NotCo: Creating Plant-Based Food Alternatives with AI

NotCo is a business based in Chile that creates plant-based alternatives for many different types of food including milk and meat products. Globally, it’s competing with companies such as Impossible Foods and Beyond Meat.

NotCo uses an AI-centric approach to developing plant-based alternatives. It has an AI that it calls “Giuseppe” that is used to create initial recipes. A basic overview of the AI can be found in the following video from YouTube’s “The Age of A.I.” series:

Essentially, “Giuseppe” is trained to generate a recipe using ingredients in its database in order to match a particular taste, color, or texture of a given food. The AI is trained on many different food combinations and is also given access to the underlying chemical components of the food. Given this info, the task because a pretty straightforward machine learning prediction problem. Given qualities of food that a human might care about (taste, texture, color, etc.), predict what ingredients, in what proportion, could be used to make that food. Data to train on can be made available by chefs attempting to make plant-based substitutes for foods or mixing random foods together and describing the qualities. The AI can then learn from this information and begin to predict some of the chemical interactions.

 

Value Creation

One of the current challenges that NotCo faces is that even though its AI is surprisingly good at guessing recipes, top chefs are still needed to perfect them. For example, when Giuseppe makes a recipe for plant-based “fish”, the taste-testers report that while it is a good estimate, it’s not quite right. This means that chefs will need to modify the original recipe in order to create something that actually mimics the properties of fish needed for it to become a realistic substitute.

Why is this a challenge? Because the use of the AI is supposed to create value by making it easier to find these recipes, which cuts down on the costs of hiring chefs and allows them to create recipes more quickly. However, if chefs could have just come up with the recipe on their own in a similar amount of time (and for a similar amount of pay; computer scientists are expensive), and Giuseppe’s creations still require chefs to perfect, then using Giuseppe actually doesn’t give NotCo an advantage at all.

However, this also presents an opportunity. While it may be the Giuseppe is not up to par right now, it may be that with future algorithms and data, the AI will be much more cost-efficient as well as “creative” than chefs currently are, since AIs are generally able to compute more possibilities for solutions than humans. By collecting this database and building themselves as an AI-centric company now, NotCo may have the advantage in the future as their AI begins to blossom.

Right now, I would recommend that NotCo work on creating a system that allows for many more tests of Giuseppe. Similar to the boat case we analyzed in class, where having two boats tested at the same time could improve testing, if Giuseppe can create hundreds of different recipes and a pipeline can be made for having people taste and give feedback on multiple different recipes, then Giuseppe will gain large amounts of useful information for its prediction system much more quickly.

The same infrastructure that allows for the rapid testing of Giuseppe’s creations could also then be extended to other chefs as well. Similar to how AWS allows Amazon to generate revenue from a system that it can use internally, developing a system for quickly taste-testing food and receiving feedback may be a way for NotCo to develop another revenue stream beyond their food products.

 

Value Capture

On the value capturing side, NotCo is uniquely positioned as a company from Chile. Instead of starting off with creating plant-based meats, as many American businesses did, they began with mayonnaise, a popular condiment in Chile. I would recommend continuing this practice of creating food substitutes that match the culture of the location where the business is located. While not having meat products may make it difficult to compete in the American market, having a plant-based milk product (as NotCo does sell) may make it possible if they decide to move in that direction. For now, I think gaining a strong following in their own country for creating the best plant-based substitutes, and then branching out, would be the best strategy. It ensures that they have a steady revenue stream that’s able to support failure in different markets and also increases their awareness and branding before expanding.

Previous:

TOMRA – Potatoes to the Right, Rocks to the Left

Next:

Youse: Using gamification and AI to deliver innovative auto insurance

1 thought on “NotCo: Creating Plant-Based Food Alternatives with AI

  1. Very interesting! This is one of the few cases I’ve seen where AI is being applied to taste, and it will be cool to see if/when it gets to the point where it can provide taste-accurate recipes without chef intervention. I wonder though if there is a limit to how accurate these recipes can get, since I imagine some of the data being fed would be limited by a person’s descriptive capabilities and subjective taste.

Leave a comment