Accelerating innovation with AI

Developing a new product, be it medicine or industrial, chemical or food product, requires the assessment of a large number of possible ways in which various ingredients can be combined to create the end product with desired properties. AI can accelerate the innovation process by enabling a rapid fire analysis of hundreds of thousands or millions of possible combinations. Let's look at how McCormick, a company in a very old industry, food industry, is using AI to accelerate the development of new products. McCormick is the world's largest, the company in the seasonings, sauces and condiments business. These ingredients deliver flavor to a dish. Each new generation, from baby boomers to Gen X, to Millennials to Gen Z, seems to have a stronger preference for new and bold flavors. Historically, McCormick has relied solely on several hundred product developers based in labs within 14 countries. Some are food scientists, nutritionists or chefs. Others are chemists or chemical engineers. Ever since the company's founding in 1889, developing a new product has been more art than science. For any new product, the developer starts with a basic recipe, also known as a seed formula. He or she then tweaks this formula by adding, deleting, or changing ingredients. As developers came more experience, they tend to gravitate towards a favorite set of ingredients. They develop go to solutions that they trust. While this approach has worked for over 100 years, the company's leaders wondered if AI may help do a better job. Could they develop an intelligent platform to create product formulas, one that learns and it gets better with experience? In 2019, McCormick signed a formal collaboration with IBM to explore the use of AI for new product development. The collaboration has proved highly successful, and the company's now has several hundred new products in the market conceived entirely by AI. McCormick started by training the AI algorithms on historical data about recipes and their ingredients. The company's database included over 400,000 formulations prepared from over 14,000 ingredients, many in different languages and spread across countries. After considerable effort as standardization and centralization, the database added up to 1 billion data points. These data became the raw material to train AI algorithms on a number of attributes, such as what ingredients can substitute for each other? An example would be lime and lemon. What ingredients complement each other? An example would be basil and oregano. Which ingredients are incompatible with what applications? For example, if the application is a dry seasoning, then the flavor recipe should not include liquids. What defines a success? Which suggestion has the highest probability of success? Based on learning from the vast database the AI models are able to generate several formulations with a high probability of success. These formulations might use unique ingredients, different combinations of ingredients, and or different ratios of ingredients. Product developers and marketers have been surprised at entirely novel seasonings that, in hindsight, makes sense, but given that tunnel vision were beyond the imagination of a human developers. So how does AI accelerate product innovation? In the traditional approach, each developer relied solely on his or her experience. In contrast, the AI models rely on the collective expertise of all product developers in the company's entire 130 year old history. If AI can help develop new seasonings and sauces, which are extremely traditional products. Think about the opportunities that your company may have to do something similar in its domain.