Retailgentic, presented by the ReFiBuy Media Network and hosted by Scot Wingo, is where Agentic Commerce meets retail innovation. We help retailers and brands prepare for the future of agent-driven shopping with actionable news, expert insights, dxis, forward-looking predictions, and practical guidance on Agentic Commerce Optimization (ACO).
But, you know, there needs to be some type
of, like, trust mechanism that whitelists,
uh, you know, uh, good agents, you know,
versus, like, the malicious agents, right?
So, so there has to be, like, some
industry standard in there, you
know, that all the merchants sort of
adhere to and, uh, stuff like that.
So I hope…
You know, because today we have to do
all kinds of gymnastics to avoid bot
blocking to kind of make this happen
because, you know, you know, we, uh,
because every- everybody-- You know,
it's the dolphin in the tuna net, right?
Everybody gets caught in the
same, uh, in the same net.
Welcome to the Retailgentic Podcast,
where we explore the intersection
of retail, e-commerce, and AI
agents, which we call retailgenic.
Our mission is simple.
We are shining a bright light
on the new agentic AI shopping
megatrend, which will disrupt the
trillion-dollar digital retail market.
The stakes are high.
Google, OpenAI, Microsoft,
Perplexity, Amazon, Apple, Meta, and
Anthropic are all fighting to be the
consumer AI shopping agent winner.
There are gonna be some tough decisions
for online brands and retailers to make.
What happens to checkout?
How should payments work?
Should you continue to invest
in retail media networks?
What does the world look like when
over half of your traffic is not
from humans, but from AI agents?
These existential questions
are the tip of the iceberg.
On this podcast, we're talking to industry
experts, analysts, and influencers
to get answers to these questions,
identify trends and best practices.
This is Retailgentic, your
podcast partner in navigating
our AI shopping agent future.
And now here's your host, Scot Wingo.
Welcome to the Retailgentic Podcast.
On today's episode, I'm excited to
welcome Manil Uppal to the show.
Manil is founder and CEO of Cart.ai.
Thus far on the show, we've had
startups, uh, doing a variety
of things in agentic commerce.
So we've had an agent company like
Generous that's doing agentic gifting,
and then also a bunch of companies that
are doing agentic commerce infrastructure.
These are companies that are building
the tooling that agent builders
are gonna need to build agents.
So companies like Generous would
use these other types of companies.
Uh, examples that we've had on the
show so far are Nakuta and Simple
Checkout, both on the payment side.
Cart.ai
is kind of, uh, is one of these,
uh, tool providers and, or
infrastructure providers, I'll say.
And what they've built is a
system that gives agent builders
the ability to agentically
complete any retailer's checkout.
So it goes and you-- the user says,
uh, "I have found a piece of apparel
using your agent," and then the agent
builder can say, "Okay, go buy that from
site A, B, or C, um, and, and finish
the, the transaction for my user."
This is both extremely
cool and very hard to do.
How are they gonna do it?
Well, stay tuned for my
conversation with Cart.ai,
and we're gonna learn how.
Manila Oopala, founder and CEO of Cart.ai,
welcome to the Retail agentic Podcast.
Well, Scott, happy to be here.
Um, been a fan for a long time.
Watched every episode, and, uh, you know,
finally excited to be on here with you.
Awesome.
Uh, I've been wanting to get
you on here for a while, and
you were like, "Too early.
It's too early."
And finally I twisted your arm
and got you on, uh, on the pod.
So I'm excited, uh, to reveal to
our audience what you're working on.
I think, uh, I think they're
gonna be really intrigued by,
by what you guys are doing.
Yep.
Uh, yeah, we were in stealth mode, uh,
earlier, so that's why, you know, I was
kinda staying, uh, a little bit hidden.
But, uh, I'll tell you this, um, you
know, we started working on this,
uh, early in the year and, uh, it
really felt a very isolated space,
and you were the only voice out there.
So, you know.
So thank you for doing what you're doing,
uh, for the agentic commerce world.
Yeah.
For a while we thought we were crazy, but
then everyone came along for the ride.
So, whew.
It was a little touch
and go there for a while.
