Unlocking Chemistry with Code? A Look at AI in Crop Protection Discovery

Index

1. Overview

2. Why Ai

   a. Speed

   b. Cost

   c. Outcomes

3. Examples by Company

   i. Agrematch

   ii. Enko

   iii. Bayer Crop Science

   IV. Syngenta

   V. Corteva / AgPlenus

   VI. UPL

4. Challenges

5. Final Thoughts

6. Additional Resources and Context

Overview

In 2024, the major Crop Protection and Seed companies invested over $7 billion in R&D:

However, not all of those dollars are being allocated to Crop Protection. Referencing old data from Phillips McDougall, a higher percentage of spend goes to seed and trait R&D:

It still indicates that several billion dollars are invested in R&D by seven of the largest companies into crop protection each year— and it has been around that level for several years.

Still, there has been a downward trend in new molecule discovery after peaking in the 1980s:

You can find other recent molecule discovery’s here.

With Agbioinvestor showing that discovery now taking over 12 years and more than $300 million from start to commercialization.

This is not a new phenomenon. It's an agricultural parallel to what’s been long observed in pharmaceuticals: Eroom’s Law¹ . As technology in other domains has accelerated (eg: Moore’s Law in computing), drug and crop protection discovery have moved in the opposite direction— becoming slower, riskier, and more expensive.

The slow down is happening while herbicide resistance is growing rapidly around the world and pest outbreaks increase in severity. :

This is all to say, innovation in crop protection is more necessary than ever, yet harder to achieve than ever. With more stringent regulatory environments, and demand for environmental friendliness, the conventional discovery pipeline for herbicides, fungicides, and insecticides is struggling to meet the needs of farmers, and crop protection companies.

That’s where the artificial intelligence is hailed as a saviour. It is being talked about as a catalyst to reverse this stagnation.

The promise? To cut R&D timelines and efficiency, improve hit rates, and guide the design of more sustainable, safer, and effective crop protection products.

But is that promise being delivered on today?

Why AI?

Crop protection companies are increasingly using artificial intelligence within their discovery process. AI helps screen billions of molecules in silico to predict their efficacy, toxicity, and environmental behavior. Companies use machine learning models, generative design, and retrosynthetic analysis to identify promising candidates and optimize their synthesis² . AI offers a way to shift the R&D equation by unlocking three interlinked levers of upside: faster timelines, lower costs, and better outcomes.

a. Faster Discovery

Historically, discovery in crop protection has been about mass amounts of lab work— screening tens of thousands of compounds over years, hoping a few turn out to be viable. AI changes this by shifting much of the discovery process from the lab to the cloud.

Virtual screening² and in silico modeling can evaluate billions of potential molecules within weeks compared to hundreds physically tested over months. Protein design and target-based discovery go a step further where AI can identify specific vulnerabilities in pests (e.g., enzymes critical to survival) and then reverse-engineer molecules to block those targets, going through mass amounts of molecules out there in more rapid fashion.

b. Lower Cost

The average cost to bring a new active ingredient to market now exceeds $300 million. Much of that is spent on failed candidates— compounds that looked promising in vitro but fell apart in field trials or regulatory review. AI reduces these costs not by cutting corners, but by improving decision-making upstream in the process.

By predicting which molecules are most likely to succeed based on biological activity, toxicity, environmental persistence, and synthesis complexity, AI tools can help R&D teams focus on high-probability candidates earlier. This avoids wasting millions and increases optionality for R&D spend.

c. Better Outcomes

Perhaps the most important benefit AI can unlock is improved R&D success rates, in combination with the above benefits. Traditional discovery still relies heavily on analogs of known chemistry.

AI expands the potential.

It learns from datasets including past successes and failures to uncover novel chemical scaffolds and new modes of action and can parse immense amounts of molecules.

Brian Lutz, VP of Agricultural Solutions at Corteva, recently shared with US Congress just how vast the chemical universe is— around 10⁶⁰ possible molecules, far exceeding the number of grains of sand on Earth (10¹⁸ ). Traditional screening methods can only cover a fraction of this space. AI is changing discovery by enabling targeted searches for molecules that work on pest-specific proteins, aiming for high specificity, environmental safety, and human safety.

Instead of relying on randomness, AI is said to bring speed, precision, and predictive design. In one recent example, Lutz shared that AI modeled 10,000 molecules in weeks, uncovering dozens of promising new crop protection candidates that traditional methods would have missed— many of which are now in testing.

AI also supports better multifactor optimization. In one model, a molecule isn’t just evaluated for efficacy, it’s evaluated on things like cost of goods, environmental safety, formulation compatibility, and regulatory fit.

Examples by Company

One challenge I have had is getting agribusinesses to share specific outcomes and implications surrounding AI use in discovery. Because of that, I wanted to research and find examples to try and convey the potential benefits in more quantifiable terms.

Startups

There are a host of start-ups, that primarily partner with large incumbents, leveraging AI to improve discovery:

Subscribe to Upstream Ag Professional to read the rest.