Preparing for brands that talk

With voice assistants, brands are figuring out how to insert themselves into a new kind of conversation and the high-stakes method of search.


This article appears in the October 2017 issue of strategy.

Marketers today would be forgiven for hearing voices in their heads. They may not all be identical, but they share certain elocutionary quirks and the same slightly halting delivery. They’re mostly female. And regardless of the words they use, their message is more or less the same: don’t get left behind.

In the quest to be present wherever consumers are ready to buy, voice assistants are a growing concern. There are the well-established ones from the tech companies: Alexa for Amazon, Apple’s Siri, Google’s Assistant, Cortana for Windows and Bixby for Samsung. But some brands are looking at building their own personalities, from bots to voice assistants, like Domino’s pizza-ordering AI personality, Dom.

The market for smart speakers in Canada is miniscule: Google Home only launched here in June, and Amazon has yet to release the Echo north of the border. But in the U.S., adoption is on the rise. More than 45 million voice-enabled digital assistants are in use (roughly 70% of the smart speakers are Amazon’s and 24% are Google’s), per eMarketer, and that number will rise to almost 67 million in 2019.

Facebook is working on its own hardware, including a smart speaker to compete with Amazon and Google, and a video chat device named Aloha that could be released next spring, according to various reports. In August, Google and Walmart announced a partnership (U.S. only for now) between the retailer and the Google Assistant. Walmart customers can link their accounts to Google to receive personalized shopping results based on previous purchases.

The devices are still clunky as purchase tools, and industry experts consulted for this article agree we’re in “the gimmick phase” of how brands are using them. Customer service is still the most practical use for chatbots and voice assistants, a June report from eMarketer said. When it comes to shopping, the platforms are best suited to reordering or shopping for products that don’t require much research. Consumers have concerns around security and privacy, payment, and the ability to search products, the market research company said.

But agency creatives and developers say clients in Canada are already asking how to prepare for voice in order to avoid being left behind like some were with mobile. How will voice assistants affect search and purchasing? How does online content need to change for a voice-enabled world? Do brands need to develop their own voice or personality to serve up that content, or do they leave it to the Alexas, Siris and Google Assistants of the world? And if they do want to develop their own voice, what do they need to take into account?

The multi-million dollar question

In June, not long before announcing its acquisition of Whole Foods, Amazon released a video for its Dash Wand. A woman asks it for recipes and uses the device to scan items in her expensive fridge. The ingredients that automatically appear on her shopping list are soon delivered in green Amazon Fresh cooler bags.

As the dinner party prep proceeds, her husband tells the device they need white wine. “Your favourite white wine will be here between 4 p.m. and 6 p.m.,” Alexa responds. When he tells the Wand to dim the lights, down they go. The guests arrive for the dinner party, which resembles a scene from a Hallmark movie with a conspicuously large budget.

It’s the vagueness of the “buy white wine” command that has some marketers worried. Alexa typically responds with only two suggestions when asked to make a purchase. According to a July report from research firm L2, those suggestions are usually Amazon’s Choice products, meaning brands must earn that label to be found on the devices (Amazon uses a combination of Prime eligibility, sales performance and customer reviews for the distinction, rather than going by the top search rank or the best seller).

As brands tinker with bots and voice assistants, they’re also preparing for, or at least contemplating the consequences of, a world where voice is a dominant form of search. It already accounts for 20% of searches in the Google app, according to the company.

The new holy grail for search is “position zero,” says Neil Mohan, VP of digital and creative technology at Edelman. That’s the featured snippet Google offers, often from a Wikipedia entry, which appears above the number one ranking and includes the first few lines of text from an entry.

To succeed in voice, brands will not only have to produce relevant content to be delivered in that medium, but make sure that content is positioned as the first or second result. Otherwise they risk not being found.

That may involve developing content designed to answer specific queries the voice assistants are likely to get, and filling gaps where no answer exists. Mohan says he’s working with Unilever to reconsider website design and make sure copy is written and coded to work for voice search. They’re imagining situations like someone cooking, hands occupied or too messy to touch a device, asking a voice assistant about an ingredient – the kind of question its brands would hope to answer.

“I want the brand to be there with the kind of answer a trusted friend would give,” Mohan says.

That means writing audio-friendly text: staying away from certain words that voice assistants struggle with and being mindful of the length.

L’Oréal also sees opportunity in the new hands-free search world, and wants to be there to provide expertise to the woman asking for beauty tips mid-application, says Martin Aubut, the brand’s chief digital officer. L’Oréal is learning about the types of questions consumers ask through its chatbots (more on this below), and is tagging all its content to be pushed to the consumer in the right context.

“You will be able to provide the right answer to the right question in the right format, which is voice or text or sending a picture,” he says.