So, uh, uh, tell us about your background.
You have a really interesting background
that I think will help set up why
you're working on this problem.
Yeah.
Uh, yeah, for sure.
Uh, so, uh, I think, uh, my
entrepreneurial journey, uh,
started about 12, 13 years ago.
Um, so, um, um, this Cart.ai
is my third venture.
Uh, my very first one, um, uh,
about 13 years ago was, uh, a
company called Lash, uh, Delivery.
Uh, and this is, this is like
the very early days of same-day
delivery where, uh, we were a B2C
e-commerce app, uh, that delivered
alcohol to your door in 30 minutes.
Um-
Oh, like a Drizly.
Uh, yeah.
It was around the same time Drizly
was starting in Boston, where we were
starting in, uh, in, in, in Texas.
At that time we didn't know about
Drizly, you know- Yeah … ships
in the night type of deal.
But- … uh, and then we added food
to the, uh, equation as well, and
we were the only app that could do,
like, food and alcohol in one order,
and it was a pretty complex business.
Uh, you know, I had fun building that.
It was, uh, a hard business, uh, because
we also had our own logistics and our
own drivers and, and things like that so,
you know, um, multi-dimensional that way.
Uh, I think that company got acquired
pretty quick, uh, within three
years by, uh, by H-E-B, which is,
uh, a huge, uh, grocery player,
um, here in the Texas, uh, area.
Uh, and then we took all the
lessons learned from that and,
um, started, uh, our next company
called Delivery Solutions.
Um, some of the lessons
were, "Hey, you know what?
We're good at B2B, we're good at tech."
Uh, you know, not the best B2C marketeers.
Uh, you know, that was a bear,
you know, customer acquisition
from a B2C standpoint.
Not the best logistics
operators either, you know.
It was a noisy part of the
business and, and things like that.
So, so Delivery Solutions was- A
pure SaaS, uh, platform, uh, that
invented the consolidation of,
uh, same-day last mile, right?
So we made a bet, you know, uh,
seven years ago that, um, you
know, same day would be a big
part of retailers' omni-channel,
you know, e-commerce, uh, right?
At that time it was only maybe grocers
and alcohol doing it, and we said,
you know, this is gonna be ubiquitous.
So to all the retailers trying
to implement that, we give them a
standardized API, um, you know, to access
same-day delivery, and on the other side,
we were connected to everybody, right?
The Uber, Lyft, Instacart, DoorDash,
name it, and you know, so we created that
sort of, um, you know, PayPal of like,
you know, same day, if you will, right?
Mm-hmm.
Would it do like a little RFP that
says, "Hey, I need a delivery,"
and people would bid on it?
Or was it just kind of
a, a access to multiples?
It was more like, uh, rules
that the retailers set i- in
terms of, you know, how…
And you know, maybe in certain regions
they want the delivery providers
to do a bid amongst themselves.
Maybe in certain regions they had
a contract with a specific one,
you know, and things like that.
So we had the orchestration engine
and the rules and all of that
stuff to kinda make it happen,
uh, you know, for those use cases.
So, so that company got acquired
by UPS, uh, um, you know, uh, three
years ago, um, and, uh, I had a
bit of an earn-out there and, um-
Did they make you move to Atlanta?
UPS, UPS loves Atlanta.
Yep, they're out of Atlanta.
Um, and then, uh, I think I got out
of that, uh, we finished that in
September, and, uh, and then here
I am on, on this Guard AI journey.
Tell us about the grocery.
Were you at H-E-B long enough
to learn about the grocery biz?
That, that's kind of a, uh, my, my
other podcast partner, Jason, uh, has
been saying I should learn more about
grocery, and I've been resistant.
But I've, I've recently, more
recently been drug into it and,
and it is an interesting business.
It's got a lot of complexity to it.
Yeah.
We, uh, not too long.
I mean, you know, uh, I think it was
about a six-month earn-out, uh, for us.
I mean, when you throw H-E-B resources
and m- money to a problem, it,
you know, things move fast, right?
So we had certain things we had to hit for
scale, and we were able to hit it, like,
quite, uh, quite easily and, you know,
and, and they were very good at their
tech and executing and stuff like that.
