The 2024 ABA-style attention to AI-related professional-conduct issues has been mirrored in Australian state-bar action. Specific Mata v. Avianca-style failures in Australian courts have not (as of 2026) reached the same publicity-and-discipline level as the US cases, but isolated incidents have surfaced.

Malaysian professional services. Malaysia’s professional-services sector includes substantial domestic firms (Skrine, Shearn Delamore, Lee Hishammuddin Allen & Gledhill, and many others in legal; the Malaysian arms of the Big Four plus substantial domestic accounting firms; the Malaysian arms of major consulting firms plus domestic firms). The professional-services economic context differs from advanced economies in scale and structure: smaller domestic market; substantial cross-border practice (particularly Singapore, ASEAN, China); important role of regulatory-and-government-relations practice given the country’s industrial structure. AI deployment in Malaysian professional services is at mid-scale relative to global frontiers; the major firms have adopted some AI capability through partnerships with global tools, but the deployment depth is substantially below the largest global firms.

The cross-border services question. A specific dimension of regional professional services concerns cross-border practice. Singapore-based firms substantially serve Malaysian, Indonesian, and Vietnamese clients; Hong Kong-based firms serve Greater China and the broader region; the Big Four operate as global networks. AI deployment that supports cross-border services — translating between languages, understanding multiple-jurisdiction requirements, integrating diverse data sources — produces substantial value in the region’s specific structure. The 2024–2026 trajectory has produced increasing integration of cross-border AI capabilities; specific regional consortia of firms have explored shared AI infrastructure with mixed commercial outcomes.

11.14 The 2026 frontier across the three sectors

The logistics, agriculture, and professional-services sectors operate on different time-constants but share specific cross-sector convergences in 2026 that warrant attention.

Convergence 1 — the operational-AI compounding pattern. Each sector exhibits the same broad pattern: operational AI (route optimisation in logistics; precision-application in agriculture; document-automation in professional services) has produced demonstrable compounding value over years. Industry-platform AI (Tradelens in logistics; Operations Center contests in agriculture; the cross-firm-data-sharing aspirations in professional services) has been more difficult. The pattern matches the patterns in finance (operational ML works; platform-AI-banking is harder) and healthcare (operational hospital ML works; platform-medical-AI is harder). The cross-sector durability of the pattern is informative: AI deployment compounds within firms; across-firm coordination is structurally harder.

Convergence 2 — the data-rights question across sectors. Each sector exhibits the question of who owns and controls AI-relevant data. Agricultural data ownership (Section 11.6); legal-and-professional-services data (the question of whether client-facing AI tools’ learning is firm property or client property); logistics data (the Tradelens debates were partly about data ownership). The cross-sector resolution of these questions is unsettled; the pattern is likely to produce regulatory action in multiple sectors over the 2026–2030 window.

Convergence 3 — the labour-displacement-and-augmentation pattern. Each sector exhibits AI augmentation rather than displacement, with specific role-and-task heterogeneity. Agricultural labour displacement is concentrated in the specific roles that autonomous-equipment substitutes (tractor operators in tilled-row agriculture); logistics labour displacement is concentrated in routine warehouse-and-driving roles; professional-services displacement is concentrated in junior-professional document-and-research tasks. The augmentation-and-displacement balance is similar across sectors; the broader implication is that the 2024 macroeconomic projections of substantial-employment-loss have not played out as projected, but specific role categories have been substantially affected.

Convergence 4 — the cross-sector AI-infrastructure integration. The three sectors increasingly use the same underlying AI infrastructure: foundation models (typically OpenAI, Anthropic, or Google Cloud-based); cloud-compute (typically AWS, Azure, or Google Cloud); specific operational-AI platforms (Snowflake, Databricks, others). The cross-sector dependence on this infrastructure produces both the efficiency benefits (shared capability across sectors) and the risk concentration (sector dependence on a small number of foundation-model and cloud providers). The infrastructure question is the same one that runs through Chapter 10’s analysis; the sector-specific implications are similar.

Convergence 5 — the regulatory-environment evolution. Each sector faces ongoing regulatory adaptation to AI deployment. Logistics and customs regulation (the global trade-and-platform questions); agricultural regulation (data ownership; sustainability; equipment certification); professional-services regulation (the bar-association and accounting-board responses). The cross-sector regulatory pattern is uneven but consistent: existing regulatory frameworks adapt to AI deployment with substantial lag; the adaptation produces both protective effects (against premature deployment) and friction (against beneficial deployment). The 2026–2030 trajectory will be substantially shaped by how this regulatory adaptation continues.

The logistics, agriculture, and professional-services landscapes have produced some of the contemporary AI period’s most-detailed cases — both successes and cautionary tales. Together with the finance, healthcare, retail, manufacturing, and marketing-media-energy sectors of preceding chapters, these sectors cover much of the major commercial-deployment landscape. Chapter 12 covers the remaining sector contexts and the cross-sector themes that the comprehensive analysis raises.