Voice search may also force brands to shell out a lot of money, says Mitch Joel, president at Mirum.

“You can imagine what that cost is going to be once [search engines] understand the value and merits of being the A1 answer in a world where you’re not going to see five, six, seven, 10 results on a page,” he says. “It may start as a function of who can provide the best result but it’s going to quickly evolve to who can pay the most for that coveted spot.”

Joel sees the era of voice search as a return to the days when “advertising was primarily a scarcity model.” Before digital media, where advertisers have endless opportunities to reach a targeted audience, if a brand’s ads weren’t on prime time TV or didn’t make it into the morning newspaper, they simply missed out on that audience. The same could be true for voice.

It could also mean a challenge to “Google’s insane monopoly,” Joel says. The shift in search query format could provide competitors with an opportunity to challenge for the lead in next-generation search.

First chatbots, then chatter boxes

More emphasis on voice will create opportunities for brands to go further upstream to partner with tech companies, like RBC’s partnership with Apple allowing customers to pay their bills and send transfers through Siri (PayPal also has a partnership with Apple).

But brands’ focus on voice goes beyond the implications for search and how to interact with the proliferating Siris of the world. Some are also looking at developing their own voice personalities to interact with consumers on devices.

beautygifter-622x349As a warm-up to a full voice assistant, chatbots (a smaller investment) allow brands to learn how consumers interact with AI and to experiment with brand personality. In May, L’Oréal Canada launched its Beauty Gifter bot with Montreal digital agency Automat to help consumers on Facebook Messenger find the right gift. Through a series of questions, the bot learned about the recipient’s beauty profile and made product recommendations.

It’s part of an effort to test Messenger as a transactional platform, CMO Stephanie Binette says. The bot was less about developing a personality, since it was a single service speaking for L’Oréal’s 39 brands, and more about learning to provide personalized content to consumers at scale.

“In that sense, the brand tone was not as necessary. Obviously that’s where we want to go, but this was test and learn,” she says.
L’Oréal also debated how focused the bot should be on the sale. “You don’t want to be too functional. You want to create a relationship,” says Aubut, the chief digital officer. He compared the interaction to dating, where the two parties build trust with simple questions.

Michael Wandelmaier, associate product design director at Huge, has been working on chatbots and voice services for brands, mainly on the customer service side, and has faced the question of how much personality the assistants should have.

His research suggests it comes down to timing. There are occasions when consumers want the assistants to “get out of the way and deliver the information” rather than demonstrate any personality, he says. “And there are times when it makes sense to manifest a lot of the brand values in order to set expectations, to remind people where this is coming from.” Deciding when comes down to context.

Ideally, a bot could adapt to a customer who is in a bad mood and making a complaint versus one who is shopping or seeking information. Language processing may be getting refined, but it’s not nearly as tone sensitive as a real human, Wandelmaier says.

So for customer service, some built-in casual quirkiness can feel inappropriate and even irritating.

Many times, “a simple, straightforward assertion of fact is received much better than anything with a layer of personality to it,” he says. “The moments for personality are really bookending conversations: at the beginning introducing themselves, and at the very end.”

Just as L’Oréal is learning from the Beauty Gifter before developing voice services, RBC launched its AI financial assistant in August, which could eventually become the brand’s customer-facing voice.

RBC’s CMO, Alan Depencier, describes Nomi (named from the insight that customers want the bank to know them) as a kind of Siri for RBC, the bank’s own AI brand personality. It will provide insights about spending and account activity when customers log in to the mobile app, and learn which ones are useful to a specific customer through a rating system. It will also use predictive technology to make saving decisions.

While there’s no voice capability in Nomi’s current version (Nomi has no image or gender yet, either), “In the future it will be where we take our voice brand,” says Depencier.

The bank used mostly in-house design, tech and innovation specialists to develop the service, with ethnographic research central to the process. RBC’s partnership with Apple is about helping customers on the channels they’re already using, Depencier says, but the bank also wanted to “build a capability and an overall brand personality and experience that’s unique to RBC.”

voice1How many voices is too many?

Huge’s Wandelmaier says the big tech companies may want to develop “a centralized brain for their customer experience” that transfers across all devices – the voice assistant in your car, for example, will remember the conversation you had yesterday with your in-home smart speaker.

But there are risks for brands positioning their assistants “as extensions of the self,” he says. “This is at odds with what brands would like to be – essentially the brand ambassador talking to the customer.”

Wandelmaier thinks there will be opportunities for brands to bud in on the territory of the assistants like Siri and Alexa. (Some of the assistants are themselves cooperating. In August, Amazon and Microsoft announced that Alexa and Cortana would soon be on speaking terms. Amazon CEO Jeff Bezos predicted that assistants would one day be smart enough to pass requests to whichever assistant is best suited to answer, without requiring the command, “Alexa, open Cortana,” or vice versa.)

