Start time, so I'm
going to get started. Thank you all so much
for joining this webinar. My name is Mimi An, and
I run HubSpot Research. Today's webinar is presented
by HubSpot Academy and the IBM Learning Lab. These are both great resources
for continuing education that you should definitely
check out after this webinar. So as a reminder
for folks online, we are recording this
webinar, and you will all be sent this link later today. So to start, we have two amazing
panelists, Alyssa Simpson and Scott Litman. And I'm going to
pass it to Alyssa to introduce herself briefly. Alyssa? Hi, Alyssa Simpson, Director
of Product Management at IBM Watson. My portfolio covers what we
call the sensory services, so teaching Watson to
see, and hear, speak, and the emotional portfolio
of feeling and understanding emotional intelligence. Thank you, Alyssa. Scott? This is Scott Litman. I'm the co-founder of Equals 3. We are a very close business
partner to IBM, and in my past, I've had the great opportunity
to build marketing services and ad tech companies that have
had a great deal of success, working with some of the
world's biggest marketers on how digital technology is
transforming their business.
Excellent. Thanks so much for joining. We're so happy to have you two. So let's start with an easy
question, and by saying easy, It's not easy. How would you describe AI? And let's keep it
brief because we're going to dig, obviously, deeper
into this in the next hour. But when you're talking
to someone, and you say, I work for Watson,
I work for Equals 3, and we dabble in AI,
how do you explain the concept to someone who
may not be familiar with it? So from my end,
what I look at is different than
traditional computing. There are two attributes
we really focus on. One is the ability to look
at immense amounts of content and understand it in a
fairly human-like way. I mean, truly, the computers
don't understand it as a human does. But different than
traditional computing, cognitive allows
that understanding.
And the other is the learning
nature of these platforms. These cognitive
solutions learn by usage. The more you use them,
the more you train them, the smarter they become. And I look at that as these two
big differences versus, say, traditional computing. Yeah, that's really
well-aligned with how we look at it at Watson. We take it a little
bit of a step further– that basis of understanding
large amounts of data– excuse me– reasoning
and understanding what that is in the context
of where that information is coming from; being able to
learn by interacting with humans as there's feedback from that;
and that interactive piece, so being able to
interact and partner with humans who are
providing that training data or making decisions from
that unstructured data. I love that. Because I'm not from that
space, I tend to say, it's about enabling you
to be a better marketer.
You make better decisions,
help you be faster and smarter, leveraging data
that you probably couldn't analyze
yourself because there's just tons and tons of
data being generated. But now I see that using
that term enablement can be really, really
confusing because it mixes up with all of these existing
technologies that are available for marketers today. So what are some
of the confusion that you've seen with marketers
around the topic of AI? We think robots. We think cars that drive
themselves, and a lot of people can't really grasp how
marketing can be affected by AI. So when you're speaking to
prospects and customers, what is the message
that you give? I think the first myth that
we like to dispel at Watson is that this is not magic. There's no magic ball
anywhere, and that AI, like anything else,
is a technology, and you can break it
down into small pieces to be used for a particular
application, such as marketing.
So I think the first
real hurdle to get folks who are unfamiliar with
the world over is that there's not a magic button. There's really exciting
companies like Equals 3 who have built
what appears to be magic and magic
buttons on top of a lot of different technologies. But at its core, these are
transactional technologies that can be applied in
specific applications, such as personalized
marketing and maybe Equals 3 can expand a little bit on
what you're doing there. But that's the myth we
like to dispel here. Yeah, that's the–
actually, that's a great setup because we
do run into that problem, where people see a demo, they
see it all work, and it's like, oh, it just magically works.
And so one of the things
we've had to educate people on is we have found there's
a significant difference between a cognitive project
and a traditional one. In a traditional world,
at least for us, IT does all the development,
all the back-end work, and the business user
waits for the result. But here, because
it isn't magic, we actually can implement
Lucy very quickly, but then it's incumbent
on the business user to train their new associate,
their new companion, and they actually have to
spend a fair bit of time in training and mentorship
so their companion can be effective. And so rather than
a finished system and then waiting
for IT to do more, it's actually on the business
user to do the training and really do a lot of the work. Interesting, so out
of the box, you're not going to get this
magical assistant that can automatically sync with
all of the data sources that you have and give you
the recommendations you need. You need a little bit of time
to work with the system itself because while it's smart,
it's not that smart yet.
I want to assume, Scott? Do you have anything
to add to that? Well, that's actually–
that will end up being a great segment,
segue into the demo because I can actually
show how this all works. Excellent, well
going back to Alyssa, tell us a little bit about
IBM Watson and the product portfolio that you manage. You've talked about things
like emotional elements that you teach Watson. Can you give us a little
bit more color on that because I think that's something
that we, as marketers, at least today, without touching the
product know very little about. Yeah, absolutely. So Watson is a fabulous
part of the broader IBM company, where we're like a very
well-funded startup within IBM. And so we're out on
the bleeding edge here, creating
technologies that are, at the platform level,
a series of APIs that each do discrete functions.
