- We spent the afternoon
discussing the intersection of AI
and equality, and now
we'd like to show you
how Salesforce is using
Agentforce to bridge that gap
because we know that humans
with agents drive
customer success together.
- All right, how's
everybody feeling? Okay.
How are my agents out there feeling?
I mean, let's bring up the energy.
I am excited to share with you all.
We're going to see agents in action.
We've heard a lot about AI
you've heard some great information.
I hope everybody is taking notes.
And we're seeing how important
it is that we all lean in
and we learn and truly understand
how our relationship is with AI.
And like I tell many- AI is kind
of like the discovery of fire.
It can warm your house or burn it down
unless you know what we're doing, right?
You want to be able to put
guardrails on things.
And we're going to see actually that
what we're going to see in
this amazing demo are three
important things I want you
to take away from the ability
and we've seen it at individual levels.
If you've worked with ChatGPT
or any of those LLMs, you're going to see
how we embed it at an enterprise level
and how we do that at scale.
And also what's very important
when you're taking AI
integrating it into an
enterprise software is not just
scale, but so fundamentally
important is trust.
that you use in there is
done in a trusted way.
And the other thing
that's really important
for our customers is that it meets them
where they are in the flow of work.
So you're gonna be able to see
these three amazing things in
action with my friend
who's absolutely amazing,
who's going to walk you through the demo.
And I'm going to be out
here because I'm going to,
I'm going to answer him a little bit.
Let's bring Hari out to come on out
and do this agent in live.
Hari, say hello to the people!
Leah. So good to see you.
Oh my gosh, thank you for
having me. Hello everybody.
It is such a pleasure to be
here with all of you, especially
with my dear friend Leah.
As she mentioned, my name is Hari
and I'm actually a member of
our company messaging team
And what that just means is
that I have the opportunity
to connect with audiences like all of you
and share incredible stories
of Salesforce innovation.
And as Leah mentioned, this
is a very exciting demo for us
because we're getting to
dive in to Agentforce.
Agentforce is a brand new set
of tools within our platform
that helps humans and agents drive
customer success together.
And today we get to
actually bring it to life
through the story of one of
our amazing customers, Saks.
And before I begin, I
actually want to take a moment
to also salute my amazing
demo drivers that are here
with me, Ruscha and Jonathan,
please give them a hand.
You know, demo storytelling
is a team sport,
so we're going to get into this.
Now I wanna step back for a
moment from where we are today.
And you know, I was
obviously very excited to,
to be offered this opportunity
to speak next to Leah,
but I was a little nervous
because I knew I needed to look my best.
So before this I had to, you
know, did some online shopping
and I called around to a couple retailers
and in the end when I was
following up about one
of my orders, I had an
experience with customer service
and it went a little bit like this.
- Hello. And thank you for
calling customer service.
If you know the extension
of the party you are trying
to reach, you may dial it at any time.
- Our store out, we're gonna,
we're gonna hang that up
because I think we've all
been there before, right?
There's nothing worse
than calling a retailer
and getting shuffled around a bunch
of automated switch portal options.
It's not the best. But
thankfully our customer
Saks does things differently.
They are a customer first company
and they've completely
reimagined the service experience
So now let's actually
hear what it sounds like
to call Saks about a
recent order that I placed.
- Hi, I'm Sophie, your Saks AI assistant.
How can I help you today?
- Hey Sophie. So I'm actually concerned
that the sweater I just
ordered doesn't quite fit me,
doesn't feel like it's the right size.
- I'm sorry to hear
that. I'm happy to help.
Is this about the Saks
Fifth Avenue collection?
Cashmere baseball bomber sweater you
purchased on September 9th?
The only issue is it's
just a little tight.
- I apologize that the
sizing wasn't right.
Given your past purchases,
a medium should offer a comfortable,
relaxed fit. How does that sound?
- You know, I like medium
Leah's nodding right here.
It's just, I wanna make sure it's not too
- This sweater is 100% cashmere,
so we recommend dry cleaning
to maintain the natural
fibers and prevent shrinking.
Would you prefer a medium size?
- I think we're going with medium.
I think we're going with
medium. Yeah, that sounds great.
- Given your shipping address,
it should arrive within three
to five business days. Will that work?
- Ooh, no. I've got obviously
a big event coming up,
Any way to get it in two days.
- Unfortunately that's
the quickest shipping
Would you like me to
transfer you to someone
who can provide further assistance?
- No, you've
been a huge help Sophie.
Thank you. But no, not
for now. Thank you. Okay.
Already that was a better
service experience, right?
And fun facts. That
agent that we just heard from
and is an actual AI agent that Saks built
and deployed in their business in just two
Now in that interaction,
Sophie could do a lot, right?
Sophie was able to tell
me about order details,
delivery options, but Sophie
could not expedite my order
because Sophie couldn't
access inventory data.
with Agentforce we can give
her the skills she needs
- So we're gonna go behind the scenes,
- We're gonna go behind the scenes.
We're now in an amazing
tool called Agent Builder.
Leah, you know, a thing or two about Agent
Builder, right? You're
a fan of Agent Builder?
- I have built a few agents in my time.
I can imagine across there
we're building them over there!
That's what I hear. That's what I hear.
Agent Builder is where you
can customize the actions your
agents take, whether autonomously
or with humans in the loop.
And you can see here
on the screen the agent
that we just spoke with, Sophie,
you can also see the channels
where Sophie's active, like
SMS like email and voice.
Let's now actually click into topics.
So topics for an agent are basically the
They're the work that you
would assign an AI agent,
like in this case order management.
And if we click into order
management, what's notable is
what we don't see, we don't
see complicated dialogue trees
because this isn't like
creating a chatbot.
Everything you see here
can be crafted using
So we can see the details of the topic,
we can see the label, we
can see the description of
what it is and how it gets used.