Having Google Home find a recipe from a specific publication could leave room for a brand to intervene and answer the query in its own voice, for example, rather than Google Home reading it out. Dominos already has such a partnership with Google: “OK Google, order Domino’s,” a user says, and the brand’s own AI assistant, Dom, takes over the process.

“At least in the near-term, I think that’s what all major brands playing in this space are going to want to do,” Wandelmaier says. “It’s essentially the most visible part of their return on investment: not only can we serve the customer better, but they are experiencing the brand when they interface with these services.”

But he’s not sure that’s the right approach for the user. Interacting with dozens of different brands a day could become overwhelming.

Brands should focus on designing for a “utilitarian experience,” Wandelmaier says, helping customers achieve what they’re looking for through the user interface. “That is still more palatable and practical in the short-term than a jaggy, branded and human-emulating-style experience,” he says.

One way to avoid some of the challenges around creating a brand voice and personality is to allow users to do it for you. Google-owned navigation app Waze is taking this route. In August, it launched #NextWazeVoice, encouraging users to record themselves delivering Waze’s stock navigational lines (“in 800 metres,” “speed trap reported ahead,” etc.) to then share with friends.

The app had already partnered with brands on voices as promotional tools (Owen Wilson’s character from Cars 3, for example, in a partnership with Disney). With #NextWazeVoice, it’s inviting YouTube creators to record their own custom voice packs to share on their channels, offering a new platform for influencers to reach fans in a particularly intimate environment.

“This is a totally new format for content creators to entertain and engage their fans and ultimately promote their brand,” says Mike Wilson, Waze’s country manager for Canada.

Families and friends recording their own versions is a model of personalization that could be adopted by other brands. Several experts consulted for this article said personalized brand voices are a future possibility.

In the meantime, there will be road bumps for brands developing voice tools. In addition to concerns about privacy, there are social barriers. The “uncanny valley” phenomenon where robots that are nearly – but not quite – identical to humans evoke unease or revulsion is true of voices, too.

“The technology has matured but our understanding of human nature has not progressed that fast,” says Vassilios Alexiou, group director for creative at Huge. “A lot of people just don’t like talking to machines.”

As with the Google Glass and IBM’s Watson, voice assistants might be better suited to the workplace than for mass consumption right now. Alexiou says Huge is working with corporations on tools for workers in labs and other spaces whose hands are occupied and need to interact by voice.

“The consumer-facing stuff is further along the journey,” he says. “We’ve got to learn a lot more before we can create very successful applications of that.”

So why are they all women?

In Spike Jonze’s 2013 film Her, Theodore Twombly (played by Joaquin Phoenix) is asked whether he would like his AI operating system to have a male or female voice. “Female, I guess,” he answers, without seeming to give it too much thought. So he gets “Samantha.”

The decision isn’t a perfunctory one for brands. Many have criticized companies’ choice of gender for their voice assistants, inserting females into the secretarial tasks of scheduling, assisting, etc. traditionally held by women.

Some people object to using assistants because they don’t want their kids “learning to shout at a female,” says Vass Alexiou, group creative director at Huge.

The onomastic histories of voice assistants are complex. First of all, why give them names at all? Researchers at Northwestern University’s Kellogg School of Management tested people’s trust of technology using a driving simulator named Iris (Siri backwards), and found that drivers trusted it more than those using an unnamed simulator.

Names that are easily distinguished from other speech (so as not to accidentally activate the assistant) are also important. But there is always a binary choice.

Siri means “beautiful woman who leads you to victory” in Norse. Alexa was inspired by the ancient Egyptian Library of Alexandria (many have asked “Why not Alex?” then). Cortana was an AI assistant in the game Halo, where she appears as a mostly naked hologram. Google avoided personification, going with “OK Google” to trigger the Google Assistant. But the voice is still a female’s. Only Siri allows users to change to a male voice.

Citing various studies, the Wall Street Journal reported that people respond better to AI if the robots are given gender cues. Both men and women see females as more welcoming, warm and nurturing, according to an Indiana University study. People prefer to learn about computers from men but to receive counsel on love and relationships from women, a Stanford University study said.

The WSJ pointed out that there are cases of brands choosing male voices – such as Dom, Domino’s pizza-ordering voice assistant. (The target for that one is fairly predictable.)

Agency Fjord addressed the dilemma for brands in its online guide to designing voice interfaces. Applying a gender identity has ramifications, “especially because the resulting impulse is to then add a ‘her’ to every product we can,” the guide says. “Instead, we should pay attention to the unexamined decisions we’re making to avoid digitizing existing power structures under the guise of a ‘default’ identity.”