Modernization Hub

Modernization and Improvement
Rapid Prototyping for the Public Sector (Cloud Next ’19)

Rapid Prototyping for the Public Sector (Cloud Next ’19)


[MUSIC PLAYING] DAVID STEPP: So just quick
introduction, who are we? Who is Accenture
Federal Services? So let’s start with
Accenture, LLP. Global firm, 460,000
plus people serving 80% of the Fortune 500 clients. Accenture Federal Services
is a wholly owned subsidiary. And so we’re about 9,500
plus people, mostly located in the DC area. And we serve only
federal government. So we don’t do any
state or local. So it’s all the
three letter agencies that you can talk about. So what does that
look like exactly? And so we like starting
here to just kind of set the stage in terms of the
scope of our solutions and the problem sets
that we’re attacking. So 320 million customers
is our customer base. $4 trillion is the budget
of the federal government. There’s millions and
millions of vehicles. If you think about the Army
and the DOD, Air Force– even the US Postal Service
has over a million vehicles that they’re tracking
in terms of maintenance costs, and collisions, and
insurance, and repairs, and about 2 and 1/2
million employees. So that’s the largest
employer in the United States, and it’s one of the largest
employers in the entire world. So anytime we’re talking about
rapid prototyping, anytime we’re talking about
our solutions, these are global problems with
hundreds of millions of people that are being affected. Anyone know what this image is? It’s the last image
of the Curiosity Rover that just went dark a
couple of weeks ago, a couple months ago now. And its last message,
it’s pretty poetic. This is like a poetic
interpretation. My battery is low,
and it’s getting dark. So the reason I put
this up here is, again, to kind of prove a point of
where our problem space is. So Curiosity was
launched 14 years ago. So 2005, think
about the technology that you had with
you in 2005, right? Flip phones. I was still in college. I had much longer hair. I was in the engineering
program, right? Looked a little different. But we sent this piece
and hunk of metal through space into another
planet, and it lasted 14 years. It was only supposed to last
two years or three years. I don’t know the exact,
but it lasted 14 years. And it sent back all
these amazing pictures, including this,
which is not stars. This is actually
just a sandstorm. It’s fuzz. It looks really cool, but
it was its last image. MICHAEL RAMOS: Yeah,
and to really add on to what Dave kind of just
honed in on, the problem landscape that our
government faces is extremely vast and complex. And to that point, there’s
room for innovation all over, whether we’re
talking about, again, DOD came up with a GPS,
vaccines, health and longevity. The problems are so
complex, it would be really nice to invite our
commercial partners in and try to solve everything for us. But the government is on
their own in a lot of cases. And they’re working hard. And again, that’s the
importance of rapid prototyping. DAVID STEPP: Yeah. And these are game
changing innovations, life changing innovations,
world scale innovations. The internet, right? Where would we all be
without the internet? That was an innovation
and investment from the federal government. Google, National
Science Foundation– we wouldn’t be here
without investments from the government. And so how do we solve
these massive problems, these intergalactic problems,
these world changing problems quickly? We use rapid prototyping. So we both work outside of
the Accenture Federal Digital Studio. So we have a very cool,
sleek, urban place down in Washington, DC. Lots of wood, and exposed
beams, and hip people with cool laptops, and cool
haircuts, and cool shoes solving these problems. And so we kind of approached the
problem in four different ways. So, vision and organization,
value and design, technology, and scaling. I’m going to step
through in detail what we mean about each of those. But first is kind
of our mindset. It’s the Michelin
star restaurant versus the Zagat restaurant. So when we do delivery
for our federal customers when we’re on a
project for Agency X, and we bring our
domain expertise to solve those mission problems,
that’s our Zagat approach. That’s the Mexican
restaurant that has the best burrito
in the entire city, and they make that
burrito every single day. And they do it really excellent. We take the Michelin
star approach, right? So every night, we have a
different menu, a tasting menu. We work with
seasonal ingredients. We work at the
latest technology. Blockchain, sure, let’s try to
add Blockchain to something, right? Cloud services, sure. The new cloud services
that were just announced this
morning, Anthos, right? We’re going to already
start playing with that. Here, my lead DevOps engineer is
ready to get his hands on that. So vision and
organization, I really like starting with this because
I think every time you’re approaching a problem,
you have to think about where you’re
going to be currently, where going to be three months,
six months, 18 months, and five years out, right? It’s the road that
you’re running towards. So we have capabilities,
and we have solutions. So capabilities
are our verticals. We’re not going to be
everything to everybody. We’re going to do web
and mobile development. We’re going to do DevOps. We’re going to do platforms. And we’re going to do
applied intelligence. And then we’re going to
have our solutions, right? And so later on today, we’re
not going to just bore you to death by PowerPoint. We’re going to show you a bunch