References for this chapter

Logistics and supply chain

  • Lee, H. L., Padmanabhan, V., and Whang, S. (1997). Information distortion in a supply chain: The bullwhip effect. Management Science 43(4): 546–558.
  • Armstrong & Associates (2024). Global Logistics Market Report.
  • UNCTAD (2024). Review of Maritime Transport.
  • McKinsey & Company (2024). Last-mile delivery: Future scenarios and AI deployment.
  • A.P. Møller-Maersk (2018–2024). Annual reports.

Tradelens

  • IBM and Maersk (2018). Tradelens launch announcement, August 2018.
  • Maersk and IBM (2022). Joint statement on Tradelens discontinuation, November 2022.
  • Lloyd’s List (2022, 2023). Tradelens closure analysis.
  • Gartner (2023). Tradelens postmortem.

Last-mile and parcel logistics

  • United Parcel Service Inc. (2013–2024). ORION programme communications and annual reports.
  • FedEx Corporation (2024). Annual report.
  • Deutsche Post DHL Group (2024). Annual report.
  • Amazon.com Inc. (2024). Annual report.

Asian logistics

  • Port of Singapore Authority (PSA International) (2024). Annual report and Tuas Mega Port communications.
  • Maritime and Port Authority of Singapore (2024). Annual report.
  • Westports Holdings Berhad (2024). Annual report.
  • Northport Malaysia (2024). Annual report.

Agricultural AI

  • Wolfert, S., Ge, L., Verdouw, C., and Bogaardt, M. J. (2017). Big data in smart farming – A review. Agricultural Systems 153: 69–80.
  • US Department of Agriculture (2024). Agricultural Outlook reports.
  • Australian Bureau of Agricultural and Resource Economics and Sciences (2024). Agricultural Commodity Statistics.
  • Climate Corporation (Bayer) (2014–2024). FieldView product communications.

John Deere

  • Deere & Company (2017). Blue River Technology acquisition announcement, September 2017.
  • Deere & Company (2017–2024). Annual reports.
  • US Federal Trade Commission (2021). Right to Repair report.
  • American Farm Bureau Federation (2023). Memorandum of Understanding with Deere & Company, January 2023.

Vertical farming and indoor agriculture

  • AeroFarms Inc. (2014–2024). Public communications and bankruptcy filings.
  • Plenty Inc. (2014–2024). Funding rounds and operational disclosures.
  • Bowery Farming Inc. (2024). Closure announcement, November 2024.

Australian agtech

  • CSIRO (2024). Agriculture and Food research division reports.
  • Queensland Alliance for Agriculture and Food Innovation (UQ) (2024). Annual research summary.
  • AgriDigital Pty Ltd (2024). Public communications.
  • Agworld Pty Ltd (2024). Public communications.

Malaysian palm oil

  • Sime Darby Plantations Berhad (2024). Annual report and AI strategy communications.
  • FGV Holdings Berhad (2024). Annual report.
  • Roundtable on Sustainable Palm Oil (2024). Annual report and certification standards.
  • US Customs and Border Protection (2020, 2022). Withhold Release Order actions on Malaysian palm-oil operators.

Legal AI and Mata v. Avianca

  • Mata v. Avianca, Inc. (2023). Order to Show Cause; Sanctions Order. S.D.N.Y. Case No. 22-cv-1461.
  • American Bar Association (2024). Formal Opinion 512 on AI use by lawyers.
  • Park v. Kim (2024). E.D.N.Y. — sanctions for AI-fabricated citations.
  • Harvey AI Inc. (2022–2025). Funding round announcements.
  • Thomson Reuters (2023). Casetext acquisition announcement, August 2023.

Accounting, audit, and consulting AI

  • PricewaterhouseCoopers (2023). USD 1 billion AI investment announcement.
  • Deloitte (2024). NVIDIA and Anthropic partnership announcements.
  • Public Company Accounting Oversight Board (2024). AI-related guidance for audits.
  • McKinsey & Company (2024). QuantumBlack capabilities and outlook.
  • Boston Consulting Group (2024). Gamma division communications.

Professional-services labour and economics

  • Acemoglu, D. and Restrepo, P. (2020). Robots and jobs: Evidence from US labor markets. Journal of Political Economy 128(6): 2188–2244.
  • Goldman Sachs Global Investment Research (2023). The potentially large effects of artificial intelligence on economic growth. March 2023.
  • Stanford CodeX (2024). AI in Law programme reports.
  • Susskind, R. and Susskind, D. (2015, updated 2022). The Future of the Professions. Oxford University Press.

Australian and Malaysian professional services

  • Law Council of Australia (2024). AI use guidance.
  • Australian Solicitors’ Conduct Rules (current version, 2024).
  • Bar Council Malaysia (2024). Professional conduct standards.
  • Malaysian Institute of Accountants (2024). Annual report and AI guidance.