November 4, 2016 | Sean MacPhedran
“Bots are the new apps.”
Satya Nadella, CEO, Microsoft
"HO! HO! HO! My elves are hard at work uploading my intelligence to the cloud!”
Santa Claus, CEO, Northern Manufacturing & Distribution, Inc.
One of the emerging technologies that promises to radically alter the way people research products and services, interact with brands, and ultimately complete transactions, is the conversational interface. Bots, programs that use chat as an interface on platforms such as Skype and Facebook Messenger, have been around for some time, but several technology trends are rapidly accelerating and converging in a way that is going to give them the level of impact previously imagined in science fiction.
As a sketchpad for us to think about this out loud at SMITH, we’re trying to sort out how to upload Santa's magic into the cloud, but I’ll get to that later. First, the trends.
The Explosion Of Conversational Apps
Conversational apps are continuing to steadily grow in use. Facebook Messenger has 1 billion Monthly Active Users (MAU). Facebook-owned WhatsApp also has about 1 billion MAU. Skype has more than 300 million. Line, popular in Asia, has over 200 million. The most basic conversational text interface, SMS, is used by just about everyone with a phone.
All those people looking at their phones while they walk around are probably engaged, to some degree, with a conversational interface. Facebook is not a shortsighted company, and it’s notable that they’ve decoupled Messenger from Facebook as a standalone product with its own app and, now, Messenger.com.
As an aside on the potential impact of conversational apps for commerce, Africa's SMS-driven M-Pesa mobile payment system that started with simple PIN code driven SMS text transactions has sparked a trend that now has over 40% of Kenya's GDP flowing through mobile payments.
The Growing Open API Ecosystem
While people talk to other people through software, software is also talking to other software. Application Program Interfaces (APIs) allow various programs to communicate with one another. For example, if you want to add Bing Search to your site, use Google Maps in your app, or have your phone turn your lights off and on, there’s an API for that. Sites like IFTTT even allow users to create “recipes” between various apps and IoT devices that are interoperable with one another.
If you take a close look at the websites you visit every day, you'll start to notice subtle, visible API's at work. Many sites let you create a profile or log in via Twitter via their API, for example. Most are designed to be invisible, orchestrating relationships and data flows on the back-end as B2B technology services and platforms. A notable API platform that we have explored with Santa is Microsoft's Cognitive Services suite. Another set of API's of note are AT&T's, which help integrate everything from IoT to DIRECTV.
Last year, two researchers at the Center for Global Enterprise conducted an incredible study cataloging and visualizing this ecosystem, which by their count numbered more than 11,000 APIs, each one capable, with some programming, of speaking with the others. You can learn about their project here.
AI, Bots And Natural Language Processing
People talk to people. Software talks to software. Imagine two groups of people at a party speaking different languages at different ends of the room. Suddenly, everyone is able to speak the same language. The entire structure of the evening changes in a split second.
The scope and scale of artificial intelligence (AI) spreads far outside the topic of this post, but it’s safe to say that the world’s largest companies are placing some very large bets on AI being the next big thing that creates order and simplicity for just about every one of our daily activities.
IBM has Watson.
Microsoft has Cortana.
Apple has Siri.
Amazon has Alexa.
Google has Google Now.
Siri’s sister, Viv, has her own company.
And Facebook is offering brands a new way to chat with customers.
The bots of yesteryear were powered by relatively basic programs. Today’s bots can run on the vast depths of systems contained in frameworks built on continuously evolving machine learning. Some of these AI systems are "exposed" with API's - meaning that they have interfaces we can use to sample some of their early features in other software.
These APIs do more than detect intent from casual language. For example, they can determine sentiment—in other words, they can tell how a person is feeling by the tone of their typing and respond more appropriately. They can also be used to communicate commands through APIs to other software, whether it’s the embedded OS of a garage door opener or the product search interface for Amazon.
Voice Interfaces vs Screen Interfaces
Which brings us to the most novel trend, the voice interface. Most of us who have tried talking to our phone or laptop have likely noticed that the technology still has a bit of a hearing problem. But just as Atari graphics turned into 3D models that are becoming indistinguishable from reality in videogames, so too will voice interfaces gradually become as natural as the real thing as technology learns to not only parse our intention from casual typed language, but to distinguish between our various local dialects and individual speech patterns as we talk out loud.
WIRED just published a cheeky IQ comparison between two leading platforms for voice interface systems that are intended to be the new digital hub of the home, Google Home and Amazon Echo. I've only used Echo, but after about 5 minutes of use the novelty, amazement and humor of its jokes gave way to the realization that it's pretty much the same thing as having an invisible butler for anything connected to the Internet.
