WORKPLACE CONVERSATION GETS SMARTER
(Dennis Mortensen, CEO of X.ai, on the challenges and opportunities of voice-powered technology in the office)
"I think it'll reach a point where you'll be confused if you can't just
ask any question out loud at home like, 'How old is Barack Obama?' or
'What date is Boxing Day?'
That practice will soon become so normal that you'll be confused when
you walk into someone else's apartment and can’t just ask a question
aloud and receive an answer. I think that's when these products will
move into the business realm."
Interview with Dennis Mortensen, CEO
of X.ai
Next time you feel like teasing your
aunt for asking Alexa what time it is, consider: the smart speaker has emerged
as the fastest-selling tech product since the smartphone. Less than four years
after the Amazon Echo launched, an estimated 20% of all American adults have a smart speaker at home. By
2020, smart speakers will have a user base of 76.5 million in the U.S.
alone, according to a forecast by eMarketer.
Just as “BYOD” marked a
culture-changing technology migration from the consumer world to the office
with smartphones over a decade ago, the rapid home adoption of smart speakers
puts them on a similar crash course with their next destination—the workplace.
Only this time around, you probably
won’t need to bring one to work. Employers are investing not just in smart
speakers for work and productivity applications, but in voice-enabled AI
applications that will allow employees to interact with virtual assistants to
get work done without touching a keyboard or even a touchscreen. It’s not just
a smart-speaker revolution, of course—it’s about the advancement of natural
language processing (NLP) capabilities that can be embedded into everything
from speakers and AI chatbots to mobile apps.
Voice assistants have drawn big players
to help grow the marketplace. Amazon, for one, jumped into the fray by
launching Alexa for Business last November. Suggested uses include placing shared
speakers near printers to order ink and using personal speakers to set up
meetings, start calls and retrieve information from popular customer
relationship management apps like Salesforce.
But to become really smart, AIs such
as Alexa need skills—the smart-speaker equivalent of the apps on an iPhone. In
industry parlance, Alexa is “horizontal,” a generalist AI that mostly handles
simple commands. Complex commands go to “vertical” programs specialized for one
function. If you say “Alexa, get me a meeting with Phil in accounting next week”—a
complex command that requires using your calendar, knowing who Phil is and
being able to talk to him—then Alexa needs to call upon a vertical skill
meeting scheduler.
But that just scratches the surface
of virtual assistants’ potential. Already, companies are putting the technology
to work with AI assistants and chatbots for sales and lead generation, customer
service queries, product configuration and other needs to boost productivity,
reduce costs or improve the experiences of customers and employees.
For insights on the future of the
technology, we spoke with Dennis Mortensen, CEO of New York City–based X.ai, which has created a virtual assistant (called either Amy
or Andrew) that sets up meetings while handling all the mundane details: It
considers constraints on times and locations and negotiates the best solution
for all parties.
x.ai
While that may not sound all that
groundbreaking, it hints at important productivity gains these assistants can
bring to the workplace, especially as they’re beginning to multitask. Setting
up a meeting for you is just one example of that, but Mortensen sees hundreds
of other voice applications unfolding in the coming years that will all boost
productivity.
What’s the biggest advantage voice
assistants can bring to the workplace?
They can accomplish complex
objectives, not just perform individual tasks. A new user interface (UI), at
least conceptually, doesn't change any of the software that we've been
interacting with up to this point. The reason I think there will be an impact
is that the conversational UI will be so good at fielding full-on objectives.
For instance, if you jump into a
spreadsheet on your phone or desktop, you're doing a micro-task, like changing
a cell or doing a sum. An objective is something like putting together a report
covering two quarters worth of business. That's not a task you can achieve
today in Excel.
How do we get to the point where
workplace AIs begin fielding those complex “full-on objectives,” as you put it?
I think there are two real tech
barriers to achieving that kind of advance. One is a bit unsexy, because it’s
education. You can install Photoshop and figure it out if I tell you to do
something because you're used to the graphic UI. But in the voice setting,
which is completely open-ended, it takes a different type of education. We've
certainly seen that. [Our customers] are asking us, “How do I stop Amy or
cancel a meeting?” So, I'm thinking, You could use the exact words you just
wrote to me: “Amy, stop the meeting.”
Before you outline the second
challenge, why is it that users are struggling with figuring out that command
on their own?
Because we haven't trained people to
use any piece of software in the conversational UI universe. The only reason I
feel comfortable as the head of a startup that’s putting out a product that’s
ahead of the [UI] learning curve is that Google and Amazon and Samsung are
mass-educating the market and doing so very aggressively. That kind of
education work is the first real challenge [to getting voice assistants into
the workplace], but it’s already happening to an extent.
The next biggest challenge on the
tech end comes in two packages. One is ambiguity. You or I believe we're
crystal clear in our requests because in our heads, we know exactly what we
want. But there's a lot of ambiguity in how we express what we want and so
recognizing that ambiguity and clearing it up is a dramatic challenge.
