How I: Use Google Prediction API to improve customer service Video Lecture | "How I" by Google Developers - IT & Software

[MUSIC PLAYING]
MANI DORAISAMY: My
name is Mani Doraisamy.
I'm the founder of Guesswork,
a machine learning startup.
We help CRM companies to
build predictive intelligence
in their product.
There's a lot of buzz around
machine learning and big data
nowadays.
I want to talk about how
it impacts ordinary uses.
If you think about
successful products,
they have three elements working
together very well-- software,
infrastructure which
runs that software,
and then the users
who make it useful.
If you think about CRM,
for example, Salesforce
disrupted the CRM
market by combining
two of these elements, the
software and the infrastructure
portion of it.
And thus, SAS was born.
Software as a service was born.
We think the next
big CRM will be
the one that combines
user intelligence along
with software as a service.
Think about how sales people
interact with software today.
That kind of intelligence will
be built inside the software.
This is something that
every business software
will have in the next decade.
This predictive intelligence
that machine learning offers
will augment human
intelligence and make
software more intelligent.
If you think about CRM, CRM
records every interaction
about customers.
But even with a wealth
of data, it still
doesn't understand
what customers want.
But on the other
hand, think about what
Google does in
Search and AdWords.
It understands
customer intention
with a few keywords
they type in search box.
And that makes it the largest
business in the world.
Now it's time that CRM learns
from what Google does, and then
does the same kind
of customer intention
prediction inside CRM.
With the launch of
Google Prediction API,
Google has opened
up that secret sauce
as to how it predicts customer
intention to the developers.
At Guesswork, we have taken
Google Prediction API,
and we have simplified
it just for one use case,
for one vertical-- to
predict customer intention
for CRM type of applications.
Also, we have
built a rule engine
on top of Prediction API.
Because most of these
predictions about customer
intention are kind of a
combination of both rules
as well as machine learning.
So we have built that rule
engine to improve accuracy
on top of Google Prediction API.
So I'm going to show you a demo
of how you can use Guesswork
to predict customer intention
and also personalize
customer interaction.
So let's take an example
where a customer sends you
an inquiry or a feedback.
And then we want to understand
what that inquiry is all about
and automatically respond
to those inquiries.
So by doing that, you
can make the sales people
more productive.
But at the same time, by
understanding what customers
are asking, you'll
be able to understand
what the customer intention is.
So you will understand what
kind of products he likes,
or if he's most likely to
stay with your company,
or is he going to churn out.
So that's the demo that
we're going to show.
So here is our dashboard.
So I'm going to
click on New Project.
So once I click on
New Project, you
can select one of
these three templates
that you already have.
So let's name this
project as Autobot.
And let's select the
auto-response engine template.
And once you click Next, you
can see the design screen.
Here, you get the customer's
feedback as the input.
And we're going to
predict three information.
So for example, let's
say a customer gives
a feedback saying,
the food was good.
Now, in that case, we
predict that the category
of this feedback is food.
And the property is quality.
And the type of the
feedback is praise.
So based on these
three predictions,
we're going to
say, what should be
the feedback that the customer
representative or the sales
person should do?
So let's click on Next.
So this gives you
a screen where you
can upload the training data.
In this case, you're
going to leave it
to the default training data.
So Guesswork
automatically comes up
with a sample set of training
data to populate the project.
So once I click on Publish
in the fourth screen,
Guesswork automatically
creates the machine
learning project underneath.
So it uses Google Prediction
API and then creates
the classification algorithm
that's needed for it,
and also creates the
rule engine that's
needed to automatically respond
to these enquiries or feedback.
So now that this is published,
I can go and click on Test.
So let's see how that works.
So I'm going to click on Run.
So now you can see
the try it out screen.
Here, I can give a feedback
saying, food was good.
And when we click on test, you
can see that it automatically
predicts that the category
is food, property is quality,
and type is praise.
And it automatically comes
up with a response saying,
we're glad that you
enjoyed the food.
Hope we get an opportunity
to serve you again.
So that's the demo.
So what you saw was the ability
to automatically respond
to the inquiry and improve
the productivity of the sales
staff or the customer
service representative.
But at the same time,
by understanding
what the customers
are speaking about,
which product is
he talking about,
and whether he is
praising the product
or complaining
about the product,
you know exactly
what the customers
want so that sales people
can upsell or cross-sell
new products to the
customers from the CRM.
I want to conclude
with two predictions
about the future of
predictive intelligence.
Now, RelateIQ
acquisition by Salesforce
has caused huge interest in
machine learning in CRM space.
We think other business software
verticals will follow suit,
be it health care, HR, law, or
any other software verticals.
The second one we
think will happen
is that as machine
learning infrastructures
like Google Prediction
API flourish,
vertical services like
Guesswork will emerge.
Because there is a lot of need
to fulfill machine learning
needs in business
specific verticals.
So we think that will
happen in future.
Thanks for listening.
This is Mani Doraisamy
from Guesswork.