How PepsiCo Is Using AI and Machine Learning To Deliver Success

A production line at PepsiCo. Credit: Adobe

One business who realized that using artificial intelligence (AI) and machine learning is a business need, no longer a competitive advantage is PepsiCo. The food-and-beverage company behind brands such as Pepsi, Gatorade, Tropicana, Lipton, Frito-Lay, and Quaker sells products in more than 200 countries and brought in $64.7 billion in annual revenue last year. From robots to machine learning, PepsiCo uses AI and machine learning throughout the organization in many ways.

Snack Delivery Robot

There’s a six-wheeled mobile vending machine robot tooling around the University of the Pacific chockful of PepsiCo snacks and beverages from Hello Goodness—a healthier line-up that includes SunChips, Baked Lay’s and bubly sparkling water. Named Snackbot, these self-driving robots are a partnership between Robby Technologies and PepsiCo. Students can order their snacks from the Snackbot app, and then the robot will deliver it to more than 50 spots across the campus without charging a delivery fee. The bots have a range of 20 miles on a single battery charge, and they can even navigate at night, in rain or up curbs thanks to onboard headlights and all-wheel drive capabilities.

Snackbot represents the solution to the needs of strapped-for-time college students and their preferences identified through PepsiCo’s research. There are three to five Snackbots on campus to keep up with demand.

Manufacturing with Machine Learning

The Frito-Lay (a subsidiary of PepsiCo) manufacturing plant is benefiting from machine learning. One project uses lasers to hit chips and then listen to the sounds coming off the chip to determine texture. Algorithms process the sound and determine the chip texture to automate the quality check for Frito-Lay’s chip processing systems.

From this beginning, Shameer Mirza, senior research and development engineer at PepsiCo, realized several more applications of machine learning could impact process control within the factory. Next, Mirza developed a machine learning model that could be used with a vision system to be able to predict the weight of potatoes being processed. This led to considerable savings for the company because it no longer had to spend $300,000 per line (they had 35 in the U.S. alone) for weighing elements. Mirza’s systems used only a camera and the machine learning model and are essentially just additional data points collected with no additional cost.

Another project still in development would assess the “percent peel” of a potato after it had gone through the peeling process. By understanding this data, it can help the Frito-Lay team to optimize the potato peeling system. This project alone is estimated to save the company more than $1 million a year just in the United States.

PepsiCo is launching a global training course on advanced machine learning and computer vision for its internal research and development associates this year to broaden its team’s abilities to use these technologies to continue finding insights that will drive efficiencies in its manufacturing facilities.

Vera Streamlines the Hiring Process

PepsiCo used Robot Vera for the first time to phone and interview candidates for open positions in sales, as drivers and to fill factory vacancies in Russia when HR professionals needed to fill 250 jobs in two months. Vera was developed by Russian startup Stafory and is capable of interviewing 1,500 candidates in nine hours, a job that would take humans nine weeks.

Advanced speech recognition software and tools from Amazon, Google, Microsoft and Russian technology company Yandex allow Vera to make calls and screen candidates for open positions such as fork-lift operators, factory workers, and sales staff. Its software can scan CVs to determine if a potential candidate has the right experience for the position, can respond to yes and no answers, ask follow-up questions and send out follow-up correspondence. It can also forward transcripts of a call to a human HR specialist for further review. So far, the reception from the majority of candidates when dealing with the robot has been positive. There was a bit more hesitation from human HR professionals. It turns out one of the biggest hurdles is “reprogramming humans” to feel comfortable with the technology.

Read the source article in Forbes.