And each one of those
functions serve something a little different, but they
are basically like LEGO blocks, and companies like Equals 3
are stacking them together to solve a particular challenge
in a particular industry– so in this case, marketing. What a lot of people
don't necessarily realize about
Watson, since there's a lot of attention
and hype around AI, is that how much it's actually
already being used in reality, and you might not
know it because it's sort of like this– I had a friend at one point call
AI this digital toilet paper. It's something that you
don't necessarily know exists, but it's there. So major banks, major
insurance companies that you already
interact with, lots of companies, large and
small, are using this today. And they may or may
not be super clear that they're using
it behind the scenes to help automate some of
the customer interactions that they're serving you with.
We're really big into
transparency at Watson, and so we like to be really
clear and open around what data is being used to train, why. So in the emotional intelligence
space, as you were saying, that's a really fun and
exciting area for us. We have three
different technologies in that space today. We have tone analyzer, which
takes text and analyzes the tone of that conversation. So today, actually, we released
a major update in that space, focusing on the
customer care market.
So for example, if you're
interacting with a customer service agent, and
you are perhaps calling a company because
you're frustrated, likely, we can understand
that you're frustrated and help escalate that
conversation faster, or direct you to
the right place, and really be attuned to
what those emotions are in the customer care space. That might be a little
different from emotions that you're expressing
more generally on social media or other venues. AI– the emotion stuff
is really interesting, and because it's not language
that we, as humans, often talk about, and
reasonable humans can disagree around what
emotion is contained in a particular phrase. Emotion is a very multifaceted
way of expressing yourself. You have your facial expression. You have your tone of voice. You have the actual words
that you're saying themselves. You have the context
in which you're coming into the situation. Are you on the phone? Are you on social media? What is that background that's
gone on to that interaction that you're having? What do I know about you? But you may look happy today– Alyssa is pleasant,
or she's not angry– but maybe I am angry,
and I'm not telling you.
So it's a really tricky space. It's really exciting. We're really proud of
what we're doing there. And one I know that Equals
3 is using quite a bit is personality
insights, which is taking personalities and
understanding intrinsic natures about people. And when we say someone
has a really strong EQ, we mean they're good at reading
people and understanding. And so how can we build
suites of technologies that help understand
people better, and interact better, and apply
those towards delightful client experiences? Yeah, that's fascinating. The minute you talked
about scanning for tone, I was thinking of a certain
airline that issued an apology that nobody appreciated. I wonder what Watson
would have had to say if we had run that
text through the system. I wonder if that would
have gotten published. That's actually a
pretty classic use case around understanding
what people are saying. Or even if you've
written an email, and you might not be aware
that it comes across as really assertive.
So you can have–
there's companies that have built little widgets
that integrate with Watson and say, hey, this is
an aggressive email. Do you really want
to be aggressive? Here are some suggestions
of how you can alter things, so there's a lot of
exciting work going on. Yeah, and I know some email
marketers would probably love to get their hands on that type
of technology and make sure that their emails are effective
in conveying the tone that they're– Exactly, that matches
with your brand and that matches
with what you're trying to communicate
because not everyone has a good third party,
independent editor to review what they're saying
and how it comes across, in the same way
that me, as a human, I'm always interested in how
people are responding to what I'm saying and how
I'm communicating it. And I might think that I'm
being pleasant, and open, and conscientious, but I
may come across as abrasive, and assertive, and
different from how I'm interested in
communicating my emotions.
So fascinating. Scott, so Alyssa
set you up, too. Tell us a little bit more
about Lucy or your product. What is it? What can it do? Show us a little bit about
what you're working on today. So Lucy's the
cognitive companion to the marketing professional. She's built for the Fortune
1,000 and the agencies that serve them. The problem that we
set out to solve, working with IBM and
Watson, was really the idea that marketers have
so much content in so many different systems.
They have the content that
they on their own databases, marketing analytics, website
analytics, media data. They have third-party data, like
Forrester, eMarketer, Kantar, and others, and they have
all their own own documents, all of the PowerPoints,
PDFs, and the like. And if you're OK,
should I bring up Lucy? Totally. So I'll just turn
on screen sharing. Yeah, I love seeing real-life
applications of what AI can do, especially, again, in a
marketing space because even to me– I've done a lot of
research into it. It's just this cloudy
topic where I don't really know what it is. So to see an actual example
is always super valuable, so I hope that's the same
case for the audience that we have here. Absolutely. So this is Lucy. She's a software as a service. She lives in IBM's
Bluemix environment, and she has three
major components– research, audience
persona modeling, building really tight persona
models along the lines of what Alyssa was just talking
about, and then helping with media planning.
And so what I'm going to
show here is research, and you can see I've
asked a question of Lucy. What is the latest information
on self-driving cars? In this instance,
I've got a demo that's around automotive
marketing and Tesla specifically. Lucy gets trained around
the data of the company that hires her, so we have to be
pretty specific in that regard. So here's what she's done. When I asked the question,
what is the latest information on self-driving cars? She comes up with a
list of responses. These can come from databases. They can come from PowerPoints,
and PDFs, and documents in our file systems,
or it can come from third-party relationships
like the eMarketers, and Forresters, and the like.
