Bringing a new crop protection product to market is traditionally a slow, painstaking process. Researchers may spend 10 to 15 years discovering, refining, and testing thousands of molecules before a product reaches farmers, but AI is beginning to change that.
Rather than replacing scientists, AI is allowing researchers to predict which molecules are most likely to work long before they’re made or tested, dramatically reducing the number of unsuccessful candidates and helping researchers reach better decisions sooner.
Speaking during a recent webinar hosted by the International Federation of Agricultural Journalists,
Martin Clough, Syngenta Crop Protection’s research and development head of digital collaboration and sustainability, said the agriculture sector was experiencing an unprecedented convergence of technologies.
“It is probably the best time to be working in agriculture, because several technologies are coming together at once. We now have the ability to manage big data, we have generative AI, and we have a much better understanding of how chemistry and biology interact. Together, these have created a step change in innovation over the past five years,” he explained.
One of the biggest opportunities in AI lies in helping researchers identify completely new biological targets. Instead of improving existing chemistries, AI can analyse gene sequences and biological pathways to identify previously unknown ways of controlling pests and diseases, including those that have already developed resistance to current products.
According to Clough, the goal is to develop products that are highly effective against the target organism while remaining safe for people and beneficial organisms.
Traditionally, discovering a new crop protection product follows a sequence of separate steps. Researchers first search for molecules that control a pest effectively. They then refine those molecules to improve safety, reduce environmental impact, lower manufacturing costs, and improve sustainability. Each stage requires evaluating thousands of potential compounds, often taking years.
Generative AI has fundamentally changed this workflow.
“We can design molecules on a computer and optimise them according to all our parameters in silico using predictive models,” Clough explained.
The result is that candidates previously requiring years of laboratory screening can be narrowed down to the most promising within months.
Driving biological crop protection
With the move towards biological crop protection, AI is proving especially helpful, as billions of micro-organisms need to be analysed before product recommendations can be made. AI’s ability to analyse such masses of data means scientists can find solutions far quicker.
Jérôme Cassayre, head of biologicals research at Syngenta Crop Protection, says AI can be applied to integrate microbial genomic data with plant trait information, enabling more precise predictions of microbial functions.
“By using AI-powered models on large datasets, it is possible to understand how microbes influence plant physiology, stress tolerance, and productivity. This approach has the potential to guide and accelerate the discovery of novel microbial solutions for agricultural use,” he explains.
Developing biological products is especially challenging since these living organisms are highly sensitive to their environments. Changes in climate can therefore alter the efficacy of the product.
However, by analysing microbial genes and predicting how they will perform in different environments, better selections can be made for product development.
Cassayre notes that current research includes building data collection platforms that will subsequently be used by AI tools to generate correlations between the chemical composition of various biological active ingredients and the biological response of plants.
This will enable the identification of promising biological traits and guide the development of future applications in crop growth, stress resilience, and nutrient-use efficiency.
Ultimately, Cassayre expects AI and data-driven research and development to revolutionise the use of biological crop protection by enabling both faster innovation and higher-quality products.
“These technologies help us understand and predict how biologicals perform under different environmental conditions, allowing farmers to make more informed decisions about product use. This scientific approach builds trust through consistent results and clear performance data, which in turn drives broader adoption,” he says.
“It is also creating a positive feedback loop: better data lead to improved products, which increases farmer trust, driving further adoption and data collection.”
The field still has the final say
But despite AI’s growing capabilities, computers cannot replace field testing. Clough noted that every promising molecule still needs to be manufactured and evaluated under real farming conditions.
“Those results are then fed back into the predictive models, allowing the AI to learn from each season and continually improve its future recommendations,” he explained.
Syngenta already has more than 500 000 historical field trials available to train these models. Clough added that newer datasets are becoming even more valuable because advances in sensors and digital technologies capture far richer information than traditional assessments.
AI is also changing how researchers evaluate crop performance.
“Subtle improvements in plant vigour or crop health can be difficult to detect visually, particularly when assessing something as nuanced as regenerative farming practices. Hyperspectral imaging allows researchers to detect changes beyond the visible spectrum, providing objective measurements that are more accurate and more consistent than human observation,” said Clough.
“These richer datasets also improve the quality of the AI models, creating a cycle in which better measurements lead to better predictions.”
While AI is accelerating research, Clough cautioned against expecting immediate breakthroughs. The generative design systems currently being developed are still relatively young, meaning commercial products remain several years away.
Even so, he believes AI will substantially reduce the time required to reach research decisions by replacing some experimental screening with highly accurate predictions.
“The aim is to bring better products to farmers faster and more cost-effectively,” he said.
Both Clough and Andre Piza, Syngenta Group’s global head of digital agtech, emphasised that AI is a tool, not a replacement for scientific expertise.
Clough said his team already uses around 30 different AI tools, each suited to different research challenges.
“It’s not that we don’t need scientists; AI simply augments our capabilities,” he noted.
Piza added that farmer confidence is just as important as the technology itself.
“The customer needs to trust what we are doing or they won’t use the tools. Being transparent and respecting farmers’ data is key,” he explained.
He also noted that reducing research costs could make it economically viable to develop products for smaller agricultural markets that have traditionally attracted less investment because of the high cost of product development.
