What if you
could have your
That's exactly what
every Williams-Sonoma
customer will have,
thanks to Agent Force.
Every meal perfected,
every table set,
every holiday handled.
Recipes, techniques,
tools, every answer
at your fingertips.
This is what AI
was meant to be,
and how Williams
-Sonoma found the
ultimate recipe for
customer success.
please welcome to the
stage madhav tatai
executive vice president
and general manager
of agent force at
salesforce for a
conversation with samir
hassan chief technology
and digital officer
at williams sonoma all
right well it's nice
to see everybody samir
thank you for being
here great to be
incredible customer so
like the audience like
the video showed um
we got to build uh an
incredibly interesting
use case with you uh
olive which is the
agent on the williams
sonoma website it helps
customers figure out
plans for how they're
going to entertain
uh figure out products
that they might be
interested in um but
the question i have for
you is you know williams
sonoma is across
all your brands this
is like one of the
most canonically human
experiences You go
into one of the stores,
it's very personal.
How did you think
about recreating
that experience
using agents?
Yeah, it's a great
question, Madhav.
Like you mentioned,
let me just tell you a
little bit about
Olive first, just for
context. So Olive is our
Williams-Sonoma agent,
and of course, she
can help with customer
service issues, simple
ones, complex ones.
But what's really
exciting is that we've
been able to sort
of capture a lot of
the essence of what
the brand's about.
Which is helping
with life moments,
helping with
entertaining, helping
with planning,
helping with cooking,
helping with a lot
of these, you know,
like kind of life
moments that center
around what the
brand is great at.
And, you know, as
we started to think
about what does
agentic AI look
like for the Williams
-Sonoma brand,
you know, what we
found and what was
effective for us
is that, you know,
even though we were
talking about AI,
we didn't start
with the AI.
We actually didn't
even start with
the data. We started
with the humans.
We asked ourselves,
what does the optimal
human experience
look like for these
different scenarios?
And I'm blessed
to work at a company
where I'm surrounded
by experts in their
space. We've got,
you know, test
kitchen chefs. We have
merchants that
understand entertaining.
We've got interior
designers with our
furniture brands.
We've got all these
different experts.
And before we even got
into the AI, before
we even got into the
data, you know, I
and my team just sort
of sat down and just
picked their brain.
How do you do this?
When a customer asks
this question, how do
you think about this?
How do you think
about different
types of customers
and personas?
And that was just a
data goldmine. And
that's where a lot of
the sort of light bulbs
popped about, okay,
this is what we can
do, is try to figure
out how do we sort of
encapsulate these amazing
experts and leverage
the AI. And then we
moved to the data,
right? We took some
of these kind of
concepts and ideas,
and we mined and mined
and mined our data.
We've got years and
years of years of
great first-party data,
you know, customer
service engagements,
store engagements,
transactional data,
recipe data,
expert data, test
kitchen data, all
these different
sort of sources
of great data.
And then we
moved to the AI.
Then we started to
sort of take this
and iterate and test
into these experiences.
We sort of socialized
it internally.
We shared it back
with those experts.
We said, pretend
this is a customer.
let's see how this agent
does and we iterated
and fine-tuned and
ultimately what was
born was all of which
was something that
we were very proud of
and we've continued
to build on her
capabilities but we found
that if you start
with the human and not
with the AI and not
even the data you can
end up really creating
some pretty compelling
experiences that
are grounded in the
way that people like
to engage with people.
it's an incredible
experience
and you know
we talk a lot
about AI use
cases, right?
We're seeing an
incredible explosion of
coding use cases,
engineering use cases.
We see customers
using these agents
internally to drive
productivity for their
employees. We see
process optimization.
There's a lot of very
impactful use cases
that are out there
today. But this one,
this one is facing
your end customer.
It's got to be an
agent that you really
trust to be out there, to
have these interactions.
So putting something
like that out
there, you lined out
the steps, drive the
outcome, figure out the
data. But what have you
learned along the way?
Like what's worked,
what's not worked?
What's been something
that you've really
taken from this journey?
Yeah, I think
we've had to break
some of the
conventional thinking
in e-commerce and
digital experiences. I
think historically
e-commerce has been
built for transactional
efficiency. You
think about search,
shop, cart, checkout.
The metrics are
all about how do
I drive conversion
and how do
I do it as quickly
as possible.
Amazon has
built a massive
business on this concept.
And even when
you look at the
LLMs sometimes,
they're very quick
to try to give
you an answer.
And I think what
we found is that
there's a lot of,
especially when you
talk about discovery,
There's a lot
of value in the
experience, in the
back and forth, in
the exploration.
And if you drive
that engagement
and you're not
so quick to jump
to, well, here's
some products.
Here's something
to go buy.
But let's explore
this, right? And again,
this sort of started
with the human
piece, which is, okay,
we sort of probed
some of our best
sales associates
and customer service
agents and folks
and said, like, sometimes
the best response
to a question is
another question.
and and what we found
is that if you can
kind of engage people
and drive engagement in
the exploration in
the discovery it can
drive a tremendous
amount of value right so
you know the engagement
is the conversion in
some cases and we've
seen what with sort of
embracing this approach
and guiding customers
and letting them
explore whether it's
product again or
entertaining or some of the
things that all can
do or some things that
we do with you know
our other brands in
Williams-Sonoma, Inc.,
Pottery Barn, West Elm,
around helping
customers with interior
design and some of
these other areas,
we've just seen the
conversion results
follow extremely
strongly, like multiples,
multiples from a
conversion value
when you can engage
customers in that
experience. Yeah, and
in some ways, that is
replicating what
you do in the store,
which is a deeply
engaging experience
versus a transaction.