So she's showing the
list of responses, and you can see on the
left her confidence score. Her confidence is based on
her natural understanding of language, which
comes from Watson, as well as the training that
is given to her by the company that's hired her. So here, I can see her
responses to this question on self-driving cars
with a 94% confidence. Now, keep in mind this
is a trained Lucy. She has found some
information from eMarketer on level of interest,
attitudes, and opinions about self-driving
cars, so great stuff. In the bottom
right, where it says "was this answer relevant?" I can say, yes, give
Lucy four stars. That's the training that edifies
and gives her confidence. And as I go through this,
I can see other examples. I see more information
from eMarketer. I can see yet additional
reports from eMarketer, and as I go through
this, she's giving us the components of
eMarketer reports that she thinks best
answer the question. Just below her confidence
in those eMarketer reports, she has some great
data from Statista. So throughout, I can be
grading these responses, saying how she did a great job.
That impacts her confidence. If I see something I like,
I can save it to a project. So by clicking on the star
here, I can pick a project and save this to it. And so that's how we end
up interacting with Lucy. We ask a natural
language question. She goes through all the
data that's available to her, shows her confidence, and
gives you her best responses. Now, another example
is I want to find a SWOT analysis for Tesla. So I'm going to ask, do you
have a SWOT analysis for Tesla? Now, in a world
without Lucy, what would happen is I would think,
I'm in a marketing department. We have dozens or hundreds of
people here, and I might say, I know somewhere we
created this, but where? And I might post to an
internal social network, like a Facebook for
Business or to a chatter. I might email around. I might not find cube walls. But the chances of my finding
such a specific component as this within the
thousands of documents that are in an enterprise
is very, very difficult.
I'm more likely to recreate
it than anything else. But here I ask, do you have
a SWOT analysis for Tesla? I just as easily
could have said, what are the
strengths for Tesla? What are the weaknesses? What are the threats? And Lucy found it. So where did she find this? And in the bottom
left, I see the source. In the eMarketer and
Statista reports, that source would have
taken me to my subscription. In this case, the source
is going to take me to the specific file. And so when I click
on Source, you can see Lucy's
downloading a file. It's a PDF. It's a 46-page PDF that is
like so many documents that are within an enterprise,
where a singular document could answer dozens of
different questions.
So as I scroll through
this, this is a document that Lucy read, but she
answered with the precision of the specific answer
that's within this document. So as I eventually get to page
19, I see that SWOT analysis, and so it's not
just Lucy saying, here's a series of documents
or here's the documents. Here's the component
in that document, and then I can save that
component for later reuse.
Now, a lot of what Watson
has been known for, when you saw Watson on Jeopardy,
was the amazing ability to look through huge
amounts of content, text, and we think of that
as unstructured data. So these initial examples,
either my own documents or the license documents from
eMarketer and other sources, are examples of
unstructured content. The thing is, marketers need
to work with data that's structured and unstructured. They need to ask questions
of their marketing automation platforms like HubSpot. They need to be able to
ask questions of databases like Google Analytics, or
Omniture, or other website analytics.
They need to ask
questions of media data from sources like Comscore,
or Kantar, or Nielsen. So here I'm going to
ask a question, which is, how much did BMW
spend by month last year? This is an example of a
natural language query that's going to go against a database. Without Lucy, I would have to
go into a platform like Kantar, or Nielsen, or Comscore
to ask this question. I'd have to be trained
on how to write scripts or how to do
reporting, but here, we bring a natural
language interface to this source of data.
So here we're working
with Kantar data. You can see the data that we
connected via API to Kantar and extracted to
answer the question, and you can see the
visualizations of this data that we're able to provide. So if I ask something like, who
are the competitors for BMW? that's another question that
can be answered from data that exists at Kantar. And here we see the competitors. If I want to ask
a question like, how much did BMW spend versus,
say, Jaguar versus Audi and versus– we'll add Ford, and
Kia, and some others– by month last year? You can see that she
will go out to Kantar and formulate this question,
come up with a response.
And it's really
pretty amazing how we can work with various
structured sources of data all through this natural
language interface, and Lucy works with
this very quickly to give us what can be some
fairly complex reporting and brilliant. So there's one other thing
I wanted to show you. And you brought up
United Airlines, and you brought
up, what could we learn from checking things
like tone from messaging? And so one of the things
we're doing with Lucy is we are combining
multiple sources– news sources and social all
together– in one component.
So an initial example,
this is brand insights, and what Lucy does
is she is reading through roughly a million
pages of content a day. And here we've looked at United
Airlines over the last 10 days, so we're looking at about 10
million pages of news content coming from common news sources,
like Washington Post, New York Times, CNN, Reuters,
and 1,000 others. And what Lucy is
doing is she's saying the sentiment in articles about
United Airlines is really, really negative,
74% to the negative, only 12% to the positive.
There is a ton of content here
that Lucy has gone through. You can see, under the
sources and articles, the volumes of mentions. So New York Times has written
about United Airlines 33 times in the last 10 days. And if I click on
this, I can see which articles were considered
negative or positive. We're looking at that
tonality per article, and so we can easily
go through the list. If I want to see, when
did it get bad for United? I can click on United Airlines
the brand under the topics, and I can see there were
500 negative mentions on the 10th, 1,000 bad ones on
the 11th, 900 more on the 12th. This is just a crisis
that's just been– you can see that bubble when
the news was just so bad.
