Unlocking your Enterprise Data with Google Cloud Search

so much for being here. I know it’s the last
session of the day. So we’re going to make
this fun and interactive. I’m Diana Britt. RICARDO URENA: And
I am Ricardo Urena. DIANA BRITT: Between
the two of us, we’ve been at Google
for 20 plus years. So we’re both grandparents. Most of that time
for both of us, working with our search
and ML technology, so thrilled to be here today. Today, we wanted to just
share with you what we believe is the market opportunity. First of all, why do anything? If you’re looking
at prioritizing your projects and
your companies, how might we with Cloud Search? Then, we’re going to talk
about the capabilities. So we’ll give you the
fundamentals around that. But then also, more importantly,
give you some stories and use cases around how would you
apply the technology, and what other customers have
done to drive value for their organization. And then lastly, we’ll
walk you through– we’ve actually got a very
tried and true process to help you evaluate. So let’s make this
simple, so you can get to a go, no-go, or
a decision to move forward. So when you think
about Cloud Search and you think about most
organizations, a lot of folks have this challenge. They can’t find the information
within our organization. But is it a priority
to solve for? In some instances, maybe not. But what we’re seeing now
is more and more large companies are actually
moving to this how do I apply machine
learning and AI to make better
business decisions and automate my
business processes. And think of Cloud
Search as your first step in that journey. So first of all, you’ve got
to gather the information, then you’ve got to
find the information, and be able to search
on that information. And then you can take that
to documented understanding workflow, and API. And then you’ve got to take
that to process automation. Or if you’re in
manufacturing or aerospace, it might be robotics automation. And from an
architecture standpoint, it’s all about what
are these business decisions that we’re
able to actually gain out of this, right? So if you’re taking
it from a decision and really automating
a process, how do I get better
demand forecasting? How am I proving my mean
time to resolve a customer situation or de-escalate? How am I able to
turn on my engineers so that they’re actually
better able to do product launches and share
code across the organization, so they can get better
applications developed? And there’s a cost
of doing nothing. So McKinsey IDC and others have
basically said, it costs you– it takes you 20% of your
time on an average week to actually go and
look for information. OK, great. If you can’t find
it, 38% of your time is spent trying to recreate it. So back of the napkin math– if you’re an organization
with 2,000-5,000 employees, you’ve got a cost
per employee of x. You can come up very quickly
with a return on investment for your CFO to say,
why should we even think about solving for this? And we’ve all done
a really good job. We’ve got information in Slack. We have information in Box. We have information on our
product lifecycle management. We have information in SAP. But at the same
time, it’s very hard. You can’t search
across it, right? So it’s stuck in SharePoint,
Lotus Notes, databases, you name it, right? And so that is what
we’ve unlocked. So we have taken the
best of what Google does, which is Google.com,
what we all know, and applying that in
a very secure manner for your enterprise. And so with that,
I’m going to have Ricardo talk through a
couple of the slides here. Here you go, Ricardo. Thank you, Diana. RICARDO URENA: So
the first slide here that I want to focus on
is the why Google, right? And this is for your
business, right? So one of the things we do when
we do search is really focused on who is the user. What is the identity
of the user? What does she have
access to, right? What has she interacted with? | Because we take that information
to essentially generate a knowledge graph, so that we
can personalize the experience. This is what we call
identity-based search. We can also do public
website search, where you have anonymous users. But in this case, if
you know the user, you can tailor the experience. The second point is,
what is the company? What is the industry, right? Is this a manufacturing company? Is this a financial
services company? Is this public sector? Or is it health care, right? So each industry,
each company is going to have their own grammar,
their own vocabulary, right? And so we want to be
able to understand the industry terminology,
utilize machine learning, but also allow the businesses
to load their own synonyms and ontologies so
that we understand, from a personalization
perspective, the context, right? And then really, the third point
is the application context, right? If you’re a manufacturing
company and you’re manufacturing products, and
say you do a search for lead– so lead in
manufacturing, how long does it take to manufacturer
a lead time, right? That has a certain meaning. But if that
manufacturing company deploys a CRM application, a
