At January's Consumer Electronics Show in Las Vegas, the world's biggest cosmetics company, L'Oréal SA, showed off a small machine that it believes could transform the way consumers look after their skin.

Powered by artificial intelligence, the 6.5-inch Perso device does several things in quick succession. It links to a cellphone app to study the user's skin condition, including fine lines and deep wrinkles. It accesses geolocation data to assess that day's local environmental conditions that can influence skin, such as humidity and the UV index. The user can add information about any personal niggles, such as unwanted pigmentation. All that data is then crunched and a dollop of personalized skin care — in the form of moisturizer, serum or under-eye cream — is immediately dispensed at the top of the machine. Perso's neural network helps it learn the user's preferences and thus become "smarter" over time.

"It's the beginning of the marriage of physical and digital," said Guive Balooch, vice president for research and innovation at L'Oréal. The goal is to put personalized cosmetics "into the hands of everyone." L'Oréal says it plans to offer both the skin care version of Perso sometime in 2021 and will launch separate devices for customized foundation and lipstick later.

Like other businesses, consumer brands have long used artificial intelligence behind the scenes to speed up humdrum tasks, from predicting consumer behavior to tracking inventory and making manufacturing processes more efficient. Now, food, drink, beauty and fashion brands are harnessing AI's futuristic appeal to sell a range of novel products to everyday users. The companies predict such products will appeal to well-heeled consumers, but it is too early to say whether they will also become big revenue spinners.

The Procter & Gamble Co. jumped in early. In 2017 it launched the AI-driven Olay Skin Advisor, which uses a selfie to assess a user's current "skin age" and recommend a skin care regimen. The Cincinnati-based company now has put machine learning into the humble toothbrush: its $220 Oral-B Genius product, launched in October 2019, uses sensors to monitor pressure, track brushing time and detect the brush's location and orientation. Data sent to an AI-based app offers the user feedback about parts of the mouth that require extra attention. It is a way to "coach users to achieve better brushing habits," P&G says.

Two factors are behind the embrace of AI. The consumer industry is struggling through years of anemic growth, and AI offers a way to sell newfangled products at premium prices. At the same time, brands are keen to parse, understand and use their growing repositories of consumer data in a way that delivers a commercial payoff. One approach is to move from mass-market products to narrowly-defined ones.

"You're no longer making a product that you're going to sell everywhere in the world," said Steve Laughlin, general manager for the consumer industry at International Business Machines Corp., a major AI technology provider. "Companies are instead developing regional, subregional and hyperlocal products" by using AI to sift through local trends, Laughlin said.

In June 2019, Spanish food company, CEREALTO SIRO FOODS S.L.U., which releases more than 250 new cereal products in global markets each year, teamed up with IBM to launch an "AI-informed" snack specifically for the U.K. Its AI tool listened to what consumers and experts were saying in newspapers, magazines, Twitter, blogs and online forums and concluded that British shoppers wanted more foods low in sugar, salt, fat and additives and high in fiber and protein. Thus Cerealto's first AI-informed snack — rice cakes with pea and lentil — was born.

Other companies are using AI to hasten product innovation. It usually takes a beer-maker up to two years to test the plethora of flavors and aromas in beer samples using traditional lab methods. In partnership with Microsoft Corp. and two Danish universities, Carlsberg A/S has embarked on a "beer fingerprinting" project that uses AI-based sensors to map a flavor fingerprint for each sample. The approach could cut the time it takes to research taste combinations and processes by about 30% and could bring more distinct beers to market faster.

"The project is getting quite advanced and we hope to make it commercial soon," said Birgitte Skadhauge, vice president of Carlsberg's research laboratory in Copenhagen.

About 4,000 miles away, in Hunt Valley, Md., McCormick & Company Inc. has made a bold bet on using AI to help it create novel spices and flavorings at a faster pace than it has achieved in its 130-year history. Finding a new flavor is a hugely complex, lengthy process. A mix of science and art, it involves the analysis of millions of data points relating to ingredients, consumer taste preferences, historical formulas and flavor palates. A developer must not only figure out which ingredients to combine, but also the ideal ratio. Even a tiny change can make or break a winning formula. It's a task suited to smart computers.

With IBM's help, McCormick in 2019 launched its AI platform, dubbed ONE. It uses machine learning to sift through thousands of ingredients and raw materials to identify patterns and novel combinations. The effort has borne fruit; McCormick recently launched a series of AI-led recipe mix flavors, including Tuscan Chicken, Bourbon Pork Tenderloin and New Orleans Sausage.

Other companies are using AI to add a dash of pizzazz to their usual marketing efforts. In late 2019, the U.K.'s Diageo PLC launched its AI Whisky Selector app, which requires a user to answer 11 flavor-related questions, such as: Do you like smoked foods? How often do you eat bananas? The app then crunches the data and recommends a single malt that is the closest match.

Nike Inc., meanwhile, sought a fix for the 60% of people who it claims wear the wrong shoe size, which results in a flood of returns to online stores. Its AI-powered scanning app, called Nike Fit, measures the shape of each foot using 13 data points. According to the U.S. footwear company, the "hyper-accurate" scan enables any cellphone-wielding buyer to find a good fit in seconds.

Sweden's H & M Hennes & Mauritz AB (publ) offers an online tailored shirt service in Germany called "H&M Yours" in collaboration with fashion technology company ZyseMe. According to H&M, the algorithm behind the offering uses machine learning, statistics and real data from users to predict measurements. An AI-based tailored shirt costs between €34.99 and €39.99. In an email, a spokeswoman declined to comment on sales numbers but said the company was "happy about the current development."

"No one would claim that using AI to develop consumer products is mainstream yet," said IBM's Laughlin. "But the innovators are starting to dig into how to do this."