John Deere – Bringing AI to Agriculture

John Deere, known for its iconic green tractors, has been making big strides into AI-driven precision agriculture.

Introduction: John Deere

John Deere is a giant in the agriculture industry, with nearly $30 million in annual revenues and a $48 billion market capitalization6. While it was  founded in 1837 as a tool manufacturer, today the company focuses primarily on manufacturing heavy equipment for the agriculture industry. The company’s iconic green tractors are one of the best-recognized machinery brands in America.

John Deere has long prided itself for its technology-forward mindset. The company’s main technology focus is on what it calls “precision agriculture,” which incorporates technology into farming processes to increase productivity and yield8. John Deere began incorporating aspects of precision agriculture in the nineties with self-driving technology in tractors, and today’s farm equipment is as high-tech as the average car.

John Deere’s ML Strategy

Machine learning has huge potential to revolutionize the way we farm. Computer vision and machine learning technology can be used in every step of farming: tilling soil, planting seeds in the optimal locations, spraying fertilizer or nutrients, and harvesting5. Machines that harvest corn, for example, typically drop a percentage of the corn on the ground. A machine with blades that could dynamically adjust to the width of a corn stalk could increase yield simply by catching more of the harvest8. Other areas with the biggest potential are crop and soil monitoring, predictive analytics, and agricultural robots.

In the last few years, John Deere has made several big plays in agricultural AI. John May, President of Agricultural Solutions and Chief Information Officer at Deere, has said, “As a leader in precision agriculture, John Deere recognizes the importance of technology to our customers. Machine learning is an important capability for Deere’s future.”In September 2017, the company acquired Blue River Technologies for $305 million, a move that attracted a lot of attention in the agriculture industry2. Blue River Technologies is a Bay Area startup that uses computer vision and machine learning to reduce the use of herbicides by spraying only where weeds are present, optimizing the use of inputs in farming – a key objective of precision agriculture. John Stone, the senior vice president of the Intelligent Solutions Group said at the time of acquisition, “AI and machine learning is going to be as core to John Deere as an engine and transmission is.”2

The Blue River acquisition wasn’t John Deere’s only big move of 2017: they opened John Deere Labs in the trendy SOMA district of San Francisco, with the plan to centralize their software development there4. The move reflects the difficulty companies have attracting top tech talent to traditional agriculture cities like Des Moines. With the new lab, the company is hoping to cement its reputation as one of the leaders in agtech, particularly AI and machine learning.

Recommendations

While the company has made some big moves, there’s still a lot of ground to cover to truly integrate ML into the company. One of John Deere’s highest priorities needs to be focusing on the development of Blue River’s see-and-spray technology. The acquisition has a lot of promise, but unless John Deere can help Blue River deploy their solution at scale, they’ll lose out on the AI revolution. The first step is to deploy the product for cotton farmers, and then to expand to soybean farmers soon after.

In the longer term, John Deere should take advantage of the advances in edge computing hardware to move some of its data processing from the cloud to the edge. Right now, John Deere tractors come equipped with a 4G LTE modem to upload and process data3. If the company plans to employ more computation-heavy algorithms at scale, the upload time will become cost-prohibitive, especially in rural areas with poor service. If the company can host some of the data pre-processing at the edge, possibly through an installed GPU on the tractor, it can deploy more algorithms at scale.

The company should also incorporate more machine learning capabilities into its Farmsight suite of products. Right now, Farmsight focuses on data processing and visualization, with some rudimentary decision analysis for farmers. But the amount of data the tool collects, as well as its wide install base, would make it the perfect system to test-drive ML capabilities.

Finally, the company could make a push into the predictive maintenance space. Machine learning is a perfect tool for predictive maintenance, and John Deere’s experience and expertise with large assets gives it the necessary capabilities to succeed where so many companies have failed.

Open Question

John Deere has recently attracted controversy over its refusal to allow farmers the “right to repair” their own equipment, instead charging high fees to use proprietary diagnostic tools9. How can John Deere ensure that farmers can operate and repair their equipment as it becomes more technologically complex?