Yeah.
Let- let's talk a little
bit more about H-E-B.
So, so I didn't…
You know, I'm in North Carolina,
I didn't really know about them.
In Texas they're big, but everyone
in grocery kind of looks at,
looks at them and admires them.
What…
Tell us more about what makes, what
is unique about this H-E-B thing.
Um, yeah, at least when we were
working with them, uh, I think their
vision, um, was just, you know, beyond,
and, and their pace of execution.
And there was this one, uh,
executive there who, you know, Paul
Taffanad, who's now in, uh, Google.
Uh, and he was the one who made the
decision to kinda acquire us as well.
And, uh, you know, he approached tech as
a whole separate business within H-E-B.
He had his own building where
he had his like, you know,
separate from the mothership
where all the tech, uh, folks at.
And the speed at which they moved, and
the speed at which they evaluated stuff
and made decisions, you know, back then
was, we were, like, just amazed, right?
And even worked with other grocers,
you know, since then, uh, never
have seen that level of speed and,
uh, you know, just like, uh- The
surgical execution, what I saw at
H-E-B, you know, uh, in my time there.
So, so kudos to that
team and, and, and Paul.
Yeah.
Yeah, so you've done, uh, two startups.
You could probably
retire if you wanted to.
Um, and then you're, you're
crazy, you keep going back to it.
Like, what, uh, what, what made you decide
to jump into the AI deep end of the pool?
Um, you know, I, I tried the whole
take a trip and, you know, find
the soul in retirement stuff.
It c- didn't kind of work out for me.
I was like, uh, you know, within
two weeks I was like, "Nope," you
know, uh, gotta have a purpose
in life and stuff like that.
And this problem of, you know,
agentic commerce, uh, this was
gnawing at me f-for a while.
Uh, it, it all started, I, I used to
know a company a couple years ago,
uh, a, a small API that would, you
know, if you're a DTC brand, it would
tell you, you could go to the brand
website and it would tell you where the
product is within the retailer, right?
And, and they had a API for that.
And I, you know, uh, the founders,
I knew them, and I'm like, "Man,
this doesn't go far enough."
Like, why just redirect
me to the retailer?
Just offer the checkout.
And they're like, "No, that's hard.
We, you know, can't do it.
How…
You know, the product's there.
They have their systems.
We can't…"
It was hard.
Yeah.
Yeah, we, we, we call that where to buy.
There's, like, all these
big where to buy solutions.
Yeah, we had one at ChannelAdvisor.
Yeah, the where to buy, but, like,
I wanted let me also buy because I'm
there at the site and, you know, I have
intent, and, you know, uh, and maybe that
retailer is just a mile down from me.
Why can't I just, like, you know?
And that morphed into a bigger question,
why can't I buy stuff if I see it, right?
Um, you can only do that
if you're in a marketplace.
Like, if I'm in Amazon experience,
yep, I can search and buy, right?
But everywhere else, there's
always this giant redirect, right?
From, from where you discover products
to where you buy products, right?
And, and I'm like, we gotta solve that.
Like-
Yeah.
And the redirects are a drag because even
if it's a customer, even if it's like, you
know, I have logins at all the big guys,
Target, Best Buy and all those things, but
I can never remember them, so I just end
up going to the guest checkout every time.
It's just like such a pain in the butt.
And, and, and also it's a big drag on
conversions and, and, and stuff like that.
So that's kinda like, uh…
So I'm like, "You know what?
I'm gonna solve this.
Let's, let's go figure…"
This is like early last year, right?
I'm like, uh…
But then, you know, I've also worked with,
in delivery solutions, like all the top
retailers, you know, in the country and,
and I know what the retailers want, right?
They, they don't want another stack
to deal with, another channel to
deal with, uh, where they gotta
like, you know, uh, maintain that
and, you know, so on and so forth.
So merchants and retailers want
merchant native orders, right?
And, and that's two things, right?
It's, you can be merchant native
from a payment standpoint.
Uh, you, you know, you, you
gotta have their payment systems
clear the payment, you know?