We can see hashtag analysis. We can see image
associations, and you see what was a United customer
being dragged out of the plane. So all of this is being combined
by bringing social and news sources together into one place. And by the way, we
also compare how sentiment runs in news, which is
74% negative, and social, which isn't quite as harsh, which
is a little surprising. In any case, what
you're seeing are the research components
of Lucy, the ability to ask natural
language questions against unstructured
content that's licensed, like eMarketer, ask questions
against your own data, like the PDFs, ask
questions against databases, like the Kantar
database, as well as how we're able to use
Watson's ability for measuring tone and sentiment to look
at huge amounts of content and to do things
like brand insights.
So Lucy has all kinds of other
features we'd love to show, but this gives you
a good idea of how we're working with those
core Watson components into a package solution
for marketers in Lucy. That is so fascinating,
especially the middle piece, where you talked about
being able to scan assets. I know that that's a pain
that we even feel at HubSpot.
We're not perfect. We have lots and lots
of PDFs, and files, and we have it in our
internal wiki system, and it gets lost totally. Being able to search
our own archives like that would be just– I want it. Sign me up. [LAUGHTER] Even so, I wanted to go back
to the first example, where you were training Lucy by
simply giving it a star rating. I think that's just a
wonderful visual testament to how simple it can be because
I think a lot of people, when they think about AI,
they think about having to dump in a lot of
data to train it right, and then complicated
algorithms get spit out, and you have to have a
PhD to really navigate your way through the
system to make it do the thing that you want to do.
But these overlays of the
technology that you're building, what IBM
is enabling, it can be as simple as the
Netflix thumbs up, thumbs down, you're going in the
right direction. It could be as simple as that
to teach or train your AI system to do what
you need it to do, which is a wonderful example. I have no question for that. I just wanted to point that
out, that that's fabulous. Cool. All right, so thanks, Scott,
for the demo, super, super interesting. So from the HubSpot's
perspective, we've been trying
to dabble in AI. I don't think we're as far
along as you two, obviously, but we wanted to definitely
share with our own customer base on what marketing
automation can become with the help of AI. We've got our own kind of
natural language processing bot, where we allow people to
dig into their CRM prospect, look for new leads, and
also create new blog posts, just bypassing our menu
navigation system completely. We can just have the bot say,
create a new blog post for me, and it'll pop out and
give you the link.
I think for a lot of folks,
though, the concept of AI is a little bit
frightening, and so do you think that with all of
these new technologies that are being built– I think AI will definitely
change our jobs, for sure, but do you think
that jobs will be gone, that we are
going to be obsolete, that the machines are
going to take over? Obviously, you can
tell from my tone that I have a bias
in how they answer, but I think it is
something that is part of every single
conversation nowadays that's around AI, so I'd love
to get your thoughts on that. So for us, the whole idea of the
name Equals 3 is about the idea that 1 plus 1 equals 3, that
better than the individual, or better than the machine,
are the two together. So I think that we
will see scenarios where people look at how they
can make changes to staff based on automation. We're certainly seeing
that in many industries.
I think the business
that complements the talented individual
with the AI companion will outperform those who don't
adopt or embrace AI at all, or those that rely too heavily
on the AI to do the job itself. And so we're pretty
bullish on the idea that this is all about
supplementing and enhancing the individual. Now, I think what's
going to happen is that we're going to see
more expected or demanded by the marketing department,
more expected and demanded of the agency, and that the way
they keep up with that is AI.
It's going to
enhance their service delivery and their
performance, but we look at it that more will be expected
and more will be achievable. People be able to drive better
results and better outcomes because of their
embracing of the AI, versus the
displacement of people. Yeah, I think that that's
100% with how IBM really comes to market and talks about this. We see this as man plus
machine, and Jenny's gone on about that many times. It's about the partnership here
between humans and cognitive technologies. We, actually, at
IBM, when we say I, we talk about
augmented intelligence, which is all about augmenting
what a human is already doing and extending that to
be able to do things that they could not have done before.
So one example in the marketing
space from another client working, actually, with IKEA,
a company called iTrend. IKEA was interested in
social media listening, similar to what
Scott just demoed, and they actually wanted to
understand– if you've ever put together an IKEA product,
it can be a challenge sometimes. And sometimes people
get frustrated, or they do really creative
things with a bookshelf, like turn it into a bed
that IKEA didn't necessarily anticipate or think of. And so they did a project
that, again, extended the reach of the
marketing team by looking at Google video, YouTube videos,
and understanding, visually speaking, where were the
IKEA products that they were particularly
interested in appearing in those videos
and then what was going on in the context
of those videos.
Was it positive? Was it negative? What did it associate
with in the product SKUs that actually sold? And so that was an
example where there were hundreds of thousands,
millions of videos. They couldn't possibly have done
that with their marketing team. It's something that,
from a human perspective, it's way too hard,
way too overwhelming. But if you can do that using AI
by training a visual classifier to understand visually
where are those products, and which ones look
like ones I sell, you can start to have
an intelligence that was not possible before. But that was only possible
because the humans trained the visual classifier to say,
hey, here's what it looks like. Here's what I want to see.
