Digital assistants: how not to fall off the edges
A shorter version of this article appeared in Campaign.
Digital assistants, a staple of science fiction, from HAL to Jarvis, now promise to be genuinely helpful to millions of people. After a shaky, over-promised start, products in the market today have come on in leaps and bounds. What opportunities are digital assistants opening up for brands? And what are the new challenges they face?
In the middle of last year Google revealed that 20% of mobile searches are done through voice. Google and Facebook announced their own assistants that integrate with existing messaging channels, taking advantage of the explosive growth in the volume of messaging. Facebook announced its bot platform, with over 30,000 bots created so far. Microsoft showed how its Cortana assistant can introduce third-party chatbots into conversations at appropriate times, elegantly demonstrating how finding the right bot to use among those tens of thousands of choices is increasingly likely to be done by an assistant (how much of your marketing budget do you spend on persuading robots to select your brand?).
Devices like Amazon’s Echo, Google Home and LG’s Hub Robot expand interaction with assistants from focused (I take out my phone and look at it or talk to it) to ambient (I can casually “talk to the room”). These products’ ability to recognise what people say is remarkably reliable, and they have been tuned to work well for some common tasks such as controlling entertainment and home automation, and simple information like diary appointments, travel delays and weather.
Today this ambient interaction works in a room with a device. The next step will be ambient interaction you can have wherever you go. Rony Abovitz, founder of augmented reality company Magic Leap, describes his company’s goal as “all day, every day” technology. He envisions a device so light and convenient that you would wear it comfortably all day. It remains to be seen how quickly and at what cost they can achieve this, but in the meantime Vue Smart Glasses, due out in July, promise unobtrusive and permanent audio connection to your smartphone.
Ambient interaction is powerful because it removes that last barrier to starting a conversation. I don’t have to select anything; I can just start talking. This creates great opportunity to have more contact with customers and help them in a timely way. It also creates some important new challenges.
One of these I call the scope scaling problem.
Imagine I have an ambient digital assistant. I’m out shopping, and I like the look of a watch in store. The sales assistant takes it out and a try it on, looking in the mirror. My assistant, of course, is always ready and waiting.
“What’s my balance?”, I might ask. That’s pretty easy to answer: just check the balance of my current account and report it.
But people do not naturally speak in commands. I’m more likely to ask “Have I got enough for this?”. This isn’t the same question: to answer it accurately we now need to figure out what “this” is, and its price. We could solve this by having a further conversation (“How much is the thing you’re talking about?”), but this creates an annoying shift of attention.
What if I ask, perhaps most naturally, “Can I afford this?”. That’s quite a bit harder. Not only do we need all the same information as before, but we now need to put it in context. I may have enough money now; will I still have enough money after the rest of the month’s expenses? Or is this something I was saving up for specially, maybe in another account? And do I have to pay for it all now? Does the retailer offer finance? Can we do better?
If I would ask “can I afford this?” I might just as easily ask “Should I buy this?”. This is a lot harder! We need all the same information as before. But there are lots of other considerations.
Will everybody in my family be happy I’m spending £4,000 on a watch? I might be attracted to this watch because it’s the one that the Apollo astronauts wore to the moon. But do I know that they wore the classic model with a manually-wound movement and a plexiglass crystal? And that this one has a modern automatic movement and a sapphire crystal?
Is £4,000 a good price? If I want to wear it now, maybe; if I’m prepared to wait for one shipped from overseas, might I do better? But I want it now and I’ve got the money. Is it the assistant’s job to tell me that I’m looking at a fake?
This is the scope scaling problem. An ambient assistant has an unparalleled opportunity to embed itself deeply in people’s lives. And in an ambient conversation there are no guard rails. Over time, when things seem to work, we naturally extend the scope of what we ask.
Amazon’s Echo was built in large part on the work of a company called Evi, that it acquired. Evi’s founder, William Tunstall-Pedoe, told me that they had learned, through their experiments, that people strongly prefer action. If you say to your Echo, “play music”, it plays some music. If you don’t like the music, you’ll be happy to change it. If it doesn’t play music, but instead asks you to make some more choices first, you’ll find it annoying.
At any time, an agent can give up and — hopefully with grace — explain that it cannot answer a request. Every time we “hit an edge” in this way we create a moment of annoyance and leave the door open for someone else to take control of the conversation.
Many brands today are taking their first steps in conversational interfaces, chatbots and digital assistants. Getting the basics right is the proper focus. That should not distract them from the need, in parallel, to think hard about their strategy. What role do they want to have in a world of constant, rich, ambient interaction with digital assistants? What are the attributes of their brands that will help them? What kinds of relationships should they have with what kinds of partners?
How important is this? Smart phone users, as is often quoted, typically check their phones 150 times a day. Always-available ambient assistants, with their lower barrier to interaction, greater understanding of context and greater ability to be proactive, could drive ten times as much interaction. It’s worth getting it right.
