AI Amy and the Customer Experience
Dennis Mortensen, CEO of x.ai
Virtual assistants supported by artificial intelligence have been high profile over the last few years, buoyed by growing presence of Amazon's Alexa, Apple's Siri, and Microsoft's Cortana. But these assistants are limited to a device, screen, or software. New access options to artificial intelligence services are growing.
Among the intriguing choices is Amy, a virtual personal assistant that schedules meetings without the need of an app or mobile device. Users send an email to initiate the algorithm that finds the right meeting time and handles the back and forth communication. x.ai, a New York City startup, launched Amy in beta service and is expanding options this fall.
To learn a bit more about Amy I interviewed Dennis Mortensen, x.ai's CEO and co-founder. He is a highly regarded analytics and big data pioneer, having successfully delivered a number of company exits related to the space. He is also an accredited Associate Analytics Instructor at the University of British Columbia. He is author of Web Analytics: Tracking, Reporting, and Analyzing for Data-Driven Insights (Wiley, 2009), and a frequent speaker on analytics and data.
The idea for Amy grew from Mortensen's own experiences a few years ago when he scheduled over a thousand meetings—most of which required updates or rescheduling at least once. During our interview Mortensen described how AI is better forrepetitive tasks than current apps. “Managing your calendar is a really poor mobile experience. Handing over that task to Amy is such a wonderful customer experience, and it quickly shows up the limits of any mobile app for scheduling and calendar management. A desktop experience, where you have full screen view of your calendar, might give you the illusion that you can/should do this yourself. But this is really a task better done by an AI agent.“
What influence has customer experience—the product of an interaction between an organization and a customer - had on the development of Amy?
It's everything. Our whole product (and company) is the dialogue—and thus experience—between host, guests and Amy. That's why we've invested so much in developing Amy's voice. We created the role of the AI Interaction Designer precisely to do this, to thoughtfully craft these dialogues so that communicating with Amy (or her twin brother Andrew) feels as natural and seamless as communicating with a human assistant.
Storytelling is often mentioned as an important part of establishing a customer experience. Do you see marketers doing that with the AI options available today?
Yes. For us, though, storytelling is a bit more subtle. You can think of it as investing a lot in Amy's “back story.” Early on we decided that we needed to humanize Amy. So we gave her a name (first and last!), and we put some real time into identifying her key character traits. You'll notice she even has a LinkedIn profile. We did all of this for two reasons: We want both our hosts and guests to interact with Amy as they would a human assistant, which makes these interactions feel almost magically easy. And this has the added benefit of giving us good data on which to train Amy and make her smarter. We don't believe people should have to use machine syntax to operate an intelligent agent, and if they attempt to, then their weird phrasings mess up our data set. So story actually connects to data science for us.
Where do humans play a role in this, for example by intervention?
You first need to understand where humans can and should intervene. For us, we use human AI trainers to verify data annotations made by the machine (this is a type of machine learning called "supervised learning"). All of Amy's responses to customers and their guests are machine generated. Once Amy has mastered a skill—say, understanding that a person is running late—we don't need human AI trainers to verify those types of annotations any more. Instead, they can teach her new skills, like speaking another language. We believe this is the only way to create a fully autonomous AI agent for this particular vertical.
Other companies use AI to make humans more efficient; in these cases, humans are always in the loop, which is a different way of deploying AI. There is plenty of room for people to participate in marketing and analytics processes. But AI should remove all those little chores you have to do to actually get to do what humans do best—which is creating great stories and connecting with customers directly.
You've been offering x.ai as a free beta. What features and benefits can consumers look forward in the upcoming expansion?
We'll be launching three editions this fall: Personal will be free and lets you schedule up to five meetings every month. Professional lets you schedule unlimited meetings, personalize your signature, and create a list of VIP contacts (people who can have Amy set up a meeting with you directly, without you verifying the meeting request first).
Besides all the features of Professional, our Business edition will let you place your AI assistant on your own domain and have a team admin interface. Super exciting stuff for us after being in beta for more than two years.