of our actual rapid prototypes, what we’re going to build. And one of the solutions is the
next generation contact center that Mike’s been
working on really hard over the last two weeks. And then, again, people. So I want to iterate
on this point. The type of people that we’re
bringing into our organization are non-traditional people. Mike has a background
in fintech. We have people who are
econ background, people who are non-traditional
federal consultants, and we bring them because they
bring a wider range of thinking to these hard problems. But most important– and I have
it bolded at the bottom there– is the ability to learn. Right? So we’re working with the new,
with things that are announced in the morning, right? There’s nobody who’s an expert. There’s no job description
that I can post and say, let’s have five years
of Anthos experience. It just doesn’t exist. So we need to adapt on the fly. Value and design. So me and Mike work
very closely together with our human
centered design team. And so human centered design,
that was kind of a new term to me when I started working for
the Accenture Federal Digital Studio. And it’s basically a
fancy way of saying we put the human first, right? We don’t take a
technology first approach. How is the human interacting
with the systems? What’s the human problem? What’s the human experience? And so we do these things. They’re called design sessions. We bring in all of
our stakeholders. We bring in the operators. We bring in the
security experts. We bring in smart
folks like Mike, who have an engineering background. And we all get in a big room. And it’s very granola,
but it’s very effective. And we do sticky
notes and whiteboards, and we come up with
here’s the main issues. Here’s what we want to solve. OK? And then once we start
going through let’s iterating on the
technology approach and how do we actually– is it feasible and what’s
the return on investment? Maybe we can solve it,
but it costs $100 million, and the problem is only a
$500,000 type of problem. OK, well, let’s not
actually do that. And then we talk about our
business process redesign, and user interface,
and user experience. And something we like to
talk about within the studio is our liquid approach,
so liquid expectations. What does that mean? That means our customer base,
our 320 million customers that we’re supporting within
the federal government had expectations that
their day to day lives– that their interactions
with the federal government are going to be the same
as their day to day lives. So if they pull up a phone
app, and they order a car, it’s a touch of a button. They want to have that same
simple experience when they’re interacting with IRS,
Social Security, whatever age– they’re going to get
their driver’s license renewed. And how we do this? MICHAEL RAMOS: Yeah, so a rapid
prototyping team, essentially a software development
team, is not like your traditional
development team, right? If I’m on a contract
for one to five years, we’re going to get some pretty
strict requirements on what it is we’re developing, right? And with that, I can go out
and search for a developer y who has capability x. And I can be pretty
stringent on my interviewing them to figure out if
they’re going to be really capable of doing this job. Like Dave said
earlier, we’re not looking for that
within our team. We need generalists
that can solve, again, any one of these problems
in this vast landscape of problems. So while we do want to
standardize on efficiencies like JavaScript that’s kind
of taken over the full stack and even get into
the data world a bit, we want to lean heavily
on our vendor partners, and manage services, and be able
to provide custom integrations into cool capabilities like
Cloud ML Engine or Contact Center AI. And we want engineers who
could do that with Python because maybe that’s what
our clients are working with in their current systems. Maybe they’re working
on Salesforce CRM, or maybe it’s service now. And the way those
API’s work need to be configured differently. And Salesforce has its
own programming language. Again, we just want those
generalist engineers capable of taking on any
problem and solving it with the constraints
that they’re given. And to that point,
engineers are lazy. I mean, they’re efficient. So any mundane
tasks that kind of falls below a value
line of mission value– like if I’m writing
code, most of the code I’m writing for this
application, I’m hoping it’s producing some type of
value for the end user. Anything that’s not– we’re
talking about infrastructure, logging into VM, scaling out
the underlying platform– we want to automate that
as much as possible. So we have awesome engineers,
like Jordan Taylor here, who sets up an entire
infrastructure for us in Kubernetes, and Pivotal
Cloud Foundry, and Red Hat. Again, a broader
range of technologies because we want to be able to
provide for all our clients the same way that
we do internally in terms of development and
providing efficiencies in that regard. DAVID STEPP: So we talked
about the problem set. We talked about how we organize
our rapid prototyping team. Now, how do we actually go
about and implement this? So say we start with our
problem statement, right? There’s a federal agency. They have a legacy support
desk, and they’re trying to think about modernization. And modernization
means a lot of things to a lot of different people
across the stakeholders. And so they’re trying to balance
the risk between maintaining that domain expertise
for the people who’ve been supporting the mission
