Categorizing data important to success of artificial intelligence

KANSAS CITY, MO. — Organizing data, more so than having enough data, could cause problems for food and beverage companies embarking on artificial intelligence (AI) projects, said Vinay Indraganti, founder and chief executive officer of BCD iLabs, Hoffman Estates, Ill.

Customers tell BCD iLabs they have lots of data from experiments, Indraganti said in a webinar with sister publication. Food Business News. A recording of the webinar may be found here.

“When you actually start pulling that data out, we see people are scrambling,” he said. “It’s with this person, it’s with that person. It’s here, there or somewhere else, or everywhere, or sometimes nowhere to be found. So that’s the first thing that hits people is, ‘Well, we think there’s a lot of data. We believe there is a lot of data.’”

How experiments are done may vary. Research in a study on a product may be done differently than research in another study, even though both studies involved the same product.

Data may come in different forms: texts, emojis and Instagram photos, said Malay Shah, chief business officer for Singapore-based ai palette. Within companies, data sets may come from different teams, like sales and customer service. Companies might spend 80% of their time on projects getting data and data architecture correct, he said.

“The biggest challenge is in understanding all these data sets,” Shah said.

Main AI benefits come in saving time and reducing costs.

Indraganti said one company spent two years trying to add more protein into a dairy beverage. Then, after using AI for seven weeks, the company increased the amount of protein in a 100-ml serving to 13 grams from 9 grams.

Shah noted PwC has said AI may help companies lower food and beverage product innovation costs by as much as 30%. He said the Hershey Co., through AI, screened 20 flavor concepts in one afternoon when creating a blueberry muffin-flavored KitKat sold in the United States and Canada.

“Hershey was able to get inspiration in terms of what was happening in the bakery space and then bringing it into the confectionery space as well,” Shah said.
 
 Some in the food and beverage industry may wonder if AI’s growth will lead to significant job losses.

“If you’re expecting me to tell you that food scientists suddenly will have to become data scientists, well, that’s not really the case,” Indraganti said. “We don’t expect that.”

AI will bring data scientists into product development, he said, adding the data scientists and the food scientists will need to develop a common language around data sets and parameters so they may work together.

“We cannot take food scientists out of food product development,” Indraganti said.