lead is really an opportunity, right? So it has different contexts. So being able to
disambiguate whether it’s lead time for manufacturing,
or lead from an opportunities perspective is very key. So we’re able to focus on the
user, the company context, and the application context. What we have here is the
architecture diagram. And there’s a lot of data here. So I want to focus your
attention in the middle. In the middle, you’ll see that
label that say Google Cloud Search Index, right? That is the index. The index itself is
in the cloud, OK? Your data can still
reside on-premise. If your data is on the cloud,
it can reside on the cloud. Really, you control your data. You decide very
prescriptively as to what data you
want to index, OK? So that’s the first point. The second point– this
is serverless, right? So you don’t have to worry
about managing clusters, de-fragmenting indices. We give you an indexing API
to push data and a query API to get data out. At the bottom, you’ll see
that it says ACL-aware. Our product marketing
manager said really what he wanted to put
there was permissions. But permissions
didn’t fit in the box. So he put ACL-aware. So ACL-aware stands for
access control list, right? So when you index data, you’re
pushing data, the content, the metadata, and the user
and group information, right? Think about if you’re indexing
Office 365, SharePoint, OneDrive. You’re going to grab content,
and also the user information, right? So we want to grab the content,
and the metadata, and the ACLs, right? And when you think
about content, content comes in different
shapes and sizes, right? So it can be a PDF file. So a PDF file, we can take a
PDF file, we’ll grab the text, we’ll extract the OCR
information from the images, and we’ll go that
into the index, right? But content can also be like
a case record or a ticket number– so ticket ID,
status, owner, description. So it could also
be very structured. So when we talk about
content, really the point that I want to make here is
that it can be unstructured content or structured content. And that’s all
managed in the cloud. And we do that at scale, right? And the other thing we
do is we bring the search quality from Google, right? So we want to be able to
bring the search quality. And we want to give you
the ability to index as much data as possible. So on the left-hand side, you’ll
see the various data sources, right? And as part of the
data sources, Google will give you what we call
a connector SDK, which is open source, right? And then partners are going
to develop connectors. And a connector could be a
connector for a file share, a connector for a database, a
connector for PTC Windchill. It can pretty much be anything. And the role of the connector
is how do we grab data, how often do we push the
data, and what data should we push, right? And so if the data is
on-premise, that’s fine. It could be in the cloud, or
it can be in a public cloud. At the top, we have
an image of G Suite. So G Suite is just
a data source. That could also be
Office 365, right? So really, we have connectors
for these third party data sources. Or if you wanted to
build your own connector, you can build your
own connector. And really, what we’re
going to allow you to do is to do it at scale. How often do you want to do it? We really don’t care. On the right-hand side,
we have Query API. And the Query API
is I run a query, I want to get the
results back, right? So I’m going to
authenticate the user. We’ll take care of
the authentication. We’ll do the trimming. And the look and feel could be
whatever you want that to be. It could be as
simple as Google.com. We provide an out-of-the-box
look and feel. We also provide an embeddable
search widget, which is client-side JavaScript. Imagine if you wanted to
very quickly embed Cloud Search within your
Salesforce application, or with an Office 365, right? So you can have the search bar
be powered by Cloud Search. Or you can build a
custom front-end, OK? Now, at the bottom, we have
these additional services. Think about Cloud AI,
document classification, some of the Vision APIs. Imagine you have videos, right? You have videos. You have [INAUDIBLE]
diagrams, Maybe from a finance use case, maybe
loan applications, mortgage applications. And you want to do more
than simply search, right? And you want to put that
through an indexing pipeline. We’ll extract the entity,
say the key value pair information– first name, last
name, this is a semiconductor, this is a controlled
engineering diagram. We’ll take that data that
comes from the Vision APIs. We’ll feed the content plus the
metadata into the search index. And then you’ll be
able to search it. And how you search on it– it
could be through mobile phone. Maybe you integrate it with
the ChatBot integration, right? So you have a dialog
flow application where people can
interact, right? And that, in turn,
invokes the Query API. So really, the point that I
wanted to make here– there’s a lot of flexibility. We can do this for
a little bit of data or large amounts of data. This is a slide on the