Can you go find that? Tell me where this exists. So that's just a good example
where the nature of the work may be shifting a
little bit, or one may have different
responsibilities than they did before, but
to Scott's point, the winning companies are
going to be the ones that embrace this idea of both. Excellent. So before I ask
the next question, I'm just going to pause and
let folks on the webinar know that you can ask
questions of our panelists by tweeting at HubSpot Academy. Use the hashtag #HubSpotWebinar,
and someone will come in with the questions, and
we'll be sure to answer them if we have any time left over. So if you have any burning
questions for the panelists, please do tweet
at us, and we will try to get them answered in
the remainder of the webinar. So I just have not
a personal question, but more about why
you two decided to– you started your own
company, Scott, around AI. What was the potential
that you saw? What motivated you to
get started with this? What caused you to
think, this is it, and I'm going to dabble in it.
I'm going to build it because
I see X amount of potential in return in what
I'm going to build. Yeah, so I've always been
fascinated with the AI space. And when IBM showed up on
Jeopardy, it was like, wow, and that's something
I just explored. It was super interesting. And then it was about two years
ago, it became clear to us that the Watson
platform was being made available to developers. And so I sat down with
my business partners and said, what could
we do if we had this? What is the problem
domain, where if we could apply everything
IBM has invested, the billions and years they've
put into developing this platform– if we
had it at our disposal, what could we do with it? And we thought about all the
marketing technology platforms we have stood up for
customers over the years and thought about
just how much data is in so many different platforms.
If we could find a way to
bring the power of Watson to all that data, whether it
was structured, unstructured, owned, or licensed,
what we do with that? And that became the impetus. And then we started to– we worked with IBM. They were great to work with. We loved the tech,
and we started to build out the MVP around
Lucy and bring it to some of our trusted contacts. They were blown away,
and that gave us the energy and excitement
that, let's go for it and make it happen. I think, for me, I
was really attracted to Watson and the
space generally because I see the
potential to change the world for the better. It's a cheesy answer,
but it's really– it's what gets me out
of bed in the morning is exactly what Scott and
many of our other customers are doing with the technology
and how they're applying it to a whole host of different
industries and business problems.
And it delights me to
hear our customer stories and to work with those customers
around how they're actually using this to make
something easier, or better, or delightful for
that end customer that really is exciting and something
they could not do before. So that's what gets me
out of bed in the morning, and I think it's certainly a
privilege to be in my position. Do you think that– so the reason why we at
HubSpot held this panel, even though we're not
really in the AI game is that we undertook
a lot of research into AI because we saw a
lot of potential, obviously, but a lot of confusion, at least
among our marketing audience.
People could tell that it was
something that was important. It may disrupt their
jobs, but they just didn't know what it even
was, and so we wanted to unpack that a little
bit, and we've been, just had the pleasure of
working with you to even develop this webinar to
educate our audience. I kind of asked you
this question before, but why are so many
professionals just not aware of what the
potential impacts of AU is? Is it because it's nascent,
and so the tools are still developing, and there's
not a lot of messaging? Is it because we've been
told in popular culture that this is what we
should expect AI to be, and so we have
this preset notion? There's so much
potential, there's so much interest, and
yet so little clarity. Why is it– Yeah. –the case? I think one of the challenges
that this space has is it's been a promise for
the last 50, 60 years.
Hollywood has promised
this magic future world, and so there's a lot
of pre-existing ideas around what it will be
or what it should be. And I think, in some ways,
this technology and this space in general is very nascent. We're just starting to see,
in the last five years, I would say, real
businesses use this in reality at
scale in production to really and truly
solve hard problems. But it's not new technology, but
there's a confluence of data, of hardware, of accessibility
of this technology that is new, and we're able to do things
that we were not able to do 10 years ago or 15 years ago. And so I think that's
one of the challenges that we face is re-educating
people around what's magic and what's now. And then I think
the other side of it is that one of the biggest
pieces that I think we really work with our
customers on, which is what Scott touched
on a minute ago, which is that he sat down,
when starting Equals 3, and thought about, what problem
do we want to apply this to? And that really is the hardest
problem with any technology.
I can sell you a knife,
or a MacBook, or anything, but it's just technology. The magic happens when
you apply it as a chef, and you create a
masterful dinner, or if you put a MacBook in
the hands of a iOS developer, and they create
wonderful mobile apps. Like anything, this
technology is a tool. It's a new tool,
new-ish, but it's really up to our customers
and the users of this to make that magic
happen and apply it to particular industries,
particular business problems. And so I think there's
a lot of people who– one, you have to understand how
the tool works and learn it, and so that can be a
challenge to understand, hey, here's what it does. Here's what it doesn't do. You cannot make breakfast
with your MacBook, but you might be able to
make dinner with a knife. So you need to understand
the limitations of the tools and what they do, and then
you have to think about, hey, am I going to make
Vietnamese food tonight, or am I going to make Italian? You have to get specific
around what you're going to do with
that and how you're going to create a
delightful experience.
So I think that that
hurdle with AI is around, do I want to do customer
listening and social media? Do I want to optimize
my call center? Do I want to apply
this to medicine? Do I want to apply
this to health care? Do I want to cure cancer? How do you want
to take this tool and apply it to the problem
that you care about? And breaking down that
problem that you care about into its parts and pieces
can be a big challenge. I want to do social
media listening. I want to understand
everything anyone has ever said about my company
ever and magically have an insightful dashboard.