for years and years and costs. Costs are a real driver. And using automation,
and generally using automation to help
improve the domain experts and ultimately deliver a better
value experience to the end customer. And so you have some people
that are viewing this. You go, all right, well, they
control the purse strings. Let’s introduce CRM tools. Let’s introduce our automation. It’s going to drive down
our contracting costs, and we’ll have more revenue to
invest in, in additional areas. Or maybe because
they’re federally funded and their budget got cut this
year for a political reason. Or maybe you have
people who’ve seen this before, CRM tools won’t
fit within our mission. People won’t adopt it. This is the same story
we’ve gone through again. It doesn’t work. Never does. And then some
people, they’ve been doing their job for 20 years,
and they don’t want to change. They think the way
they’ve created it is the best way to get it done. And then some people
are– let’s be real. There’s the human element. Some people, when
there’s automation in technology being introduced
that’s new, they get worried. People have families to support. They’re worried about
job loss, right? And so we need to take
into consideration all of these different
stakeholders, all of these different concerns,
all of these different risks– the people, the
technology, the cost– into how we actually
build the problems. And so we think
rapid prototyping is a really good way to do
this because we can fail fast. We can build something
very quickly, and when I say
quickly, I’m talking about one day, one week,
using one or two engineers, or two weeks with one
person grinding it out, building out the next
generation contact center, which we’re
going to show you in a couple of minutes. And then actually getting
that in front of our customers really quickly, and getting
end users’ feedback, and then iterating over that. MICHAEL RAMOS: So
the development team, your rapid prototyping
team, you want to take that problem
statement, and you want to be careful with
how you unpacked it, right? Hopefully, you’re leaning
in heavily on design methods and methodologies, such as human
centered user centric design. That way, you’re working with
a team who really understands that problem they’ve given you. But with that, I want our team
to be keen on that problem, too. Because I don’t want
to just take a problem and develop something that’s
not going to resonate or that’s not going to be useful. Or worst case, that can
never be implemented with a certain
federal client, right? We really need to
consider what’s going to be feasible
to a certain extent. We want to be fit
for federal in terms of compliance, current
systems, current solutions, and integrations. Then we can move into
our layer of usability. And OK, well, is what
we’re going to build, is it going to be usable? Again, in the context of
federal, it’s pretty important. And then useful, this is
where we get to have fun. All right, how can
we add some spice? How can we add some
pizzazz and really make any solution we develop
within two weeks cool? And then with that on top,
if we’ve done these things, we’re going to have some
time to really provide some efficiencies
with a solution. And if we’re providing
any type of solution that makes the mission more
efficient for our end users, hopefully, that provides
a joyful experience. If I want to get from
point A to point B and I can push a
button, and somebody’s going to come pick me
up and take me there, that’s a joyful experience. That exists all over
the federal landscape. Cool. So back to the
problem statement, we have a legacy service
desk that’s been shut down. We have a CRM
that’s been brought into the bureau’s organization. And they want to
know how to use it. They come to our group
and say, how are we going to use this thing? It’s not really
fitting in our mission. How can we make it relevant? We’ve gotten a
budget to explore AI, to explore some cool solutions. What can we do? Help us. And what we’re
about to demonstrate is essentially that next– how do you pronounce it, Dave? DAVID STEPP: The next
generation contact center? Yeah, so about
nine months ago, we were very lucky to be added as
an alpha partner for Contact Center AI. And so what you’re about to
see is our second generation of working through that. We built out one over
the last six months, and we showed it to a lot
of our federal customers as they came through and
got a lot of feedback in terms of how does this
integrate with our CRM tool, how do you have omni
channel support. And so what we’re about to show
you in a couple of minutes, this is our reference
architecture. We worked with Google
as a strong partner. We worked with Salesforce as our
CRM tool that we’ve integrated, and then Twilio. We’re going to
show you in an omni channel the ability to call,
do telephony, live streaming to an operator, and having
Agent Assist working in place. So that means I’m
actually having a customer calling an operator. I’m splitting off of that
actual raw audio file, doing transcription on the
fly, searching a knowledge base, and then in real time,
showing that knowledge base to my operator so
that they can resolve the issue of the customer
faster and more efficient, providing them the
information that they need. And we’re using Contact
Center AI and Dialogflow to create a single
virtual agent that you can interact with via
chat, text, or telephone. Did I miss anything? He’s the engineer. He’s the smart guy who
actually built this. OK, so we made a
very California demo. It’s about avocados. And so with that, if we can