Connector Marketplace. This is the marketplace,
similar to what Salesforce has. Partners comes here. They develop connectors. This is a representative
list, right? It’s not an exhaustive list. So if you don’t see
your connector here, it doesn’t mean it
doesn’t exist, right? And then connectors
are for content. We talked about content. But also think
about people, right? When you do a search, you
find the engineering document, you want to find who is the
subject matter expert, right? So people search is also key. And so connectors come
in terms of content. And then we also have
these identity connectors to interact with Microsoft Azure
ID, maybe an HR application. We also provide a connector
SDK that is open source that you can modify. But again, an exhaustive
list of connectors to allow you to essentially
load as much data as you want into the index. All right. So this is a little bit more
specific, down in the weeds. But this is an example of
what we call natural language understanding, or natural
language processing. And the example that
we’re looking at here is a consumer example,
where I’m saying, show me all movies that have
this type of actor, right? So that’s the consumer world. But imagine within
your organization, maybe it’s a ticketing
application or a case management application. And what you want to do is show
me all tickets that are open, or show me all tickets that
are owned by Diana, or show me all tickets open in the
last 30 days, right? Those are the types of
queries that you can do against third party data sets. And you can do them in a very
conversational format, right? And the system will
understand that. And the reason we’re
able to do this is because we’re utilizing
the Google.com infrastructure, right? And so we’re very good from
understanding languages. Later on, we’ll see that
we do a lot of languages. But this is one of the
things that, in my opinion, is a big differentiator
for Cloud Search. This example talks about
search quality, right? So we loaded data
into the index, but we want to have
great relevancy, right? And so I want to
focus your attention on the 15% successful searches. Really, what that means is
that, when I go in there, I did a search. So we tuned the search
quality algorithms, and we saw that increase, right? So what’s going on
is I went in there and I found the answer, right? So Google has given
me the answer. So hence, that number went up. That’s what we call precision. Don’t give me everything, just
give me the answer, right? The other example is these
zero result queries, right? And I’ll show you a slide
and example in a second. But basically,
when I ran a query, the number of zero result
queries dropped, right? And let me tell you
why that’s the case. So in this case, this is an
example from the pharma space. And it’s this notion of
related concepts, right? Say you were in the
construction industry, and you wanted to do a
query on wall type material, like particle board,
sheet rock, gypsum, right? You run a query, and I do
a search for sheet rock. But the way it’s stored
within the system, it says sheet rock,
gypsum, or particle board. It’s actually stored
very differently, right? But what the system
is going to do is it’s going to automatically,
on behalf of the user, expand the query and bring
back related concepts. So in this case, the user
does a search for ibuprofen. And the system
says, oh, ibuprofen is equivalent to Advil, it’s
equivalent to painkiller. And so it brings back results. And so we’re able to
do that automatically using machine learning. And this is something
we get from Google.com. So again, this is something that
you don’t have to do manually, but the system just does it
automatically for you, OK? All right. So these are the core benefits. So in my opinion, this
is why Google, right? Why Cloud Search, right? We talked about enterprise-wide
search with the connectors. Open source, a
lot of connectors, a partner ecosystem essentially
allow you to index as much data as you want. The second point at the
bottom– zero administration, its serverless, right? I don’t have to worry about
whether the lights are on, right? It’s basically on if
I do 100,000 or nine billion records, right? I can push all that data. The system is going
to be up and running. I don’t have to worry about
index de-fragmentation. In the middle, we have
hardened security, right? That’s the permissions
I talked about, the ACL. We’ll have the user information. We’ll also support things such
as nested hierarchies, right? When you pull data from file
shares, or Confluence, or Jira, sometimes you have
a lot of nesting. And to be able to
do that at scale can be challenging, right? It’s one thing to do it
for a small data set. But when you have millions
and millions of records, there’s performance
implications, and you need something that’s
robust and sophisticated. The feedback we’ve
gotten is that that is a big differentiator
for us, right? We talked about world class
search, the Google.com infrastructure, which
is this area here. And so we’re utilizing
the same web stack. And then at the top right,