That's a lot of work, and Scott
and his team have proven that and really made that
easy, but there's a lot of smaller tools, and
smaller parts and pieces, that go into making
that magic happen. And so I think when people
get overwhelmed sometimes, it's because they're
trying to break down that problem into smaller and
smaller components of which AI can be applied to.
Yeah, I'll just add to
that a little bit, which is one of the practical
challenges for marketers is they've never
done this before. So I totally agree with
everything Alyssa said, but you apply that to,
yes, identify the problem. But if you've never bought it– for us, there are no RFPs
for cognitive agents. There's no marketing
departments that have pre-existing
budgets for, I'm going to put X dollars into AI. And so because of
that, you don't have people with the
experience of having run AI projects before. They haven't bought it before. They're not quite sure what
it's going to look like. So there's a huge level
of market education. Events like this are immensely
helpful because people can walk away and say, oh, I get it. I get at least that's
one facet of my job that could be impacted by AI. And as much as we all have AI
permeating our daily lives– Google is an
amazing tool for AI. It's no longer a search engine. Your Facebook news feed is
completely driven by AI. To some degree, people
use services like Siri and things like that.
So we have AI in our
day-to-day lives, but we haven't put
them, necessarily, into our business
lives in this way, and so it's new for people. So market education is
just a huge, huge element to get to that point
where you can say, what are the problems I could solve? Excellent, so I do have
a question that actually ties into what we
were just discussing, so I'm going to ask it. It's for Scott, and it has a
little to do with this, maybe, confusion around like what
AI enables versus something that exists today. So the question is, how is Lucy
different from other listening platforms? So the third example you
gave in your demo, a marketer is asking, well, that kind
of looks like Omniture.
It kind of looks
like something– Yeah. So what are the
nuances of your product that makes it more advanced? So here's the thing. If your whole life
is in social media, you're going to live in products
like Sprinklr, or Sismos, and you're going
to go a mile deep. If your life is as
a data scientist, you're going to use
a Domo, or a Tableau, or Watson analytics
type product. If you're in
marketing automation, you'll use one of
the marketing clouds. But to the VP, the product
manager, the strategist, the planner, somebody
who's omnichannel, somebody who either isn't
using all of their data, or they're sitting there
with 20 windows open at once, Lucy becomes amazingly
helpful to them because through one login,
through one natural language interface, they're able
to get at that data that would otherwise be in
Omniture and, perhaps, is only in the hands of very
few people in the organization.
They're able to
get to that Kantar data that would
otherwise only be in the hands of a few people. They'll be able to
fully utilize eMarketer because eMarketer's
got great content, but too often an enterprise
is not used as universally as it ought to be, or that'd
be true of Forrester and others as well. And so we're saying, through
one login, one natural language interface, I can query
dozens of different sources and have it all come together. Now, if my life is
only in one source, then the deeper tools are going
to be used, use that tool. If I'm a Kantar operator, and
I know how to write a script, and I know how to
do the reporting, I should just live in Kantar. But if I'm the agency
account director or the VP, and I just want to
know, how much did we spend versus our
three top competitors? And I don't want to ask the
decision science or the media team who is overworked,
I can just ask Lucy.
Got it. All right, last
question, and then we'll dig into some more of
the submitted questions. So where do you see AI heading
in the not-so-distant future? And how can we get
started in using AI today. I think we dabbled a
little bit in that piece, but tell us about what
you're excited about, new developments
that are coming.
So from our end,
and Alyssa probably got a broader perspective
on this being at IBM, but from our standpoint,
we see that it's going to permeate everything. I get so excited when we
connect to a new source, and Lucy learns it, and she
becomes smarter and smarter. And it's amazing to see how
that data gets stronger. The longer somebody
has an AI companion, the better it performs for them. And then the other part is
for us just product roadmap. We're constantly inventing
that "what's next." It's exciting to sit around with
the team, listen to customers, and get their feedback
on what they would want to see, and then make it real. Yeah, we're similar at Watson,
with a potentially little bit of a broader scope.
We're really excited about
the future of everything that we're bringing out. My teams have releases– I think we have three
releases this week, so we're constantly iterating
and releasing new stuff, and that's just my team. There's many others that
I work closely with, so we're developing at
a lightning-fast pace to keep up with market
demand for different features and functionality. As I mentioned today, the
customer care, tone models are just available. Last week, we released a
visual recognition tool around making training
easier, and we're coming up with some
more exciting stuff in the next couple of weeks.
I think more broadly,
though, there's a perception that this is hard, and it's
difficult to use and take advantage of. Something people
don't know about me is I don't have a background
in computer science. I have a liberal
arts degree, and I don't code on a regular
basis, but I use AI, and I can use these
developer tools. There's a 13-year-old, Tanmay,
who's gotten a lot of press with IBM, and he's always
the first to adopt whatever we put out there, even before,
in some prerelease beta, and he's 13 years old. But this stuff is– there's free versions
of all of it. It's easy to use, and if
it's not easy, call me. I'm not doing my job well. But the idea is that this
stuff is easy for developers of any skill level
to get started with, and there's certainly an
expertise and a training as you get more advanced
and more sophisticated with the tool that
you want to do, but at its most basic
level, these are APIs.