Next, if I’m the AI, my universe
exists as a pool of intents and a set of entities. Intents like “I'm running
late,” “I'm mandatory,” “You're optional,” or entities like “He's the
administrator.” Trying to define those intents so there's no element of that
universe that’s muddled or misunderstood is actually quite difficult. Take a
self-driving car; it can't make a complete model of the real world. That's way
too complex. So you have to make a simplistic model for it to understand, but
what does that consist of? Roads, signs, weather, animals? Defining all of
those things is really hard.
We all want a good meeting
scheduler. But there are so many other higher-level needs in the workplace. Are
we years away from more sophisticated applications?
My bet is that you'll see tens of
thousands of very specialized vertical AIs that will each do one job really
well—and all of them working side by side will have an exponential effect.
When Apple launched the iPhone, they
believed it could come pre-installed with all the apps you needed. Apple had
effectively infinite capital, they could have tried to build all those apps.
But of course they wouldn't know what kind of apps each particular person would
even need. So, they needed 2 million apps from outside developers to satisfy
the market.
You can see Amazon making a similar
bet with voice. They want to be the horizontal, enabler AI, but then they have
the skill store. In every practical sense, Amy [X.ai’s virtual assistant] would
be just one app in that store.
Your assistant does part of what has
traditionally been an administrative job. People of course are worried that,
with the rise of AI assistants, lots of jobs for humans will disappear. Is that
a valid worry?
There are some agents that will do
simple jobs that used to be a luxury service. Nobody has a driver today, a
personal driver, and that’s where self-driving cars will come in. [X.ai] is in
the same place—not many people have human assistants. Less than 0.1% of
meetings are set up that way. So, we’re democratizing what people couldn't have
before.
There are plenty of chores at work,
and I think those things are the first things we'll attack. After that, we can
talk about what happens if [AI assistants] become creative and can do the jobs
people enjoy doing. Right now, I think that's far into the future.
Where are we now with these
technologies in the transition from home to workplace?
I think the dramatic change I just
suggested is that the home products—Alexa, Google Home—are built on the idea of
the conversational user interface. There are no buttons, no screens, you have
to speak to them. I think it'll reach a point where you'll be confused if you
can't just ask any question out loud at home like, “How old is Barack Obama?”
or “What date is Boxing Day?” That practice will soon become so normal that
you'll be confused when you walk into someone else's apartment and can’t just
ask a question aloud and receive an answer. I think that's when these products
will move into the business realm. I think we need to normalize it in the home
setting to the point where it's so normal that you expect it to exist in the
business setting as well.
There used to be a time, 20 years
ago, when you could be a rebel just by bringing your own laptop to work, or 10
years ago, by bringing in your own smartphone. But getting a device and putting
it in the meeting room will be different. Rebels probably won't cut it, unless
these devices become so cheap that you can buy them by the half-dozen, kind of
like Amazon sells the Echo Dot.
Now, of course, offices are
different from homes. I certainly believe that, today, it's much easier to roll
out a text-based, conversational UI solution in the workplace versus a
voice-based solution, and I do think a workplace assistant will be a mixture.
People imagine someone shouting across the [office] when they envision using a
virtual assistant, but I think it’s more like just extending a conversation.
You and I could walk into any
open-plan office right now and there will be multiple conversations going on.
That's acceptable. I think this will be the same—I walk over to another table
and talk about something. I walk away and I don't need to do a Post-it note or
take out my phone, but rather I just say, “Alexa, add a reminder for next
Tuesday to talk about that feature.”
In the home market, a report by Activate found that 65% of smart speaker owners hadn't
installed any third-party skills—the vertical AIs—at all. Won't that be a
problem in business, too?
Most people use their Alexa or
Google device for three or four things. They check the weather, set timers and
play music. But in the office environment, most software is being evaluated by
experts, decided on, then rolled out internally with real training. That's where,
if you find the strong use case, it could be forced upon the employees, like,
“The only way you can get a meeting room booked is if you use this technology.”
That tends to turn a practice into a habit.
Do you think this workplace AI
assistant space will see a few companies battle for dominance, as happened with
smartphones?
I think it’s unlikely you'll see a
plethora of horizontal AIs like Alexa or Google Home. The horizontal AI can
answer most of your rudimentary questions and integrate with most of the vertical
skills. You have three or four main competitors today: Amazon, Google and
Apple, and Microsoft to some degree. It’s the explosion of vertical
applications beneath them that will drive everything forward.
CREDIT: vm/iStock
About the Author
Forbes Insights is the strategic research and thought
leadership practice of Forbes Media. By leveraging proprietary databases of
senior-level executives in the Forbes community, Forbes Insights conducts
research on a wide range of topics to position brands as thought leaders and
drive stakeholder engagement.
No comments:
Post a Comment