the 148 script language pairs, right? So we’re not just doing the
Western European languages. We’re doing Mandarin. We’re doing Vietnamese, the
more complicated languages. And we’re giving you
great relevancy, right? So it’s not just
saying, oh, we can support the indexing
of those languages, but is the relevancy good? And because we do this
within Google.com, we’ve brought down
the language bundles. And we have those available
within Cloud Search. And then obviously, the
certifications and compliance that you would expect– so as Diana mentioned, we
launched 11 months ago. Since then, we’ve amassed
a large number of customers across multiple industries. And again, this is a subset– finance, retail,
manufacturing, government. Essentially, any company that
has a large amount of data, and needs to find that
data, and analyze the data, they’re good candidates
for Cloud Search, right? So really, that’s– we’re
seeing horizontal adoption, and also very much
vertical adoption. This is a platform, right? So pretty much any industry. So what I want to do now
is show you two examples. Quicken Loans is in the
finance sector, highly regulated industry, right? Typically, finance,
a lot of regulations, a lot of compliance. And what they
needed to do is they get a lot of mortgage
applications, loan applications, financial
data from the government, because what they do is they
do mortgages for cars, autos. And what they do is
they take the data, they classify the data as to
the type of mortgage or loan application, they index the
data within Cloud Search. And then when a user comes
in through the website, they say, well, this is
a first time homebuyer, hence these are the documents
he or she should see, right? So they personalize
search experience on top of allowing
them to search. And they have this
application called Guru, where they can find the information. And essentially, people
go in and ask questions. This is a neat
example, in that it’s a combination of Cloud
Search plus Cloud AI, the combination of the two,
and also within the finance industry. The other example is a
multinational, PepsiCo. I’m sure you’re
familiar with some of the products they offer. They had essentially
an extensive list of legacy systems– databases, SharePoint, content
management systems, Documentum. You name it, right? And so what they did is they
deployed Cloud Search across 15 different applications, right? So everything from we’re
doing a new commercial, what are some of the imagery or
art that we need to utilize, to find, right? So that’s really from a
brand management, right? From an internet
perspective, I’m an employee. I need to find information
within the organization, right? From a people
search perspective, they have 350,000 users. So they want to find the experts
within the organization, right? So really, a variety– and the architecture
here you’ll see, and it looks very
similar, right? So SharePoint, a database,
Confluence, Drive, Documentum, websites file shares– we create an intelligent,
consolidated index. We’ll load the identities. And then they’re able
to essentially find the information, whether it be
through a mobile phone or a web UI. With that, I’m going to hand
it over to my colleague, Diana. DIANA BRITT: Great. Thanks, Ricardo. So another customer, Broadcom–
actually headquartered here in Singapore, also
operating globally, a semiconductor
organization– they’ve got hundreds of
thousands of engineers. And I’m sure, if you’re
tracking Broadcom, they are growing by acquisition. And so the largest
challenge that they have is, how do I get my
engineers, and get them aligned and
cross-sharing across the different organizations? And so they’re
integrating the Box, and into Liferay, and Oracle. Andy, their CIO,
gets very excited when I meet him
for lunch, and he gets to show me his
application that he has on his phone to actually
quickly find information. So Broadcom is a really
interesting use case for semiconductors and an
engineering-type use case. Another example is Philips. Another highly regulated
industry is health care;. And so this is their division
actually out of Europe. And they’ve got
5,000 field engineers that go out actually to
service the MRI machines, which is a mission critical piece
of machinery in any hospital. And so they’re out there
doing the warranty, the piece-part
maintenance, et cetera. But if this thing fails,
the hospital’s down. And the challenge that they
had is the documentation was in different systems. They’re on-site, they
couldn’t find the information. And so by applying Cloud
Search and using it on a mobile application,
they were able to, most importantly,
really gain some of operational
efficiencies– so one, reduce the time it took them
to find information by 27%; and then improve the time to
actually resolve the situation by 15% So again, getting
to that MTRR, which is mean time to resolve. The architecture
of Philips, very similar to what
Ricardo had presented, is actually very similar. So they’re using App Engine on
the front-end to build a very easy-to-use user interface,
going into Ping Identity and for the identity management,