And the other thing is the order
should somehow magically and
surgically be inserted into their
order management system because
so much is tied to that, right?
Your customer service, your post-purchase,
your returns, all of that's tied
to an order, you know, following a
certain flow if it's inserted in there.
So, you know, early last year I was
like, "Well, if I gotta do this, I
gotta integrate with every payment
gateway in the world," because
everybody's using different stuff.
Then I gotta go integrate with every
order management system because every
retailer's on different, uh, OMSs, right?
Like Sterling to like Shopify
to Manhattan, you know, to
Salesforce, Oracle, et cetera.
And I'm like, "Well, I'm probably
gonna kill myself," you know, just
doing integrations and I'll, I'll…
You know?
So I'm like, "I'll never be successful.
I shelve the idea."
Uh, back then.
And it was not until early this year
as, as LLM technologies rose to where
the LLMs could start doing things
like autonomously, I saw the light.
I'm like, "Wow, you know what?
An agent can do this."
And, and if I went to the browser
and placed an order from the
top, it would be merchant native.
It'll, you know, follow their
payment stack, will be inserted
into their order management.
Doesn't matter what they use,
tho- that rail's already built.
So I'm like, "Wow."
Looks
like a human.
Yeah.
So well, bought it back from the ice and
that's kind of where the journey started.
Cool.
When did you get to the point where
you knew, like you were gonna found
the, like incorporate, like you were
like, did you do a proof of concept
with some of the agentic stuff?
Or you just like knew
once you w- to run it?
Uh, you know, even after two, um…
I put too much trust in AI.
I'm like, "Wow, AI is magical,"
'cause, you know, there's so much
press on this and stuff like that.
I'm like, "Yeah, I will
do everything," right?
Like it's-- So we started like, uh,
you know, I put a small tech team
together, uh, and we started working
on it and oh, it wasn't that magical.
Like, you know, it just, you know.
And that was a good, it was a
good and a bad thing, right?
Like the bad thing is like, okay, shit,
we gotta like work on this to kinda
make this work and make this reliable.
The good thing is good, you know, we're
not gonna get like, you know, we're just
not a, not a, a wrapper over some LLM.
We really have to like kinda build
some IP to kinda make this work,
uh, and, and things like that.
So we, we said, "Okay, we gotta go
figure this out, um, and start, you
know, building towards it," right?
And, um, and that's kinda like how
we got started, uh, on that path.
Awesome.
So this year you founded Cart.ai,
and then, um, give us the pitch
so, so help folks understand,
um, you know, what it is you do.
So yeah, so let's talk about agentic
commerce just in general, right?
The general layers and, uh, you know, I'll
just share that one visual, uh, Scott.
Um, you know, I wrote this substack.
Um, and, and you know, i-in very
simple terms, there are three
layers to agentic commerce, right?
Um, you know, up top is
sits the catalog, right?
You, you, you gotta have a feed of some
kind and you know what you're selling,
right, for agentic commerce to happen.
And even before this, uh, the way
I'm defining and, you know, generally
define agentic commerce, uh, technically
is like, is the initiating surface
of where the transaction starts and
the target merchant and retailer
are two different surfaces, right?
Uh, you know, and, and, and agentic
is what like bridges the gap, uh,
you know, between that, right?
So, so obviously the initiating surface
has to have some type of catalog, uh, you
know, feed in them and things like that.
Some
would say the catalog's
the most important part.
You and I differ a
little bit, but like the-
It is, right?
I'm, I'm a little bit catalog, you're
a little bit checkout, so that's okay.
Yeah.
Yep.
Yep.
We can, uh,
yeah.
So,
uh, and even in catalog there
are different ways, right?
There are people who are pull
catalogs, you can scrape, you
can do stuff, and there are
merchants, you know, the whole GEO.
This is also where you play, right?
Help merchants get their catalog to
a point, you know, give it context
so that like AI engines can consume
it properly and things like that.
So that's like what I'm calling like the
push part, uh, of, of the catalog where,
you know, it's, it's, it's, uh, initiative
from, from the merchants, right?
