The 5 most important AI trends in agriculture

What are the technological keys for unlocking the potential of sustainability at a massive scale?

We all know that agriculture is one of the oldest cultural practices in human history. And we know that the technological limitations and possibilities of an era have always been mirrored in that practice. The rise and fall of many civilizations has been closely linked to it. That still holds true in 2020. The continued viability of our civilization depends on our ability to drastically alter the way we produce our food, the way we do agriculture. It is very simple ecological arithmetic to show that it is impossible to continue feeding today’s diet produced with today’s methods to a growing world population in a changing climate. This equation is completely out of whack. But tech might help. While not at the forefront of the general imagination, in terms of quantitative impact, agriculture will probably be the most important field of application for the technological promises of the next decade. And the term at the top of the list of technology trends that enthuse and concern us today is undoubtedly AI. I got excited about the question of what this megatrend means for agriculture and how some of it’s applications could massively accelerate the way towards a future of sustainable farming. During my research about this topic, I came across a bunch of hot companies that are actually translating the potential of AI in agriculture into scaleble solutions. Here are the five top trends that these companies are working on:

1. The precondition: Big Data

First I realized that at the beginning of using AI down on the farm is the transformation of farms into data factories. The basis for all AI applications is abundant data on all aspects of whatever the smart algorithms are used for — be it for optimizing matches on a dating platform or for improving the cultivation of plants. And since we are dealing with living ecosystems in the case of plant cultivation, that is quite a bit of data: weather, temperature, soil health, moisture and water consumption, plant condition, pesticide and fertilizer concentrations and most likely many more. Sensors in the soil, in and around plants and waterways, drones and the use of satellite imagery will be ubiquitous on farms to collect these data permanently and live. (The startup Beeodiversity even generates data about the environment’s health from automated analysis of bee pollen!). These data will then allow better, more precise and automated decisions on the whole cycle of plant cultivation, from choosing the right sowing options and rotations, to the most efficient use of resources and designing the least invasive chemical interventions.

2. The holy grail: Climate and weather forecasts

Second, it stood out for me that short and long-term weather trends are the strongest influence on agriculture, while at the same time volatile and extremely hard to predict. This is particularly true in the context of climate change. However, the right fit between weather patterns and cropping strategies can create dramatic differences in yield efficiency. So if we seriously want to use technology to generate better decisions on the management strategy of a piece of land, the cracking of weather patterns in any given micro geography is the holy grail and a particularly promising application for AI. A number of start-ups are pursuing precisely that goal. They want to help in enabling better decisions where weather is as a major factor. As a lighthouse example of an AI-based weather platform, I am fascinated by the UK start-up Cervest. It has trained its algorithms with millions of satellite images and probability theory to reduce the difficulty for decision makers at all levels of in dealing with the uncertainty of climate risk. Another company to watch in this emerging field of predicting climate as a service is US-based start-up Jupiter. The multi-facetted repercussions of changing weather patterns across so many aspects of life will make climate services a thriving field with many spillover effects across various industries.

3. The foundation: Plant Health

We regularly come across these mind-boggling statistics describing the losses and inefficiencies within the food system. We read them and can hardly believe our eyes. Better forget them quickly, right? Here is another one: About 30 percent of the world’s harvest is destroyed each year due to pests and disease. This gigantic loss triggers the massive use of pesticides, which in turn drives many species into extinction, destabilizes entire ecosystems and causes huge health crises. However, truly fascinating possibilities for dramatically reducing this vicious circle of crop failure and chemical use arise from the emergence of large platforms for aggregating the analysis of images of diseased plants. How does it work? Farmers upload images of diseased plants to the platform. Machine Vision recognizes the patterns of disease at an early stage and recommends specific and minimal interventions — instead of using broad-spectrum pesticides and all-round fertilizers. The more image material collected, the more precisely the algorithms can be trained. For example, the German start-up company Plantix already has 15 million analyzed images on its platform and 50,000 uploads are recorded daily. The 1 million users of the app are also connected to a network of scientists and experts who represent a real know-how community in a fragmented industry. This is particularly true in places like India where small scale farming is widespread (also check out wefarm, which is doing similar things). Solutions such as Plantix are a massive help in resolving the classic decision between sustainability and (short-term) yield optimization step by step. Competitors like Taranis are equally inspiring, using a slightly different approach based on different types of aerial images. The evolutionary next step will be the execution of the recommendations for action by Robots such as the Dino of Naio Technologies or the robots of Rootwave, which are commiting electricide on wheets instead of drowning them in pesticides. These and many similar developments will be key ingredients for a future agriculture depending less and less on broadly applied chemicals while at the same time increasing yields.