The power of API's, natural language interfaces and cognitive processing let you take a call, change the music, dim the lights, order a movie - or even be reminded about an anniversary - without ever looking at a screen.
So Where Do These Trends Take Us In Ecommerce?
When we can "talk to the Internet" in the same way we talk to a human, the impact of this revolution isn't limited to one or two new "cool things" that need to be integrated into an ecommerce platform to stay competitive. As these trends continue to snowball, the early adopter technologies like Echo and simple use cases for Cortana will evolve into the normalization of talking to "The Internet" which includes everything from the Internet of Things (your toaster or 3D printer) to your bank, or even healthcare assistance provided by Watson to your doctor about your cold (*with help from GSK).
Conversational Commerce means, in plain language, that we can buy things just by talking to our devices. Devices, and their software, will be smart enough to not only complete the transaction, but to satisfy all of our questions with the kind of responses we would get from talking to a subject matter expert on the topic, since they essentially have access to the sum total of human knowledge, and the processing power to sort through it.
While basic systems are here now, working and usable, the "perfect" use cases are a ways away. Gartner’s Hype Cycle breaks the overall trend into several technologies, suggesting that Natural Language Question Answering is 2–5 years from mainstream adoption, while Conversational User Interfaces might be 5–10 years out.
But tools like this are already everywhere, their seeds are built into Windows 10 and iPhones. One interesting case is Alaska Air’s “Ask Jenn” feature. While “she” can’t book your flight, she can help you navigate the site to find what you’re looking for. A feature like this is easy to continuously improve without presentation layer modifications, as the enhancements are all code-driven and have little impact on the visual layout.
It’s safe to assume that “Jenn” will evolve into a system that you can just talk to in order to book the flight you want. Through the power of APIs, you might do it through an Amazon Echo.
Imagine your family sitting down for dinner, while the Echo’s Alexa Voice Service plays music in the background. The conversation turns to finally taking that trip to Disneyland. You tell Alexa to turn the music down and search for flights. As the dinner progresses, you and your family ask Alexa questions about various options, ultimately settling on an itinerary, which you may or may not want to review as an email before taking the plunge and booking your hotel, flight, and other experiences without clicking a mouse.
Amazon's Echo already lets you complete transactions like ordering a pizza from Domino's or summoning an Uber to your house.
In a B2B use case, you might have a conversation with a future device in your office about options for optimizing a new manufacturing process. The conversation might start with your intelligent agent before breaking into a conversation with two "agents" from different companies, one powered by Watson, one powered by Microsoft's Cognitive Services.
In the short term, it might just look like a chat conversation on Skype with a basic FAQ "bot" that can access all of the engineering data for the product, and present it back in a way that saves you 30 minutes of searching.
But what about Santa? In a bit of watercooler conversation, the question came up, "If online shopping is replacing the mall, where does the mall Santa go?"
It’s a silly question, but it sparked one of those conversations that took on a life of its own. What if Santa was an artificial intelligence? Would he have some kind of digital manifestation? ...Could Santa's Magic be, if not replaced, at least supplemented by today's emerging technologies? Is there a set of conditions in which Santa could go from (earmuffs for kids) myth-real to "real-real" with the power of tomorrow's system of connected technologies, powered by AI?
Pretty soon we found ourselves white boarding dozens of approaches for a first-state AI Santa, including a charitable notion with our friends over at Giftagram.com that proved too complex to execute in time. Maybe next year. Lots of "next year" ideas.
Given that SMITH has 15 years of history working with Microsoft (for a while, we took over Microsoft Commerce Server), we have a natural bias toward their wizardry. We explored both the Watson API and Microsoft’s Cognitive Services, and decided to get the sleigh going with Santa on Microsoft’s Bot Framework with the LUIS API and Amazon’s Marketing APIs.
It will be touch and go if we decide to share anything before the holidays. As we explore the nuances of Natural Language Interfaces, the most common challenge is in imagining all of the edge cases that a human can create in an unstructured interface. Instead of the elegant binary options of buttons, we're faced with parsing the giant array of inputs driven by language, with every nuance, into a specific intent.
One thing is certain, as our disparate industries bridge the gap between "imagined and possible" and "live and bug free," someone is going to make a lot of money selling coffee.
Follow along for updates on this project by following us on Twitter.
Sean is a Senior Content Strategist at SMITH.