So if you can integrate an API– or even better,
some of our services have tooling on top of them. If your business
user, like me, you can log in and build a chat bot
using our conversation service. So as an example, I got
sick of people asking me the same question over
and over and over again about visual recognition, and
pricing, where to find docs, and everything else, and I was
like, Watson could handle this, and I built a little chat bot. And again, I'm a business user. I don't code, and I
was in my hotel room, and an hour later, I was
done, and I launched it.
And so I think that's– dispelling that myth
that this is hard is something that I've
tried to reiterate, and dispelling the myth
that it's expensive because these API cost
fractions of cent, and it's something that can– it's easy to get started with
and scale up as you grow. I feel like this is going
to be a natural question that some folks
are tweeting at us. Where can they find
these resources that you're talking
about, Alyssa? Oh, just go to Google. Go to IBMWatson.com. Get started with
a Bluemix account just like Equals 3 did
a couple of years ago, and start making API calls. But IBM Watson platform
is hosted within Bluemix. Accounts are free. Check it out. Great, all right,
so let's bring it back to marketing a little bit. I think you two are definitely
really well aware of what's available on the market. So there's some
questions around, all right, how can
AI help me deliver the right content to the right
person at the right time? What exists out there that
I can leverage to do that, and if at all? Who wants to take it? Ah, sure– oh, Scott,
do you want to go ahead? Well, I'll just say
that, as of the moment, Lucy's fairly unique
in what she does.
The ability to do both some of
the research things you saw, as well as the audience persona
modeling and the media planning capabilities within
a package solution are currently unique
for the most part. So my answer to the question
is, well, talk to us about Lucy. We have yet to have any
significant client interactions where they're evaluating
us head to head with another cognitive solution.
It's really the evaluations. Are they ready for cognitive? What are their use cases? What are their sources of data? What are the expectations
for how that data looks within a platform? So our point of view
in the marketplace is we certainly know
there will be competitors. There has to be. But at the moment, we haven't
seen a lot of that as of yet. I think certainly what
Equals 3 has built is unique and
unusual in the space. I think there are many
agencies that we're working with that are
doing components of this and using different APIs for
similar types of use cases around ad targeting
personalization as an example, or something that
are a little bit more on the fringes of
what Equals 3 does. So Ogilvy is one that has had
a lot of attention and case study, and what they've
done for the US Open or Wimbledon is a
well-documented examples of how they're delivering with
brands the right messaging for the right person
at the right time right with all of their sponsors and
many different brands involved.
Yeah, to that end,
agencies like Saatchi did some interesting work that
was put into the Cannes Film Festival that was driven. You've got agencies like Havas. Yeah, cost of rocket
fuel is another one. –that have developed
cognitive practices. But in general, the
cognitive practices are also creating bespoke
solutions versus providing packaged offerings. Got it. I think we were partnered
with Salesforce, and they have quite
a AI engine that's being built out for their
marketing and their CRM platforms. Yeah. And even on the HubSpot side,
we are seeing a lot of interest. And we're certainly working
on allowing our own customers to examine previous
emails they've sent, content that they've written,
channels that they've explored. And then, eventually,
our goal is to package that in a way, using
natural language processing, to allow people
to understand, OK, this was a successful
channel that we– where we pursued a prospect.
This is a great conversion lane
that we can enrich and enable in the future, using some
of these capabilities. I think there's definitely a
lot of marketing automation companies like HubSpot
where they're definitely feverishly working on it. It's such– Yeah, I think you mentioned
Salesforce specifically, and I'd be remiss if I
didn't bring up our– we announced recently a
really large partnership with Salesforce, and
many of those use cases are in the marketing
automation space. I think Salesforce is
really interested in around how do they take the
richness of the data that they have around
customers, and really use it to deliver
personalized messaging, and enable sales teams,
marketing teams to really delight that end customer. So we're really
excited about the work that Salesforce is doing
on integrating the Watson technologies into
their platform. Yeah. Well, and to that end,
a lot of the effort in AI and marketing, which
is not the area of our focus, is in that area of
cognitive engagement. How can I use
cognitive to optimize how I'm bidding on media? How can I optimize performance
at the e-commerce level or conversion of some kind? And there are a fair number
of solutions out there that are working in that space,
so very different than what we have been talking about
as far as research and things like that, but there's a
fair amount of energy there.
Certainly, that's a big area
that Einstein is focusing on in trying to drive optimization
of their Marketing Cloud and creating a better
one-to-one experiences for their customers, and they're
doing some very cool work there. Yeah, the ad buying
piece is really, really, I think, compelling
for a lot of marketers. Alyssa, I saw one
of your colleagues actually demoed a piece
of Watson that was just focused on ad
purchasings, optimizing, getting the right
type of channels, hitting the right people
at the right time, all powered through
the AI engine.
I think for a lot of
folks, advertising is the biggest
crapshoot for marketers because we just
don't know if we're hitting the audience
that we're hitting, the conversions are actually
going to generate the revenue. And I think tying AI
into that, the purchases that you make are the
smartest possible. And, hopefully, with all
the data sources coming in, you can tie it back
to an actual sale. That's super, super powerful,
and I think a lot of marketers today just don't have
that ability to track it and that type of details. Yeah, the Holy Grail
is the market– is the attribution and the
automation of all of that.
