Brands such as L’Oréal have found success with AI in areas such as using image recognition to enable virtual product testing and skin analysis
Despite artificial intelligence’s promise to help marketers reach consumers in ways that are better, faster and cheaper, executives can wind up wasting time and money when they actually try it, according to industry experts and analysts.
“Every marketer wants to be famous and wants to say, ‘I’m using AI,’” said Stéphane Bérubé, chief marketing officer for Western Europe at L’Oréal SA. But they also think they’re experts on the topic just because they’ve been to a few conferences, he said. “Instead of saying, ‘What can I do with AI?’ they need to say, ‘Here’s what I would like to do with my dialogue or relationship with consumers, and can AI help?’”
The challenges continue from there.
Marketers think the technology is the hardest aspect of AI, said Bérubé. “I think it’s the easy part. The tough part is finding the purpose.”
L’Oréal could have better thought out the chatbot it introduced last year to encourage people to buy products as gifts, Bérubé said. It asks both potential gift-givers and the potential recipients a series of questions before suggesting options. “The consumer experience ended up being a bit complex,” he said.
L’Oréal improved that bot over time, Bérubé added, and introduced a “virtual skin expert” this year with stronger traction. It developed both bots with Automat.ai, which builds conversational marketing bots for brands.
L’Oréal also has found success with AI in areas such as using image recognition to enable virtual product testing and skin analysis, fueled by its acquisition this year of augmented reality and AI company ModiFace, he said. “How can a consumer be on her phone at home or wherever, in a store, and be able to try a product without necessarily trying it?”
It’s tempting to spend time and money on something sexy like a skill for a home voice assistant when unglamorous applications could pay off better, said Sucharita Kodali, vice president and principal analyst at Forrester Research. “When I’m in the middle of a grocery store, why can’t I shout, or talk, and say, ‘Hey Kroger, where is the light brown sugar?’” she said.
“Voice exists, and there are use cases that I have in the store, but nobody has married that together to provide something that’s useful to me,” Kodali added. “Those solutions, unfortunately, are built by Silicon Valley bros who often don’t even understand what customers want and therefore don’t build useful solutions.”
Retailers are more effectively using machine learning to prevent fraudulent transactions, Kodali said. “Those are things that are pretty categorically strong investments.”
Brands are panning for gold in a flood of AI promises from vendors that might not have the expertise necessary to deliver. “Most of them have no idea how the technology truly works and don’t have the talent required to put it into practice,” said Andy Mauro, co-founder and chief executive of Automat.ai. “And that’s where you get into trouble.”
A marketer should put two questions to any AI vendor, Mauro said: How many PhDs are on staff, and does the company focus on the marketer’s category? “There’s no free lunch in any of this,” he said. “In order to do a good job in AI, there does need to be a certain degree of specialisation.”
Testing is wise at the start, but too much caution can be counterproductive, executives said.
“We see clients that say, ‘Oh my God, this is available, let’s do a test,’” said Or Shani, founder and CEO of Albert Technologies, which offers an AI marketing platform to brands. “And they give you the worst brand they have in the worst market they have and then they don’t understand why it doesn’t work.”
Commitment also requires marketers to prepare their teams to welcome the machines, Shani said. “They sign off and we start dealing with their marketing team or media team and they see us as a massive threat,” he said. “They need to do some internal education to make sure people understand, ‘Guys, this is online to make your jobs better.’”
When the fashion design house Natori Co deployed AI from Albert Technologies to optimise its advertising in social media, the return on its ad spending soared, according to Ken Natori, president of the company. But there was a problem: “Some people commented that we looked like a discount website because so many sale items were being advertised on Facebook , ” Natori said.
Albert was increasing the number of sale ads that Natori ran because they were working—a reasonable response for a system instructed to maximise ad performance. But Albert didn’t consider whether too many sale ads could hurt a brand’s image.
“We said, ‘Listen guys, love the return here, but the percentage of ads that are promoting sale product is too high and we’re not comfortable what it’s doing for our brand,’” Natori said. The company fixed the mix by setting some limits.