To, to make the catalog like,
um, better so that AI agents can
consume them, uh, better, right?
And then once the catalog's there
and, you know, you gotta have some
type of payment layer to transact.
That's a whole different science, uh,
you know, within itself and a lot of
companies trying to figure that out.
A lot of big players trying to figure out
how payments work in the agentic world.
And then when you scroll down, you have,
um, or I gotta scroll down I suppose.
Yeah, there you go.
Uh, you have checkout, right?
The, the execution layer.
Somebody has to go place that order, uh,
right, and tie it all together, right?
The transaction has to happen, right?
So that's what, you know, is the
execution layer and that's where Card.ai
is, you know, that's where we've
decided to play at least, you know,
for now is, is, is, is tackle that.
And even that layer, uh, you know,
has a couple different methods, right?
Uh, you know, one is
browser agentic, right?
Which is like going to the
websites, y- where the agent goes
to the website as a human, uh, and
navigates it and places the order.
And the other one's like more of an
integrated checkout where there's
some type of connection already with,
with the platform or with the, uh,
with the merchant and stuff like that.
And, uh, you know, you're able to play
out, you know, place the order that way.
A combination of both of that
is the execution layer, and
that's what Card AI plays, right?
To make that piece easy for our customers.
Got it.
So let me restate it back
to make sure that I and the
listeners/viewers understand.
So your customer is building an agent of
some kind, and they, uh, they take care
of all the catalog and all that jazz.
And then, um, and then when the customer,
their customer is a consumer, um, and
when they're ready to buy something,
they then turn it over to you to, you
know, collect the payment information
from that consumer and then navigate
the, the checkout on their behalf.
Yeah.
So basically we are, uh, yeah, we are
the ones who are helping our- customer
add a buy now or a buy for me button, uh,
you know, to their checkout experience.
Uh, I think, um, when you get into
the flows and stuff like that, uh,
they would have to kind of work
with a payment provider to tokenize
the cards and stuff like that.
So you don't wanna do that.
Okay.
So that's-- So they would use
Stripe, show you the Stripe thing,
and it would say, "I'm Link Boom."
You would get the token and-
Yeah.
The, yeah, w- you know, so somebody
will vault it and, and things like that.
And when we are ready to buy, we
will talk to that provider and get
the credentials because, you know,
we'll have to go and place the order
and, you know, things like that.
So, um, so, so yes, we do have an angle
where we have to talk to the payment,
to the payment layer as well to, you
know, to, to get the credentials, right?
Yeah.
Yeah.
And to make it more real, let's say
I built a, a shoe collector, um, a
sneakerhead, uh, agentic, uh, you know,
very verticalized, uh, consumer app, and
it goes through and detects, oh my gosh,
there's this exciting new drop at Adidas.
And, uh, once the consumer starts
paying, I, uh, I gather their payment
credentials, get a token, and then
pass it to you, and you would go
downstream to Adidas and complete the
checkout and, and, you know, cons-
you know, consummate the transaction.
Yep.
What-- So you would capture the
intent of the customer to buy, right?
And you would send that intent over to me,
and I would be the one sort of buying, you
know, on your customer's behalf, right?
Whether it is a g- web agentic,
where I go through the browser
and, you know, go to adidas.com
and finish the transaction, or maybe I'm
already connected to, uh, Adidas using,
I don't know, maybe the new ACP protocol.
We'll probably talk about
that . Or, or you know, or whatever
comes out in the future, right?
Or
AP2.
There's like a, there's a
million of these things now.
Yeah.
Uh, yeah.
So whatever that, you know.
So our job is to abstract
those complexities from,
from your shoe app, right?
So you don't have to deal with that and,
you know, we'll make that happen for you.
This kinda drives me crazy.
We gotta do-- Everyone always accuses
me of just rebuilding ChannelAdvisor
just in a different format.
So in a way you are, you know, you're
re-dr- you're, you're-- Back when you had
all these drop ship companies, last mile
companies, you, you made a lot of value by
putting a clever front end in front of it.
So, so in a way, that's what you're doing.
There's all these checkouts
that are hard to figure out.
