Brands pair with Amazon Echo

Capitol One, Fitbit and Domino's are among the marketers embracing the online retailer's voice-controlled digital assistant

Late last year, Amazon released Echo, a cylinder-shaped apparatus that operates like Apple’s Siri but isn’t a smartphone. What is it then? For marketers, it’s a new brand channel.

Though new, Echo has already integrated hundreds of "skills" developed by third parties that include brands like Capital One, Uber, Spotify, Domino’s and Jeopardy. Echo even gives users access to horoscopes from Elle magazine.

Echo’s latest brand partner is Fitbit. The wearable device maker announced today that it is integrating Alexa, Amazon’s voice-controlled digital assistant into its flagship product. That means Fitbit users can now ask Alexa (which is also available with Amazon’s Fire TV) how well they slept last night, how many calories they have burned, and how many steps they’ve taken, among other health-related queries.

"I think there’s a lot of possibility in the voice space," said Tim Roberts, EVP of interactive at Fitbit. Roberts foresees a day when Echo can take note that its owner has just completed a run and then congratulate the owner. "There are more and more intelligent ways to provide motivational feedback."

Consumer-facing voice recognition services like Alexa are the tip of the spear for the rapidly spreading use of artificial intelligence in marketing. Behind the scenes, more brands are tapping two primary components of AI — natural language processing and machine learning — to automate interactions with consumers and let machines make decisions about marketing that were previously made by humans.

Natural-language processing
Voice recognition used to be pretty bad, but in the last year or two it’s gotten quite accurate. As Tim Tuttle, CEO of MindMeld, wrote in a 2015 op ed for TechCrunch, speech recognition improved 30% over the previous year, which is a bigger performance gain than the previous 15 years combined. Partially this is because AI is now being exposed to a mass audience. When more people use it, the technology learns more quickly and improves at a rapid pace.

Rob Pulciani, director of Amazon Alexa, says he’s seeing the same effect, and he’s also seeing consumers warm to the idea of talking to a machine like they talk to a person. "Voice is very important to us," says Rob Pulciani, director of Amazon Alexa. "We believe it will fundamentally change the way people are interacting with content and devices." Pulciani says when customers use a voice service, they realize how much more natural and convenient it is.

Thomas Husson, an analyst with Forrester Research, said this evolution is a huge opportunity for brands. "Moving forward, people will not open apps whenever they need to simply glance for information, they will expect to be served proactively via timely and relevant messages," he said. "Leveraging natural language processing to answer the needs of consumers in context will be a game-changer. Smart agents like Microsoft Cortana, Amazon Echo, Google Now, Apple Siri or Facebook M aim to become trusted intermediaries between brands and consumers, somehow replacing traditional search."

Such natural-language processing doesn’t just apply to search. The North Face worked with IBM Watson and Fluid to create an AI-powered online shopping assistant that asks questions like "What kind of activities will you be using this jacket for?"

Sentient created a visual search engine for that lets shoppers find just the right boot by looking at images, rather than using search terms. And Bloomreach powers on-site search engines so they can interpret questions like, "What’s that thing in the living room you put your feet on?" said Bloomreach CMO Joelle Kaufman.

Machine learning
Josh Sutton, global head of artificial intelligence practice at Publicis.Sapient, said that he believes brands will increasingly use AI behind the scenes to "identify what matters to a given individual or a given audience at a very specific moment" and be able to provide them with the right message. Sutton said an AI system can look at "what’s going on in a person’s life" and offer them bespoke messages.

For example, looking at the data, an AI system employed by a bank might discern that a person is looking to build an extension on her house, considering a move or has kids that are soon going off to college. In that case, when the consumer visited the bank’s site she might see very customized content. "They’re not dealing with the problem that all firms have which is they treat all customers as if they’re brand new customers," Sutton said. "Rather, they’re leveraging all the insight that they should be able to create about that person to show them things that matter in the moment."

That sounds like the same idea behind programmatic. Mark Torrance, CTO of Rocket Fuel, said programmatic is a broad term that means "computers are involved" in the buying process. Most programmatic firms identify target customers and lookalikes to those targets and segment them together. Torrance said that is a form of AI. "It’s better than random by a lot," he said. "But it’s not treating those people differently throughout the day depending on the context of what they’re doing."

In contrast, Rocket Fuel and a few other firms including Criteo and Quantcast are using a more advanced form of AI that "predicts the potential of every moment" and presents ads that are likely to be received well at a specific moment, based on data about the user and the circumstance. "Our goal is to be really, really good at that so that we take people out of the equation, at least for making those kinds of decisions," Torrance said. "Machines are just much better at incorporating all the types of data that can bear on a decision. The difference is night and day."

Beyond that, machines are now helping brands make marketing decisions. Kia used IBM’s Watson to help identify influencers on Twitter who exhibited the brand’s preferred personality traits, including "artistic interest" and "achievement striving." (Kia and IBM reps could not be reached for further comment.) The BBC has also employed the New Zealand-based AI firm Parrot Analytics to gauge demand for programming. "Since unsupervised machine learning is much better than humans in identifying patterns it makes sense to use it to predict the best future outcome," said Arturas Vedrickas, VP of Product for Parrot. "In TV and Film that may be identifying what should be changed in a script to make the content more interesting to people, or how to allocate marketing budgets in terms of ideal platforms, audiences and creative, and then relentlessly optimizing based on what’s happening across earned, owned and paid media."

He added, "When it comes to identifying patterns and finding the unexpected, absolutely all assuming that the training of the AI has been performed on a good data set, machine learning models can easily consider millions of combinations and pick the right way a for a task at hand."

Finding the right tone
As machines take more of a role in representing brands and making decisions on behalf of brands, human intervention will be required to make sure the brands don’t come off as heavy-handed.

Sutton said he recently worked with a bank in the UK that has made a big investment in AI. The bank is concerned about how it comes off in direct interactions with consumers (a la Amazon Echo) and in the implicit leveraging of data. It was considering "how they use that in a way to show they know them and do it in a way that is helpful and not creepy."

Agencies may help such efforts by humanizing the brand, he said: "When you look at the role of agencies in the future, I think that’s going to be a much bigger role, not how do you get your messages across, but how do you get your messages across in a way that registers with your consumers," said Sutton.

Meanwhile, AI’s role in marketing — and our day-to-day existence — will only increase. "Naturally, we will see more and more applied AI in our everyday lives," Vedrickas said. "However, it will not be transparent to us. It is now and is going to be all about solving very specific tasks — programmatic advertising, search, natural-language processing. Things will become really interesting once we solve knowledge integration problems, in which case AI would be come all-encompassing."

Start Your Free 30-Day Free Trial

Get the very latest news and insight from Campaign with unrestricted access to, plus get exclusive discounts to Campaign events.

Become a subscriber


Don’t miss your daily fix of breaking news, latest work, advice and commentary.

register free