4. Bringing it all together: Cultivation decisions

Reading about all these and other companies with their specialized analysis of specific aspects of farm operations, I expected to find solutions that are trying to bring it all together to create one central “cockpit” for the farmer. And I found them: Companies like the celebrated Brazilian start-up Agrosmart, Kisanhub from the UK, Croatian Agrivi or Onesoil from Belarus are building solutions for the holistic use of all data points on a farm for the biggest of all decisions: which plants should be grown when and on which soil? To do this, all data on the condition of the soil, the macro- and microclimate as well as the performance, the disease and pest infestation of surrounding areas and plants must be considered and weighed in an integrated way — a process that today mostly still follows the experience and gut feeling of the farmer. This gut feeling will gradually be replaced or supplemented by agronomic models that use AI-trained algorithms to generate recommendations from the raw data for cultivation strategies or even for instruments like insurance products corresponding to these strategies’ specific risks. These solutions also have the advantage of comprehensive documentation and transparency of the cultivation process, which allows even large companies to monitor quality standards in complex supply chains and thus to communicate more credibly to the end consumer and regulatory authorities. I expect a few of these platforms to become very powerful in tightening up the inefficient and fragmented food industry by storing the relevant data and setting standards. Just as Google today manages vast amounts of data about the behavior of billions of individual and makes diverse use of it, these platforms will know about the state of farms and crops around the world and put these data to wide-ranging uses. Let’s hope these future agridata kingdoms will find ways to combine profit AND the creation of solutions for fixing the food system.

5. Managing animals

Last but not least, agriculture as we know it today is not only about plants, but also about constantly creating and structuring the lives of dozens of billions of animals. Hence herd management is another rapidly growing application of AI. The goal of this type of software is also to generate concrete forward-looking analyses and action-oriented recommendations by means of data analysis and machine learning. I got the impression that the company Connecterra from Holland is one of the innovation leaders in this field. It has started with a kind of “fitbit for cows” solution. The machine learning algorithms of the company can now be trained with data sets containing several million behavioral years of cows as well as the accompanying input variables (feed, medication, condition of the stables, …). The intelligent knowledge about herd management is derived from this constantly growing data goldmine. This knowledge is used to improve animal health, fertility and productivity at maximum efficiency and conservation of resources. The aim is to create a single platform for monitoring the lives of the animals, which constantly regulates the relationship between the animals’ life data and the input variables. Users’ reports of increases in efficiency of about 30% with the same use of resources as well as dramatically reduced veterinary interventions are giving us very realistic hope that the world’s hunger for animal protein can be met at a significantly lower costs for the planet and the animals. Companies like Canadian Wittaya or Norwegian Aquabyte are doing similar things for fish cultivation in aquaculture. If these companies’ vision’s succeed we’ll soon have automated machines taking the life decisions for up to 60 % of all mammals and 70 % of all birds worldwide (these two numbers are the percentages of breeding animals of global biomass in these two groups of animals, according to a recent study). This is quite something. Machines will then have control over a significant proportion of life on this planet. I don’t really know what my emotional response to this actually is, but it certainly is very very remarkable. And it certainly offers potent ways to reduce the heavy costs the animal industry is substracting from the planet today.

Having read dozens of articles and spoken to a handful of AI experts about the topic of AI in agriculture, I am more convinced and optimistic than ever that this technological potential will be a central factor in our fight to restore planetary healt or at least in preventing it’s further degeneration. Hopefully there will be enough kick-ass entrepreneurs willing to make this potential flourish as opposed to wasting their energy building the next meaningless attention-spying-click-based business model.

PS: Many of the trends here will be strongly used in an integrated way by vertical farming solutions. I will soon write a separate article about that megatrend.