knocking into Active Directory, connecting to the secure index. And then on the
front-end, they’ve got their usual suspects. They’ve got their product
catalog, ServiceMax, SharePoint, et cetera, so
similar the other architecture slides you’ve seen. And then the UI–
again, the UI can be what you would like it to be. For their team, they wanted
to make it very simple. They’re accessing it
on a mobile device. So they wanted the ability to
say, OK, I can see this image and automatically
I, in my head, know where that application lives,
or where that document lives. But I could also search just
by data source or by modality. The other mention that
Ricardo had was around ChatBot integration with dialog flow. So this is actually another
great use case, especially in customer service. So as you’re really
thinking about how am I solving for business
process challenges, this is a great way to
do it, integrating Cloud Search with the dialog flow. Another organization, part of
the United Nations, the Food Aid Organization, is also
leveraging this as well. And they’ve got
20,000 plus folks that are actually accessing this
application on a daily basis. And then powering your brand– so we have Alitalia, we have
ACER, we have Genentech Roche, we have DHL, we have BBVA, and
other highly regulated banking industries. So we’ve got customers
that are actually using Cloud Search to
power their website, and using it from a
brand perspective. So this all comes to a fruition
of, OK, Diana and Ricardo, that’s great, but
what do I do next? How do I engage? And what would be a process
to actually evaluate or go about this? And throughout the
last several years, we’ve come up with this
evaluation process. It’s a four to six week
process that, at each gate, gets you to a decision
making process. Do we go? Or do we not go? Right? It really helps you
build your business case that you’re going to take
to your board of directors, your executive team
to say, hey, we’re going to think 10x around
solving a challenge. We’re going to drive this
return on investment, or drive this type
of business value. So it all starts with,
what’s the use case? So it’s not I want to
integrate into SharePoint, or ServiceNow, or Salesforce. It’s, what are we
trying to solve for? And if I’m not able to
solve for that, what’s the cost associated with that? If an engineer can’t
find that information, is he now recreating
a piece-part that goes into inventory and has
a carrying cost, for example? So we’re really working
with you to really define what are those use cases and
understanding those personas in the organization. And where can we
drive the most value? And then we work with partners. And we have certified partners
in Southeast Asia, North Asia, and in India that can
work with your teams. And we’ll go through
a discovery workshop. And that’s typically a
two to four hour workshop. We go through, obviously,
a security review with your team, number one. We also then do a
business review, really going through
the use cases, and then go through
a technical plan. And the outcome of
that is actually an output and executive summary
that we actually give to you, so you can actually take
that into your organization to bring this to your
leadership, to get sponsorship, to say, do we go,
or do we not go? And typically, once
we get to this phase, you’re also getting
a sense of what is the return on
investment and the budget. And then based upon
that, we come up with the plan for
evaluation, which is a prototype versus a
full proof of concept. And in doing so,
it’s really– we’ll take a subset of your data. We’ll work with the partner. We’ll actually run it
through a connector. But at the end of the day,
it’s all about your end users, and what does that
user experience. So how do we prove that to
you is really what we get to. So we want to get
through a business fit, and a technical fit,
and get you to a go, no-go. So this has worked really
well with all of the customers that you’ve seen. So it’s really a partnership. And we’ve got a detailed plan. So it’s a combination
of you, the customer, us working with
Google, and the partner to come up with a
joint plan to say, all right, let’s get
this done, because we all know that sometimes we get
in technology projects that last quite a long time. And it’s also, how can we
get just a really quick hit and a quick win
for our business? So we’re thrilled. If there’s any follow-up,
Ricardo and I are here. We’re also happy
to share out we’ve got some great customer
stories that we can share with you, some videos,
and some additional detail and resources. So I’m sure you also know
your Google reps that are here today. And they’re also happy
to work with you. I’m based here in Singapore. Ricardo’s in Atlanta. But Atlanta is not that
far from Singapore. RICARDO URENA:
It’s not that far. DIANA BRITT: So
thank you so much. We’re honored to be here. RICARDO URENA: Thank
you, everybody. DIANA BRITT: Enjoy the rest
of the afternoon, the evening. And again, thank you for
attending the Google Cloud Summit. Take care. RICARDO URENA: Thank you. [APPLAUSE] [MUSIC PLAYING]