The global agricultural industry is under immense cost pressures. The war in Ukraine has sent fertilizer prices soaring, while the fallout from covid-19 continues to create pinch points along the cost chain. At the same time there is increasing evidence of the damage caused by climate change and its potential to reduce crop yields.
Food efficiency is also an increasing focus, as is food disposal. According to the UN, as much as 17 percent of total global food production is wasted, often dumped into landfill, further contributing to climate change.
Against this background, the cost savings generated from artificial intelligence are a natural fit. Agriculture is probably the least digitalized sector and there is a lot to be gained by using better data. Even weather forecasting, a long-established tool for farmers, is infinitely better with machine learning and enhanced data analysis.
Making the difference
Rob Appleby, CEO at sustainable food and agriculture firm Cibus Capital, sees significant potential for “AI-powered sensors to collect and aggregate enormous data banks.” By using that data, farmers can make informed, timely decisions across planning, fertilizing, irrigating and spraying to increase yields while reducing inputs and risk.
Contained environment farming – such as glasshouses, vertical farms and polytunnels – will also benefit from crop monitoring, identifying early diseases or nutrient deficiencies, and allowing predictive and preventative measures to be selectively applied. Livestock farming could similarly benefit, with AI providing improved productivity, reduced mortality, reduced antibiotic use and a decline in carbon emissions.
Additionally, Appleby sees AI’s potential to improve efficiencies and reduce waste in supply chain management, accelerate plant breeding processes in genomic selection, fast-track discoveries in metagenomic data to advance human health, agriculture and biotechnology, and both aid and accelerate the development of new alternative proteins, from plant-based to cellular meat.
AI-enabled autonomous tractors and harvesters will also develop to perform a wide array of tasks on farms, from planting, ploughing, topping and harvesting to input applications, weeding and picking. This will increase productivity while driving down labor costs, labor risk and human errors.
“Overall, AI is fundamental to the future of agriculture,” says Fabio Sofia, co-founder and managing partner of Swiss venture capital firm Zebra Impact Ventures. And the concerns that some have about AI in other sectors are not appropriate when it comes to agriculture. It is so far behind other commercial sectors that that “would really be some way off,” says Sofia.
“We’re only scratching the surface of AI’s potential in agriculture,” says Mark Kahn, managing partner at venture capital firm Omnivore. Promising applications are emerging in precision farming for data-driven crop management and predictive analytics to guide farm decisions, as well as post-harvest storage.
Examples of innovation
Across supply chains, AI could revolutionize food processing safety and quality control while streamlining fragmented logistics via demand forecasting, inventory recommendations and enhanced end-to-end tracking across complex networks.
Some examples from Omnivore’s portfolio include Niqo Robotics, an AI-powered robotics start-up building robots for smallholder farmers, and BharatAgri, a firm that provides AI-supported farm advisory services. Pixxel, a space data company building a constellation of the world’s highest-resolution hyperspectral earth imaging satellites, uses analytical tools to mine insights from data.
Appleby also sees a “huge opportunity for AI in robotics, automation and precision agriculture.” Cibus’s late-stage venture strategy has invested in several companies such as Ecorobotix, which has ultra-high precision smart spraying systems and uses computer vision in combination with robotic hardware to deliver treatments at a plant-by-plant level. This reduces herbicide and pesticide application by up to 95 percent while boosting crop yield.
Another Cibus investment, BeeHero, employs simple, inexpensive sensors coupled with advanced data analytics and AI to monitor beehive health and pollination strength. Hives on BeeHero’s platform are on average 75 percent stronger and mortality rates are reduced by around 50 precent.
Finally, Agsenze provides automated identification of health and production outcomes for livestock using 2D and 3D computer vision technology. This has what Appleby describes as a “clear and proven” customer return on investment of three times via reduced labor costs, improved production metrics, reduced antibiotic use and improved fertility rates.
Sofia adds that “there are particular vital applications in the area of agricultural biotechnology where enhanced machine learning enables rapid advances.” For example, AG-Biotech works with a small protein or molecule similar to the RNA technology used in the covid-19 vaccine. It can use targeted messenger RNA to influence a whole crop. The number of combinations that can be tested go to well over a billion using AI, as it goes through a cycle of test and learn, then test and learn, over and over again at rapid speed.
Trials and tribulations
However, there are challenges. Agriculture has historically been slow to embrace technology and there are some who will still need convincing. Existing manufacturers will need to adapt and invest fast to keep up with the development of autonomous AI-driven machinery.
There will also be companies and growers that are slow to react, and late adopters will likely struggle to compete, or lack the capital to invest.
“Undoubtably, the acceleration and accessibility of AI will change the world, probably irreversibly”
Appleby says: “Thought must be given to protecting and helping small-scale farmers that will not necessarily have the capital or ability to integrate the latest AI within their production systems.” Otherwise, there will be a significant widening of the gap between big and small farmers. And special support will need to be directed towards the developing world.
Omnivore’s Kahn highlights another issue. He fears that “existing agricultural data ecosystems are highly fragmented and varied, with little data interoperability across sources, environments and formats.” Cleaning and labeling massive amounts of diverse data for AI applications is difficult, and the lack of structured data holds back analytics and discovery potential.
Furthermore, most AI and machine learning models for agriculture are still in research or at the ideation stage. The agriculture sector still lacks domain expertise, such as joint experts across AI development, agriculture use cases and business operations.
And while the availability of low-skilled labor is becoming more problematic across the developed world, there will be a need for training and re-skilling workforces to benefit from the new opportunities.
Looking to the future
As AI advances and becomes more accessible, “the key for existing agricultural companies will be to integrate AI-driven solutions within their business models,” explains Appleby.
Kahn believes “as data collection and management systems improve, larger, higher-quality data sets will fuel the development of advanced machine learning models from predictive analytics to automated farm machinery.”
Commoditization of sensors, the Internet of Things and satellite imagery will make precision agriculture affordable to more smallholder farms. In the long term, human-AI collaboration will help the agriculture sector enhance productivity, minimize ecological impact and transform livelihoods.
“Undoubtably, the acceleration and accessibility of AI will change the world, probably irreversibly,” says Appleby.