Today, we’ll have a look at where Artificial Intelligence makes sense and how it is used alongside the agrifood value chain. I don’t think there is anything more “hype” right now than Artificial Intelligence, with everybody experimenting with new models such as DeepSeek or fun apps such as Picsart. However, in our area of focus, agriculture and food, pinning down where AI is meaningful can be challenging. Indeed, while there is no clear “killer” use case, you can use AI for everything from precision farming to creating new marketing campaigns.

As shown in the mapping above, at least 12 domains exist where we observe intersections between AI and the AgriFoodTech innovation ecosystem. We can group the applications in four main blocks:

1 – AI in Agriculture

AgTech is the ecosystem with the longest history with artificial intelligence as it has been used for a long time in:

🤖 Automation, notably in robotics, has been using machine learning techniques for a long time. One example is weeding, where image recognition is used to recognise weeds to eliminate.

🚜 Precision farming and the development of tools to help farmers monitor the health of their soil and animals. Current improvement in large language models (LLMs such as ChatGPT) enable a whole new category of players to develop tools that make farm management easier.

Crop optimisation and the development of new crops are categories where applications of AI are more recent and where it helps startups such as Inari or Pairwise reduce the time it takes to bring optimised crops to the market.

2 – AI in food product development

In food development, if we exclude the last part more related to marketing, there are four big categories:

💡The generation of new concepts or ideas, validated by “AI panels” (such as the ones developed by FoodPairing) and consumer insights gathered automatically from thousands of sources.

🏗️ Product development assisted by AI is maybe the space where we see the largest amount of communication, notably from large companies (such as this example by Unilever). AI seems to be mostly used to reduce the number of prototypes to experiment with when looking to achieve specific functionalities, notably in the case of plant-based ingredients.

🔍Protein discovery & protein design: one of the new “hot categories” we identified in our recent trends report. Many startups such as Shiru leverage AI and large databases of natural compounds to identify proteins with interesting properties (sweetener, fats, caseins…).

🧪Bioprocess optimisation: companies using AI to find the right parameters to scale bioprocesses used in precision fermentation and cellular agriculture.

AI in retail

Beyond automation (and the development of robots, which is moving much more slowly than in agriculture), the main applications are mostly leveraging machin learning techniques rather than more recent AI developments. It’s mostly a question of data analysis for prediction for sales for foodservice (how much croissants should I bake today to avoid waste and optimise sales?) as well as retail (how much avocados to order for my grocery stores for the coming week?). A more recent application category would be food safety with startups such as Spore.bio using AI to detect pathogens in CPG factories.

AI in the relationship with the consumer:

📺 Marketing: that’s the most straightforward way for companies to engage with AI, even if it is not through tools explicitly developed with food applications in mind. Recent examples include the developments by Coca-Cola of a new can design through AI or a Christmas TV ad made with AI.

🧬 Personalised nutrition: with the emergence of large language models such as ChatGPT, many things that were impossible before are now becoming possible. With these models, it is now possible to create engaging and personalised experiences with consumers based on their diets, goals, and eventually some external data (biometrics, DNA, microbiome… Due to the LLMs’ unpredictability (i.e., their “hallucinations”), we still see very few companies attempting to advise based on them. This is, however, the space I get the most excited about. With more precise image recognition (to measure what you are eating in a snapshot) and more accessible biometrics devices (i.e. Apple’s watch and its competitors), personalised nutrition could finally move forward.

Now, what’s next?

First, if we look at the different categories, there is one common theme: data. The more data you have, the best you can be to create new crops, create new products, generate concepts, create new proteins… And in many instances, who has the best data? The existing crop, ingredient and CPG companies. Beyond the fact that it has become urgent for agrifood companies to structure their data, we should look at the partnerships announced between giants and startups to see which companies could be tomorrow’s AI for food leaders.