If you go figure it out and can build an
agent to go complete it, there's a lot of
different use cases that that will benefit
and enable in our agentic commerce world.
Yeah.
Yep.
Okay.
Yep.
Yep, exactly.
So it's a one-to-many kinda
connector kinda model.
I'm a fan of those, by the way.
Yep, from a checkout standpoint, right?
Um, you know, because, um, you
know, it's always this case, right?
Like, when a revolution starts,
everything's kinda very fragmented
at the beginning, right?
And it takes time for like
things to settle down.
And even when things settle down,
uh, you know, there'll be like three
or four, you know, major protocols
that are gonna win, and then the
rest are not gonna participate.
So, so we are that sum total of
like- You know, whatever that
landscape looks like, right?
Uh, in terms of, you know, who, you
know, wh-which protocols are much…
Today if you ask me, "Hey, I wanna
do agentic universally," the only
option today is web agentic, right?
Because those rails are
already kind of built, right?
In the future, obviously, as more
protocols rise, that's gonna change and,
you know, web will probably become, like,
more, you know, like, like the last mile
or that last 10%, if you will, right?
Maybe…
And, and the rest maybe
front-ended by protocols.
Who knows, right?
But that journey is gonna be hard
to solve, and n- your shoe app
is not gonna gyrate to, like,
every change in the market.
So it's…
You're, you're better off giving it to an
execution layer with a standardized API.
That connection is never gonna change,
whereas we then take care of, you
know, as the market evolves, right?
Yeah.
And the higher surface area, the higher
inventory I have exposure to in my app,
the better expo- spo- experience it is.
You know, my app would stink
if I just supported, you know-
Yeah.
Absolutely
… one protocol or, or, you
know, what, something like
that.
And, and you cannot decide what
you show to your consumer just
based on like, "Hey, do they have a
particular protocol or not," right?
You wanna be as broad as possible.
Yeah, this is one of my big concerns with
both the Google and Amazon Buy For Me.
It's, it's kind of handy
they're named the same thing.
I don't think that's gonna last long, so
I can't, I can't believe I can't believe
two giant companies with all those
lawyers have the same name in there.
But anyway, uh, um, you know, so the
Amazon implementation requires Amazon Pay.
And the Google, all this awesome
Google agentic stuff en-ends at Google
Pay, and those are two that, that
actually have almost no adoption.
So it, it's-- They've, they've taken
this really what could be a huge
surface area thing, and they've put
it through this tiny narrow little
straw and, and it's gonna be…
Yeah.
They may expand it from there, and
maybe it's just a beta or something,
but it, it, it's always disappointing
to me 'cause I have to spend all
my time for my substack finding a
store like that will take Google Pay.
It's actually not easy.
If I'm Amazon, I can see,
"Well, I don't want anybody on
top of me, so you know what?
Walled garden.
Everybody stay out.
I'm gonna do my own thing."
But Google, I mean, imagine
the potential, right?
They have your payment, uh, you
know, in, in, in, in this wallet.
They have all your
passwords, Scott, right?
To every website.
Imagine like-
Just right up in Chrome.
You can- They're just right
up there in Chrome, yeah.
Yeah.
Imagine like you can do agentic commerce,
but now with loyalty embedded in it.
Like whoa.
You know, now what's…
You see a level five in your, uh, in
your visual, you know, your famous
visual of levels of agentic commerce.
They're the ones actually the
best enabled to kind of, you know,
knock that out of the park, right?
So.
Then the thing that's even more
frustrating is the only company that,
that's actually even better than
them is Apple, and they're just like
not even off the starting block.
Yeah.
Uh, yeah.
Yep.
Yeah.
So there you go.
So I think you'll have a, have a
role in this for quite a while.
It's gonna take a while for
everyone to get, get going.
Um, so it seems like the…
And you've kind of hinted at this,
the hard part of this is there's some
really funky checkouts out there, and,
um, you know, grocery's always one.
Uh, auto parts is another one, and
B2B gets to be really, really wonky.
Um, is there, uh, you know…
Are, are you gonna have
some limits to this?
Like when you tell your customers, is
there certain categories you haven't
done yet that are on the roadmap?
