Artificial Intelligence

Voice Commerce: User First, Channel Second

When working with an emerging technology – like voice-activated AI, for example – it is easy to get one’s head turned by novelty. The experience itself is highly novel – not to mention unique, noted Luke Starbuck, vice president of marketing at Linc Global – because unlike other attempts at digital navigation, it is not centered on a screen. Even when a screen is present, said Starbuck, it is supplemental, not central.

And in the presence of novelty, it is easy, particularly for technology enthusiasts, to get sidetracked by the newest, flashiest, “whiz-bangiest” upgrades and innovations – because of that “isn’t it cool” factor.

Flash is attention-getting, but it has no staying power. A consumer interacting with an Alexa skill might be drawn in once or twice by something novel, but it’s actual usefulness that will keep a customer coming back and interacting. As Starbuck told PYMNTS in a recent conversation, that means when building these skills, one has to start incredibly simple and take into account the distinct possibility that their users know nothing about this experience.

“People really do have some phenomenal ideas on how to use voice-activated AI, and we have seen some really interesting skills,” he says. “But if you can’t break it down into simple terms that any user can access, you aren’t going to have success – because if it’s hard to use, the odds are good that the user won’t come back.”

So, how to build skills that are simple and useful?

Not Reinventing the Wheel (But Getting It to Roll Faster)

A lot of the best skills out there, Starbuck explained, aren’t radically different than what exists today, nor are they disruptive. Often, they are skills that are mirrored in other parts of the digital ecosystem, centered around a screen, and adopted such that they work better, faster and more easily than they did before.

It’s not about finding new things for a consumer to do, but figuring out what the user needs in their current processes and routines, and then giving them a voice-activated way do them better than they have been.

Which, Starbuck noted, is about both building the skill and finding the right mechanism to meet the customer at least halfway by introducing them to the use of the skill.

“When we work with brands, we are very focused on not just building the skill, but also building out how the skill introduces itself to its users and how those users will learn to use it,” he said.

And that educational process, he noted, actually needs to be coded into the DNA for its use throughout. In order for the technology to be both simple and useful, the consumers should not have to work too hard to figure out to get it to work properly.

“Even telling a person what they can do in the environment makes a real difference,” Starbuck said. “It is an easy thing to overlook, and it can be critical – just simple prompts for customers that tell them what they can do.”
 
Plugging in to the Right Use Cases

Voice technology, Starbuck noted, is an amazing technology, which is why Linc is so focused on developing for it. But that doesn’t necessarily mean it will be great in every context. One of the key factors in building a truly useful voice-activated skill, he said, is understanding which use cases work well for voice, and which ones really don’t.

As an example, Starbuck presented product discovery – probably not a great application for voice, because the kind of data that is most useful for product discovery is communicated visually with pictures. Although voice AI is occasionally supplemented by a screen with visuals, at its core it is not a visual experience, but a conversational one.

“One has to be really tuned into chunking the conversation into smaller pieces, and really think about how people talk to each other to convey information,” Starbuck pointed out. “A lot of this is really a study of limits, and how much feedback you can give someone at one time. Our finding is if you go much beyond five or six seconds, there is probably too much information in play to be useful.”

Starbuck said the key is studying actual conversation between actual humans – to get an idea of the proper “micro-chunks for data” – and determining the natural prompts for more detail. The goal is to build a “more fluid experience” that makes the user feel like they are having a more natural conversation with an AI, instead of trying to verbally articulate a specific code to get it to work properly.

“People speak in all sort of different ways, where different words are communicating basically the same commands,” said Starbuck. “The user shouldn’t have to figure out what prompt the machine needs – the machine needs to have its skill developed such that it does that extra work. The customer or user just needs to know they are being heard.”

It’s About the User (Not the Channel)

Voice is an expanding medium, Starbuck noted. Today it makes its home mostly on speakers, but that territory is rapidly expanding. In the not-too-distant future, the user might be in their car or office – and the goal for the skills developer is to serve customers while also bringing in all the data they can them, so the offerings make sense in the context of where the customers are and what they are doing.

“Something as simple as ‘what’s up with my order’ can mean a lot of things,” Starbuck noted. “If the customer is waiting on an order, they might be asking about where it is. If they have already received it, they might be saying something is wrong with it and may be triggering a return process.”

The AI needs to be able to bring in that information in real time, and then use it to inform the response to the user – and to do this for a variety of cases. Is the user in the car? What time of day is it? Is this a recurring order?

These are all relevant questions – and questions the AI can use to “take a measure of the whole landscape” to offer the user an experience that is actually tailored to their needs.

“When it comes down to it, this is not just about on conversational context,” Starbuck pointed out. “This is about an AI that is able to take a much broader overview, so it can understand the customer in a much more holistic way. People use different language and terminology – when we are actually able to look at a full set of data points, we get a better way to connect the user to the conversation.”

The users of AI, Starbuck noted, don’t really care about using AI for its own sake – no users are excited about a channel. What does excite them is what that channel can add to their lives, the friction it can remove, and the time, effort and energy they can save by tapping into it.

Give them that opportunity – to use something really useful in that way – and it is the best possible marketing a brand can hope for.

“The reality is that happy customers come back to shop again and again,” he maintained.

And Linc will be one of the first competitors in the 2018 PYMNTS Voice Challenge with Amazon Alexa – vying to build one of those customer-delighting, friction-busting skills.

Click here to see prototypes.

——————————–

LATEST INSIGHTS:

Our data and analytics team has developed a number of creative methodologies and frameworks that measure and benchmark the innovation that’s reshaping the payments and commerce ecosystem. Check out the latest PYMNTS report on driving gas pump payments to the C-Store

TRENDING RIGHT NOW

To Top