Word count: 788

References

  1. “AI and Machine Learning Innovations Taking John Deere to the next Level.” IoT Tech Expo (blog), March 28, 2018. https://www.iottechexpo.com/2018/03/ai/ai-machine-learning-innovations-john-deere/.
  2. “Deere Acquisition of Blue River Technology | John Deere US.” Accessed November 7, 2018. https://www.deere.com/en/our-company/news-and-announcements/news-releases/2017/corporate/2017sep06-blue-river-technology/.
  3. “John Deere Bets the Farm on AI, IoT.” Light Reading. Accessed November 7, 2018. https://www.lightreading.com/enterprise-cloud/machine-learning-and-ai/john-deere-bets-the-farm-on-ai-iot/a/d-id/741284.
  4. “John Deere Drives into San Francisco Looking for Farm-Tech Engineers – SFChronicle.com,” September 29, 2017. https://www.sfchronicle.com/business/article/John-Deere-drives-into-SF-looking-for-farm-tech-12239025.php.
  5. Liakos, Konstantinos G., Patrizia Busato, Dimitrios Moshou, Simon Pearson, and Dionysis Bochtis. “Machine Learning in Agriculture: A Review.” Sensors (Basel, Switzerland) 18, no. 8 (August 14, 2018). https://doi.org/10.3390/s18082674.
  6. Marr, Bernard. “The Incredible Ways John Deere Is Using Artificial Intelligence To Transform Farming.” Forbes. Accessed November 7, 2018. https://www.forbes.com/sites/bernardmarr/2018/03/09/the-incredible-ways-john-deere-is-using-artificial-intelligence-to-transform-farming/.
  7. Peters, Adele, and Adele Peters. “How John Deere’s New AI Lab Is Designing Farm Equipment For A More Sustainable Future.” Fast Company, September 11, 2017. https://www.fastcompany.com/40464024/how-john-deeres-new-ai-lab-is-designing-farm-equipment-for-more-sustainable-future.
  8. Sennaar, Kumba. “AI in Agriculture – Present Applications and Impact -.” TechEmergence, October 16, 2017. https://www.techemergence.com/ai-agriculture-present-applications-impact/.
  9. Solon, Olivia. “A Right to Repair: Why Nebraska Farmers Are Taking on John Deere and Apple.” The Guardian, March 6, 2017, sec. Environment. https://www.theguardian.com/environment/2017/mar/06/nebraska-farmers-right-to-repair-john-deere-apple.

 

Previous:

Machine Learning at Disney: Solving Happiness

Next:

Could Additive Manufacturing (3D-Printing) Be the Answer to Boeing’s Program Accounting Woes?

5 thoughts on “John Deere – Bringing AI to Agriculture

  1. I don’t buy the JD argument presented in reference 9 about why they don’t want farmers to have the ability to do repairs. For anything software related, they could use the Tesla model of having everything linked remotely, and would be able to do any diagnosis, upgrades, repairs quickly and cheaply. For anything else mechanical or electrical there is no more risk than with a conventional piece of equipment. To me this seems like a play for more revenue by adding on hidden fees in the form of service which cannot be gotten elsewhere.
    It will also be interesting to see how legislation changes this model and forces JD and others to stop this monopoly-like practice. MA has had a law on the books since 2012 aimed at right to repair in the automotive industry, and many more states are following suit (and some countries already have, like Australia).

  2. Kate, thank you for sharing this topic, as John Deere does not typically get due recognition for its technological innovation in both software and mechanization. Advances in precision agriculture will help tremendously in meeting the growing needs of our expanding population. I think the recent moves by Deere, as you outline, show a commitment of the organization towards that advancement.

    I think the “right to repair” for farm equipment is not a trivial issue – there are key factors that make agricultural equipment different than other products, like cell phones or cars, that are also going through right to repair discussions. One issue is the complexity of the machinery – combines and other farm equipment have intricate hydraulic and electrical systems, tens of thousands of parts (many more than passenger vehicles, for example), and are put through extremely difficult conditions. The low volume production of agriculture equipment can make repairs especially difficult to diagnose as an independent mechanic. The answer to right to repair is not binary, and the issues outlined here are only a few factors to consider. Agriculture equipment manufacturers, dealers, and consumers are working through to develop a solution with more consensus and understanding.

  3. I wasn’t aware John Deere was so entrenched in Machine Learning as it aims to develop its products and am fascinated by the “precision agriculture” concept. It’s so easy to overlook areas such as agriculture and farming when speaking about technologies like machine learning and artificial intelligence but this may actually be one of the most important applications of the technology. As the world population expands, food production will become an ever-increasing important factor and using ML and AI to increase crop yield may prove to sustain human life in the near future.

    To answer your question about repairing the machine especially as it becomes increasingly complex, I would envision something similar to the current state of the software + maintenance model we see in technology companies. Historically, you had software companies provide CD-ROMs in physical packages of which a consumer purchased once and then had to figure it out while not receiving any updates. As the technology and connection bandwidth improve on these technologies, I would imagine a subscription fee service for maintenance costs on the software, as well as potential physical maintenance on the machines if and when necessary.

  4. This is really interesting, Kate! I, like other commenters, was not at all aware of the technology-forward posture John Deere has in product development. I am always impressed when a company like John Deere (a $30B revenue market leader) chooses not to rest on its laurels, instead aggressively pushing the business forward through innovation. In this case, I particularly love products like the see-and-spray technology which provides benefits throughout the value chain by adding value to a John Deere product, saving farmers money on herbicides, and helping protect the environment.

    I do have significant concerns regarding John Deere’s practices as you described around locked-in, high-cost maintenance and repairs. Aside from creating financial pressure on farmers who are already struggling to compete with massive scale agri-businesses (e.g., Cargill), thereby eroding market competition and price protection for consumers, this has the potential to create significant negative earned media for John Deere in an era where many Americans are hyper-sensitive to threats on traditional American industries (e.g., Coal, Steel, agriculture).

  5. Often ignored for more “popular” industries, the benefits to improving agriculture could not be more understated. An almost perfectly competitive industry, every move to conserve resources or increase crop yields will pay dividends to the future of the human species. The US military has similarly used machine learning to predict maintenance schedules on its vehicles, so I see no issue in believing that JD can produce the same result. I hope to one day see the agricultural industry be fully automated. Machine learning may also allow farmers to rent equipment at a low cost, rather than owning and paying high up front capital costs.

Leave a comment