That's fascinating.
So last question.
We have a
couple of minutes
here. Sure.
work, things
that don't work.
You are kind of
at the forefront
of these agentic systems.
You've done everything
from building
them, testing
them, scaling them,
operating them.
You're doing all of
these things. So
I would love, I
think everyone here
would appreciate,
where do you
think the state of
the art is? What's
working? What's
not working?
And what are you
really excited
about going forward?
Yeah, I'll start
with the sort of the
positive side, the
things that are working.
and again to kind of
pick up on the theme
that i shared earlier
you know historically
e-commerce is built
around sort of
transactional efficiency
and even if you
think about the
previous generations of
ai and machine learning
it's things like
product recommendations
it's things like
algorithmic scoring
it's things like fraud
protection it's
things like customer
segmentation right
so even when you're
talking about personalization
you're talking
about segment based
personalization it's
like putting people
into buckets and then
building experience
around those buckets
and what's really
exciting about this
latest generation of
agentic tech is that
it's when you say
personalization it's it's
it's true personalization
it's true like one
-to-one like let me
understand your exact
scenario your exact
issue and let me sort of
be able to package up
all the expertise that
we have all the data
that we have all the
capabilities that we have
to create this excellent
experience i don't
think the industry is
really um delivered on
that yet And I think
we're still, as much
as I think we're very
proud of what we built
and what we've done,
I think we're in the
early innings. I think
the best is yet to come.
So I'm very excited
about the potential
of us in the industry.
And certainly with
Williamson, the way
that we're thinking
about it is almost
reinventing digital
commerce and bringing the
human back into shopping
and bringing the
discovery back into
shopping. And I'll sort
of segue to the second
part of your question,
which is, you know,
what's not working
and what we've learned is
chat is great for a
lot of things. It can
be pretty good for
customer service. It can
be pretty good for Q
&A and quick questions.
It's pretty ***-awful
for shopping. I don't
know if you've ever
actually tried to follow
the entire shopping
experience to try to
chat your way to a
purchase, unless you're
buying something that's
highly commoditized.
um and and you know
many of us have sort of
started and the industry
has started in the
sort of chat modal
but i think the future
and where everyone's
going where we're going
certainly is where
we're focused is really
thinking about these
multimodal experiences
and thinking about
where agents start to
break the bounding box
of chat chat windows
and start to think
about immersive visual
experiences talk to thinking
about the intersection
of voice and i
think that's really
where that's where we're
focused right so when
you think about you
know i'll segue from
from the Williams-Sonoma
brand um you know to
another one of our
brands is Pottery Barn
or West Elm which really
focus primarily with
interior decorating
and furniture uh you
know if you're if
you're decorating it's
inherently this very
visual medium right like
you can't text your way
like what do you want
to I don't know I want
a nice looking thing
right like I need to
see it right um and
agents can do this right
like you talk about
this intersection of
large language models
large image models built
on top of proprietary
data with these guided
experiences like
you can really help
people guide and visualize
their way into these
very complex things
and what's really
exciting I think about
how we're positioned and
where I'm very bullish
about it is you know
we all have access
to the same models but
you know what really
differentiates and
what we're seeing where
we're able to unlock
value is our massive
amounts of sort of
first-party data, right,
and being able to sort
of activate that. So
everything from, we
talked about Olive, like
our test kitchen data
to the transactional
data to, you know,
all this data that we
have around that fuel a
lot of the content that
we build and even
the training content
internally and all that,
massively impactful.
We've got, you know,
digital and e-commerce,
but we've also got
stores, We've got
customer service
agents, but we also have
an in-home business,
right? So little known
fact about Williams
-Sonoma is half of our
retail sales from a
revenue perspective
comes from going
into customers' homes
and helping them
with their interior
design. And by the way,
it's free. So if you
ever want to take us
up on it, if you're
moving or redecorating,
just sign up.
We'll help you for
free. You don't have
to pay an expensive
interior designer.
We'll do it for
free. I mean, you'll
buy some stuff, But
the service is free.
But you sort of start
to think about where
all these things
intersect and how you can
activate these
experiences. And it's not
just about digital.
It's about putting this
tech in the hands of
an interior designer.
It's about helping get
into the customer's
homes. It's about
understanding kind of
their sort of space
and being able to help
them visualize it.
And you start to think
about the sort of multi
-channel connection.
And you've got sort
of this proprietary
data. You've got this
internal expertise I
talked about. You talk
about this sort of
channel excellence.
And then for us, you
know, we're also very
highly vertically
integrated. So, you know,
the vast majority of
what we sell is
designed, built in-house.
We source it. We
manufacture it.
We transport it
all over the world.
We handle the
delivery. We handle the
merchandising and
marketing, every single part
of that experience,
right? So AI can start
to touch and really
connect the dots across
all of it because
we've got a line of
sight into the entire
experience. So I'm very
bullish about how
all this is going to
come together for us.
I'm very excited. I
think, you know, what
we've been able to do
in partnership with
Model View and your team
at Salesforce and
what we've been with
all of on the kind of
consumer-facing side,
what we've seen
internally in terms of
Agenda tech, how
you kind of connect
it with data and
other types of both
technology capabilities
and experience.
It's super exciting.
I think we're very
early, and I think the
best is yet to come.
Well, thank you
so much, Samir.
Thank you for being
an amazing customer.
Thank you for
leading the way.