switch to the live demo screen. So what everybody
is seeing right here is our Salesforce operator pane. So this is an example of
what your support desk operator would see. And I’ll walk through
the different panels and what they mean. And again, just
to level set, this is two weeks of Mike’s time
doing some engineering. As you know, his part time
job to get ready for this– we are grinding this
out to be truthful– last night, like
true engineers do. And by last night, I mean
like two hours ago is when we finally got it working. And we rushed here
and the support staff was extremely gracious to
let us run through this and practice to make sure
the phone calls worked with the mics. So we’re not nervous at
all because it’ll work. It worked a little bit ago,
but it is a world premiere. So enjoy. Mike’s going to play the role
of our avocado researcher or farmer, whatever
one he decides to pick. It will be up to him. And then let’s get started. MICHAEL RAMOS: So I’m a
researcher out on the field, and I’m experiencing some issue. And I’ve been told I’ve got
this intelligent service desk I can call to be given some
insights and context that maybe the virtual agent
couldn’t give me. That may work [INAUDIBLE]. And we’ll show you the
virtual agent chat afterwards. So I’m calling in. It’s going through the
call infrastructure now Twilio provides. David, hit phone. Just want to make sure we
capped off the last one. There we go. DAVID STEPP: Looks like Mike
Ramos, avocado researcher, is calling me. Hi, Mr. Ramos. How can I help you today? MICHAEL RAMOS: Hi, I’m
experiencing issues on my avocado farm. DAVID STEPP: All right, so
we’re going to pause here. And there’s a lot to unpack
of what just happened, right? It seems pretty simple. So I’ll move over
here, and I wish there was a laser
pointer of some sort. OK, so Mike built a
mobile application that takes his voice
and actually does the transcription on a client. So it does the transcription
on the phone, just because in terms of latency, that’s
a lot faster than sending the raw audio to Dialogflow,
having Dialogflow do the transcription, and then
do some of the intent matching. So the mobile app is actually
splitting the audio file, so the audio file is being sent. We couldn’t get the– because of the mics and
stuff like that– the way you could hear it,
but we actually have it set up so you can do
the talking back and forth between the operator. And the transcription
is being sent to Dialogflow and Contact Center
AI to actually do three things. One, the top here, or
suggested responses, it’s interpreting what
Mike was saying, asking about his avocado problems. And it’s matching it
to our knowledge base. So our knowledge base
is literally key values. It’s questions and answers. So it tries to match that on the
question, which you see here. And if I expand it
as the operator, I have some suggested
content that I can ask back. Now at the bottom here,
suggested fulfillment, this is where it’s matching on intents. So intents, think of
them as more static type of workflows, your generic
chat bot type of interface. So if Mike was calling
about password resets, we might match here. It says easy way to reset
a password, click a button, have that fulfillment working. We have it
implemented right now, but the actual interface isn’t– we didn’t get to that yet. But Mike created a
Machine Learning model that predicts the future
prices of avocados. MICHAEL RAMOS: Yeah,
that’s the other part. We used Cloud ML Engine to
essentially create a regression model that this
researcher might want to use out on the field, given
the data they’re collecting. And then an intent
that’s built in that would show up as just price
insight on this potential field in crop yield. And what you would see
there, as you know, that just being
integrated into Salesforce and then plugging in with your
relevant Salesforce data sets. DAVID STEPP: And then,
third, the most exciting part over here, is related documents. So we uploaded a bunch of PDFs. You can upload HTML. Essentially, it’s
unstructured information. And when Mike’s
asking a question, we’re doing the transcription. It’s pulling out the
entities, and it’s going through that unstructured text. And it’s actually finding
which relevant documents that I might send him. And so, I might be able to
share this link with Mike, the avocado researcher, on
choosing an avocado variety. I’m sorry. I’m not an expert
in avocados yet. All right, so, I’m
going to ask Mike, I’m sorry to hear that you have
problems with your avocado. How can I help you,
more specifically? MICHAEL RAMOS: Specifically,
I’m experiencing problems with root rot. DAVID STEPP: So as you
see, it went through. It actually pulls up
different suggestions. There are fungicides, Mike, that
are available from the nursery. And there are a number of things
you can do before applying something like that. So instead of an operator
having to listen, interpret, open up another system, do a
search, I have this right here. I can do a call
resolution much faster. MICHAEL RAMOS:
And typically, you wouldn’t want your
rapid prototyping cycles to be just two weeks. Given four weeks, this
becomes a full integration within Salesforce,
full integration within creating actual
reports and distributing throughout the CRM
to the right people. DAVID STEPP: All right. That’s it. Well, thank you
everybody for your time. We’re here if you
want to talk to us. Really appreciate you
attending the session. [APPLAUSE] Thank you. [MUSIC PLAYING]

Leave a Reply

Your email address will not be published. Required fields are marked *