We can also see the scope.
Scope is where we can add
guardrails to make sure
that this agent focuses
only on the work at hand.
They don't do more than that.
- Can I just stop there for a second?
Guardrails are so important.
Can you tell Hari, the
guardrails are basically saying
what they can and cannot do.
- That's exactly right.
- Always keeping the
And you put a great point up
there on the natural language.
So for those coders who've
done those decision trees,
the "if then else", this is doing
it all in natural language.
So it even democratizes
building the agent.
because think about it,
We won't want an agent
that works for a retail business
to immediately authorize a
refund, say in excess of $5,000.
We would want them to first escalate
to a human service rep, right?
So if we go back to that screen,
we can also see the instructions
where Saks can actually
expand on the guardrails
that are reflected in the scope.
So here for instance, we see
that Saks wants to make sure
that Sophie is always
confirming in-store pickup
unless a customer requests delivery.
So that's what a topic is.
But to actually execute on a
topic, an agent needs actions.
And within Agentforce
there are a ton of actions
that are already available
that we can add just
using point and click.
And it's important to remember
that if we think back to
that first experience,
there were a lot of actions
that Sophie already could do.
Sophie could tell me about
individual details with my order.
Sophie could also
confirm delivery options,
but Sophie couldn't access
that inventory management
data to expedite my order.
So let's actually go through the process
of adding this custom action.
So we're gonna go now to our
Agentforce actions menu.
We're gonna say new Agentforce action.
And this is actually a
place where I want to pause
and I know Leah's excited about this.
- I do get excited. We're so in sync.
- We're so in sync. So
in sync. So in sync,
- I love this about the actions.
Now just by, I can't really
see you guys but by noise.
How many people in here
have created a flow before?
I know I have. Okay, my flow addicts
or how many people have written some apex?
Woo. Okay, we're gonna work on that.
The point being that
you all have, for those
who have built items in your current orgs,
you can leverage those
actions with agents.
So when you are building out
your AI agents, you don't have
It is leveraging all the
work that you've done before.
And for those that have,
you know, their first time,
they may not have a bunch
of actions already built.
we already deliver standard
out of the box ones.
So you know
whether you have custom ones
or whether you're using standard ones,
you have a place to start from.
That's exactly right. And so
what Leah's getting at there is
that basically everything you built
with the Salesforce platform
can be harnessed by your agent.
And everything you're doing
with the Salesforce platform is
already making your
agents smarter, better,
and more impactful for your business.
So if you go back to that
screen, we'll see that
of course we have a bunch of
different platform actions we
can add as a custom action.
But of course in this case, we
actually want Sophie to talk
Sophie needs access to
inventory management.
So we're gonna actually
create a MuleSoft API,
simple declarative created immediately.
And we're gonna quickly check the input
and output just to make sure
it looks the way it needs to.
that this custom action we've
just added is now available
to us in our list of existing actions.
And so when we're
thinking about everything
that Sophie needs to
complete order management,
we're gonna highlight that inventory API,
and we're gonna schedule
some other actions as well
that will help Sophie more
successfully execute this order, right?
Like schedule in store
pickup, check nearby stores,
and of course adjust delivery method.
So now we get to the really
fun part of this demo.
We actually get to see how it all works.
So we're gonna be in the
admin view here, okay?
And we're gonna actually just
ask Sophie if Sophie can get
our most recent order sooner
and Sophie's thinking,
figuring out next steps,
not unlike an actual human would.
And as Sophie finishes
up, it looks like yes,
Sophie can now expedite that order.
And so we're just gonna go ahead and say
that we want in-store pickup
I've got an event coming up,
I gotta look good, right?
So right after this, I'm
actually on my way to Saks
to pick up that order and you can see
that Sophie is now able to
expedite this order using
that custom action that we just added.
And I wanna pause here
for a second just to kind
of understand what
we're looking at, right?
So on the right hand side
of that screen we can see
of course that the custom action
that we added now makes Sophie able
to execute the order in
the way that we need.
But on the left hand side, we
can actually see everything
that's happening behind the
scenes to make this possible.
We can see Sophie retrieve
data, reason, chain
different actions together.
We also can see the instructions
and guardrails that we added.
And this left hand view is so crucial
because we want everybody at
Saks to have transparency into
and into how they're built
and how they function.
So stepping back from all that,
we've now added the custom action.
We are ready to execute this order.
Let's actually give Saks
like another call back
and see in fact if we
can get our order sooner.
So I'm going to give her a call right now.
your Saks AI assistant.
How can I help you today?
- Hello again, Sophie. This is Hari.
I just am calling about
my most recent order.
I wanna see how quickly I
can get it shipped to me.
- Is this regarding the Saks
Fifth Avenue collection,
cashmere baseball bomber sweater?
- We offer two delivery choices.
I can arrange for it to
be delivered within three
Or you can pick it up
today at Saks San Francisco.
Which option suits you better?
- Love that. I think in-store
pickup today at Saks San
- Great. I've processed the order
and notified the in-store client
advisor about your pickup.
The sweater will be ready for you
within the next three hours.
Is there anything else
I can assist you with?
- No, you know Sophie,
you've been such a huge help.
Thank you so much. Can we give
a hand for Sophie real quick?
- Give up for Sophie.
Can we give it up here?
Let's give it up for Sophie.
So I, stepping back from this,
it's been in less than
10 minutes, we were able
and agent force using
point and click tools.
And we were also able
to do this with trust
and transparency built into
every step of the process.
This is truly the future of AI
and this is what it looks like for humans
and agents to drive
customer success together.
We are so excited to
have you all get hands on
And I just wanna thank
you all for the time
and thank you Leah for having me up here.
Please give it up for Hari!!!
Thank you.