America’s Cultivation Corridor continued its Insights to Innovation webinar series with a wide-ranging discussion on how artificial intelligence (AI) is accelerating innovation across agriculture, livestock, machinery and research systems. Moderated by Iowa Secretary of Agriculture, Mike Naig, Iowa Department of Agriculture and Land Stewardship, the session brought together leaders shaping the technology’s real-world deployment.
The panel featured Melissa Neuendorf, Principal AI Strategist, Deere & Company; Matt Smalley, Data Science Leader, Corteva Agriscience; and Joseph Victoria, Sr Associate Director, Bioinformatics, Boehringer Ingelheim Animal Health.
In The Future Is Now session, leaders discussed how artificial intelligence is already being applied across agricultural research, livestock, machinery and production systems. They emphasized that AI enhances human expertise, solves real-world challenges, and depends on cross-sector collaboration to move from research into practical use.
The panel discussed future needs for workforce training, data infrastructure and supportive policy to accelerate responsible AI deployment in the industry.
Watch the full session and stay tuned for more Insights to Innovation webinars exploring the people, ideas and technologies driving progress in agriculture and bioscience.
Webinar recording
Post-webinar Questions & Answers
Questions posed by attendees that were unable to be answered during the live event.
Question: How do we balance the right of farmers to own their own data and having enough scale for companies to make enough money to make the AI products to gain insights?
Answer (Matt Smalley): Balancing farmer data ownership with the scale needed for AI is foundational. From our perspective, farmers should retain control over their data and have clarity on how it is used. At the same time, AI only creates value when models are trained across diverse environments, management practices, and seasons.
The way we approach this is through transparency and value exchange. Farmers participate when they see clear benefit, whether that is better recommendations, reduced risk, or improved outcomes. Our responsibility is to ensure data is used responsibly, anonymized where appropriate, and applied in ways that directly improve on‑farm decisions. When trust and value are clear, scale follows naturally.
Question: What can you tell to some people that sees AI , at lest now, as just like a high-fashion runway shows that never make it to everyday wardrobe?
Answer (Joe Victoria): With any burgeoning technology there is certainly some truth to the statement that not every AI endeavor will bear fruit. We have direct experience with cases in which models look to be groundbreaking when applied retrospectively to a specific study or process – often these can be the product of over-fitting. When the AI solutions are applied to more broad circumstances or situations, they quickly breakdown and fail. However, there are real world examples where AI has created methods which enhanced communication/collaboration, speed up process allowing solutions to the customer faster, and narrowed focus on target therapies accelerating discovery. As with any new technology – we build the foundation for future iterations to be less error prone, more reliable/faster/better. Your question of ‘everyday wardrobe’ is valid observation as many of the advances, at least in animal health, aren’t or won’t be completely obvious to the customer that AI played a role in development of the new products.
Answer (Matt Smalley): That analogy resonates. AI has no value if it stays in demos or slide decks. For us, the test is simple: does this actually change a decision someone makes in the field or the lab?
Most impactful AI work is not flashy. It is embedded into workflows, improves consistency, and helps people make better decisions faster. We spend a lot of time making sure AI tools are interpretable, reliable, and designed around real user needs. When AI becomes invisible and simply improves outcomes, it has moved from the runway into everyday use.
Question: We see massive investments in AI capabilities. How much of what you each need to do can be done only internally vs through collaborations with or acquisitions of start-up innovation companies.
Answer (Joe Victoria): As you are aware, the four main paths for discovery involve some combination of:
(1) Internal development (2) Collaboration (3) Fee for service (4) Acquisition
As you correctly state – there are massive investments into AI, because of that high demand both “fee for service” and “acquisition” options are often priced out of the animal health market. It is also an incredibly rapidly changing technology and keeping up-to-date is challenging when that is not a company’s sole focus. Because of this, in my opinion, the most practical solutions are a combination of internal development and collaboration – with the possibility of acquisition, if there is a unique opportunity where an external company has the correct blend of: Quality and know-how, animal health specific experience, and price.
Answer (Matt Smalley): It is not an either‑or decision. Some capabilities need to be built internally because they are deeply tied to proprietary data, crop biology, or core decision systems. At the same time, innovation in AI is moving too fast for any one organization to do everything alone.
We actively collaborate with startups, academic groups, and technology partners to accelerate learning and access new approaches. The key is knowing where integration and scale matter most and where partnering allows us to move faster without losing focus. The most effective AI strategies combine strong internal foundations with targeted external collaboration.
Question: In your experience, when forecast models become statistically more accurate but planning teams still rely on overrides, what signals do you look for to determine whether the gap is a trust issue versus a model interpretability issue?
Answer (Matt Smalley): Overrides are not automatically a failure. They are a signal. The question is what kind of signal.
If overrides are consistent and directional, it may indicate the model is missing important contextual factors. If overrides vary widely by user or region, it may be a trust or interpretability issue. We look closely at whether users understand why a model is making a recommendation and whether they can relate it to agronomic logic they already trust.
The goal is not to eliminate human judgment but to improve alignment between human expertise and model insight. When those two reinforce each other, adoption follows.
Question: From a productization standpoint, how does Corteva manage the trade-off between building globally generalizable AI models and maintaining biological relevance at the local farm level when scaling decision-support tools?
Answer (Matt Smalley): This is one of the central challenges in agricultural AI. Biological systems are inherently local, but AI benefits from global learning.
Our approach is to combine global models with local calibration. Global models learn broad patterns across crops, environments, and seasons, while local data and expertise ensure recommendations remain biologically relevant. This hybrid approach allows us to scale insights without losing the agronomic nuance that actually drives outcomes on individual farms.