Despite increasing scrutiny of MIT’s infamous claim that 95% of enterprise generative AI pilots fail, there is anecdotal evidence that many large companies are still struggling to successfully implement AI across their businesses.
There are many possible reasons for this, from the wrong company culture to inefficient personnel to data hygiene and infrastructure. One anonymous COO captures the prevailing sentiment among MIT researchers: “The hype on LinkedIn says everything has changed, but nothing fundamentally has changed in our operations.”
In contrast, Bayer Crop Science appears to be implementing GenAI tools at scale, resulting in measurable operational improvements. Bayer’s ELY system, which gives agronomists access to product knowledge, has increased productivity by 60% and is used by more than 1,500 front-line employees across North America, according to Amanda McClellen, Bayer’s chief information officer.
The difference may lie in an acquisition made 12 years ago that most competitors don’t have.
12 years of build
The foundation for Bayer’s current AI capabilities was laid in 2013, when Monsanto acquired the Climate Corporation for $930 million. McClellen says the deal brought not only the FieldView precision agriculture platform, but a data culture, technical talent and, importantly, lessons in digital product development that shaped everything that followed. AFN.
Fast forward to today: Bayer’s leading GenAI tool ELY recently won “AI-based AgTech Solution of the Year” at the AgTech Breakthrough Awards.
“One of the reasons we chose to develop [E.L.Y.] Because we have our own data, and we have our own insights into that data…[such as] Mr. McClellen, who began his career as a biochemist at Monsanto and spent nearly eight years in biotechnology before moving into breeding and eventually IT, is a wealth of test data from the company’s research and development, product supply and commercial settings.
McClellen said the Climate acquisition provided important lessons, including understanding “how different it is to bring a digital product to market as opposed to a physical product” and figuring out “the value proposition between the interface of those two things.” The company has spent years building its data infrastructure, amassing field test data, creating semantic tools to make that data discoverable, and establishing what McClellan describes as a mature data warehouse platform.
moat
The data moat is substantial, with 117 billion data points on seed performance.
“We have one of the largest and most complete datasets in the industry. We have decades of field test data on both products brought to market and failed products in the pipeline, as well as genetic information about those products, allowing us to explore and understand the relationships between what combinations of genes are most successful in what environments.”
But this foundation was built first for traditional AI. The company has been using machine learning and deep learning in research and development “for a long time,” McClellen said, long before the GenAI hype cycle began.
The results are visible. AI technology has accelerated crop breeding cycles and “reduced overall product delivery time by two years.” In an industry where product development traditionally takes seven to 10 years, this is a significant competitive advantage.
This proven track record has helped create what McClellan describes as an “industry-leading pipeline” worth $32 billion with approximately $2.4 billion in annual research and development investment.
A key innovation is the digital twin project. “A literal digital twin of our field test network…a replica of millions of acres of potential agricultural land,” McClellen explains. “By leveraging this high-fidelity twin, we can simulate the performance of what is going through the pipeline.”
The value here is speed and predictive ability under untested conditions.
“People depend on the weather. In real life, it depends on, ‘Was it rainy in July or was it cool in July?'” she points out. “In any given year, we can only test what the weather gives us. And with this digital twin, we can really begin to understand how our products perform across environments that we may not have experienced yet.”
GenAI: Test and learn at scale
ELY was launched as what McClellan calls a “test and learn opportunity” to explore both business utility and technology strategy. The company has undergone rigorous validation, with over 1,500 agronomists testing it for nearly a year to ensure it meets customer needs.
When asked how to balance “test and learn” with urgency, McClellen emphasizes repetition. “AI is accelerating at a very fast pace, both in its capabilities and adoption, so we have adopted an iterative methodology that allows us to experiment with new technologies while continuously gathering data and insights to inform next steps.”
The system aggregates what McClellen describes as “all the agronomic knowledge, all the product recommendation sheets” – contextual information about how to most effectively use Bayer products. “We were able to develop this tool, this product, and today we’re bringing it to North America. So field-facing agronomists are now about 60 percent more productive. They save about four hours a week because they don’t have to spend searching for all this knowledge.”
These time savings directly translate into increased customer engagement.
McClellen will outline three key pillars for broader AI adoption: sales and service (where ELY is based), supply chain and logistics, and research and development. “The future is about multiple agents working together,” she envisions. She believes GenAI could be integrated with FieldView and eventually provide advice directly to growers. She admits, “It’s not something we’ve started yet.”
Actual data moat
The decade of infrastructure construction is manifested in specific product applications. McClellen points to PRECEON, Bayer’s short stature corn product, as an example of how digital and physical products intersect.
“For farmers to get the most out of that innovation and get the most productive output on the farm, they have to combine it with the right hybrid selection, and they have to combine that with the right density, the planting density of that hybrid on the farm,” she explains. “That would not be possible without a platform like FieldView that helps us understand farm acreage and make accurate recommendations.”
This integration of unique genetic resources, decades of performance data, and digital tools represents a moat that is difficult to replicate. As McClellen points out, “agriculture is deeply specific to farming and farming,” unlike more commoditizable AI applications like customer service, which “cut across many different types of industries.”
McClellen, who grew up on a farm herself and whose father is a farmer, brings personal background to the digital challenge. “I think we’re seeing a new generation of farmers who were raised differently and have a different mindset. [approach to] The complexity of decision making and the different types of data that need to be integrated to make good decisions. ” she observes. “Digital is the obvious choice for managing all of these on-farm decisions throughout the season.”
Bayer’s approach is consistent with what MIT researchers have found to be what differentiates 5% of successful AI deployments. “They identify one pain point, execute well, and collaborate smartly.” The 1,500-person ELY pilot, emphasis on iterative methodologies, and focus on specific use cases all reflect a disciplined strategy.
But the real differentiator may be much simpler. Bayer had a 12-year head start.
“Rethinking work”
The acquisition of Climate brought more than just technology. Data cultures also take years to build and are blamed for slowing enterprise AI deployment more broadly..
Perhaps the most revealing insight comes when McClellen discusses change management for agent AI. “We have to be prepared to work differently, not just having agents do tasks that people used to do, or having people do different types of tasks, but also rethinking how work works,” she says. “If a digital agent can do something that previously could only be done with a person or a team of people, then the entire business process probably needs to look different.”
This suggests that companies are looking beyond simple automation to fundamentally redesigning business processes. Although this is ambitious, it is still largely a hypothesis. “This is something we are still in the early stages of and a learning journey, but it is something we are paying very close attention to,” she admits.
Asked about quantifying ROI, McClellen believes that quantifying ROI is necessarily complex. “Quantifying ROI is a multifaceted process, and the ultimate goal is to create AI solutions that not only improve financial performance, but also contribute to sustainable agricultural practices.”
This response reflects companies balancing short-term productivity gains with long-term strategic positioning around sustainability, but whether it represents sophisticated thinking or avoids more difficult financial issues remains to be seen.
More about AI in agribusiness:
Are agri-food companies going far enough with their AI initiatives?
Corteva’s planned separation raises questions about AI and data partitioning
Corteva: AI can transform crop protection, replacing ‘randomness and chance’ with ‘prediction, specificity and design’
Where are we in the AI bubble?