And I think, even looking
ahead around the attribution problem, which is often
a disparate data problem, it's also looking at the impact. So let's say you did
reach that customer, and they did make a
purchase, but how do they feel about that purchase? Was it a successful one? How are they feeling
about that product? Are they then encouraging
others to buy it? So there is more than
just, did it happen or not? Did I get that view? Did I get that click or not? But was that click meaningful? Was it positive? Was it– can you get further
than just a more wrote, it happened or not, it didn't.
Totally. So I think one more question. This is from someone
from Southeast Asia, so international
is at top of mind for this particular person. How is it being adopted
around the world. And specifically when
it comes to localization as people are going across
boundaries across geographies, across languages,
is AI something that can help us bridge that gap? Yeah, we're really focused
on that problem at IBM, and we have a huge amount
of resources and attention on solving that
internationalization problem.
IBM operates in 170
countries, I believe, and so we need to
have Watson understand not only the languages, but
the cultures and the context of our global client
footprint because– I'll go back to tone
and emotion, right? Those cultural norms impact how
do you understand and apply AI in different places. So one example that I use
there is something like color. For example, in Japan,
the notion of green is a concept that is
different than green in the United States. So a simple tag
like that around, hey, this tree is green,
that concept is different because a green is not a fact.
It's this abstract creation
that we have of color. And so how– when we
do global expansion, how do we not just simply
translate something into a different
language, but how do we make sure that
we are being aware of the cultures
and the learnings that we can create
context-specific, relevant AI solutions for that market? So it's a lot more
than just language. Scott, any thoughts? I think it's a great question. It's something that we're
mostly focused on really US. Although, that said,
Lucy takes in questions from dozens of
different languages, although providing
English-language responses. And we have plenty
of global customers, again, working
against English data. It's interesting
to think through, how do you start to compare
different cultural norms? How do you have content
from different geographies, and how do they compare equally
within that environment? And then IBM just has a much
better and bigger perspective on how to solve that because
they're immersed in it.
I think some of the other
challenges for expanding globally are around security
and data sovereignty laws. For example, we just opened
a data center in Frankfurt that we're really
excited about to serve our European customers,
and then we're also working with partners. So you mentioned Southeast Asia. We have SK, a big
partner in Korea, and we have other clients
who are partners of ours serving those end customers. Got it. Well, you kind of
mentioned it tangentially. The last question is, are
there any concerns of AI being hacked for trade
secrets, for data? How does IBM approach this? How do you, Scott? When you're building
up this product, you're compiling quite
a lot of inside data. What's the approach there? Yeah, we take security really,
really seriously at IBM. It's not trivial at all. We try to differentiate from
our competitors in the space, actually, on security and on the
approach that we take to data. So when you're– you reserve
the right, as our customers, not for IBM to store or
learn from your data.
You can use Watson without
us storing or using anything that you're sending to us. So that's the first big
way that we differentiate. And then we offer different
levels of separation. We offer our public cloud. We also offer premium
and dedicated options for different
security environments and what's appropriate,
given what type of data that you're looking
to do analysis on. So one example would
be our health care. IBM Watson Health is a
HIPAA-compliant, totally separate type of environment
than if you're just looking to understand
social media and someone posted this image.
What is this image of? It's a very different types
of data security requirements. Yeah, on our end,
the security side has to be supportive
of the enterprise, and so we have a couple of
really important tenants. For the agency
customers, Lucy needs to help them stay in
compliance with their MSAs to their end customers. If they've got multiple brands,
they've got internal firewalls, the right users can only see
the results of the content that they should have access to. And then for the data partners,
those that are the providers of third-party data, we need
to ensure that Lucy's helping our customers stay in compliance
with their third-party data right so that if you've got
10 named seats to source X, those 10 people in Lucy
will see those results, whereas the rest won't. And that ends up being
a very important part of how we've architected Lucy. The other thing is
partnering with IBM. Leveraging the
security they have for the data that is
within their environments has been really
critical because they've got world-class infrastructure
to support us in that.
Excellent. All right, so now this is
the real last question. Thank you, everyone, for
joining this session. My question is, when the first
self-driving car rolls out of the factory, are
you guys buying one? I'll say, I already
have a Tesla, and I use its autonomous driving
features a ton, and I love it. So it's not true self-driving,
but I got a lot of mileage– But what about without
steering wheel? I'm asking, when
there's no steering wheel at all in the car. How about you, Alyssa? I'm really excited about
self-driving car technology. I have a number of
friends with Teslas, and I think it's
really exciting. I think we're just
getting started. So I was in an accident,
a car accident, last week. I got a concussion, and
I was thinking to myself, oh, I can't wait till
it's self-driving and this doesn't happen
because it's human error today. I can't wait for it to drop
me off, and then drive home, and then pick me up so I
don't have to find parking.
That's– I'm excited for that. I'm an early adopter
of everything, but I think self-driving
cars have a way to go. We're not there yet. Yeah, it will take some time. I'm looking forward to it. All right, folks,
well, thank you so much for your time
and your insights. Alyssa, it's
amazing that you are so coherent after a concussion. Oh my goodness. But thank you again. I hope, for the
audience, that this was an insightful
and interesting topic of conversation. Continue it on Twitter, and
we'll see you next time. Thanks, both of you. Thank you. Thank you..