Or how, how do you think
about the, the plan there?
Yeah.
In terms of like how we're developing,
uh, the tech, like theoretically,
um, you know, there's-- there
shouldn't be any limit, right?
As long as checkout the checkout.
But, um, I think, uh, our
initial focus has been sort of
e-commerce based, uh, stuff, right?
That's what, you know, we're kinda
training, um, you know, our, uh,
sort of model on, and that's kind
of where mo-most of our sort of,
you know, IP is getting built on.
Um, so if I then, you know, we decided
to go tackle, uh, some B2B vertical
or, you know, uh, very quickly maybe
like grocery and stuff like that, uh,
you know, then I'd probably be like
80, 90% already there, and then I gotta
figure out like what are the 10% nuance
of a, you know, of a, of a grocery
checkout and, and, and things like that.
I would, I would say like most of it
should kind of map over, but then, um,
you know, uh, maybe there's a 10% delta on
the, on the, on the execution side maybe.
Yeah.
Yeah.
And, and a lot of our audience at
Retail Gentic is merchants, so they're
brands and retailers that sell direct
in some format or even indirect.
And let's say they're…
I could kind of see two ways they
would even be interested in at least
making it easier to work with you.
So number one is maybe they'd just like
to sell more stuff and they could say
to you, "Hey, hey, hey Manil, how do I
make my checkout more friendly for you?
You know, is there anything I can do?"
Um, and then another one would be maybe
they look at Amazon Buy For Me kind of a
thing, and they wanna implement their own.
You know, maybe I'm, uh, you know, I'm
just gonna randomly pick one out, uh, uh,
a large beauty retailer or something and,
you know, there's certain, certain things
I can't have on my site for some reason
and, you know, maybe I wanna have a, uh,
a widget on my site that can go buy off of
the direct brands, the beauty brands site.
Um, seems like those would be two
use cases that would be interesting.
W-what do you say to the retailer
that wants to make th- in those
two scenarios, like, like-
Yeah, that's a great question, right?
So, um, on, on the first one, um,
in, in how to make the checkout like
more, uh, agent friendly, if you will.
Uh, today, um, you know, if a agent
lands on a retailer website or a merchant
website, uh, you know, uh, when Cart.ai
goes lands there, we look no
different than some hacker
malicious blo- bot, right?
Like, and we get like put into
the same bucket as, uh, you know,
everybody else and, you know,
Cloudflare kicks in and other people
kick in and there's bot blocking.
And so today there's no like good
way of like whitelisting, you know,
an agent who's, whose sole purpose
is to bring you more revenue.
We're good guys.
We're waving the flag.
Yeah.
We're, we're happy to identify and
tell you who we are and, you know,
ask you, answer any questions.
Yeah.
So there in general, right, with, and
retailers need to participate in this,
but you know, there needs to be some type
of like trust mechanism that whitelists,
uh, you know, uh, good agents, you know,
versus like the malicious agents, right?
So, so there has to be like some
industry standard in there, you
know, that all the merchants sort of
adhere to and, uh, stuff like that.
So I hope, you know, because today we
have to do all kinds of gymnastics to
avoid bot blocking to kind of make this
happen because, you know, you know, we,
uh, 'cause every- everybody, you know,
it's the dolphin in the tuna net, right?
Everybody gets caught in the
same, uh, in the same net.
And to the second point, that's
a, you know, there's some
fascinating use cases now, right?
Think of affiliate models, right?
Uh, where You know, if I'm an affiliate
today and I'm, you know, I'm, you
know, showcasing a particular product
for a particular retailer, it's all
about, like, affiliate links with
certain codes at checkout, and then
there's redirect and stuff like that.
Can you imagine each affiliate now
having their own sort of checkout?
But- Right … but the checkout's
happening with the merchant natively,
but it's happening somewhere else.
I mean, that's, that op- unlocks,
like, so many doors, right?
Yeah.
Keep, keep the customer up at the
affiliate site and deliver a, a paid-for
order, which is y- you end up with
this whole ar- affiliates constantly
argue over, like, who gets attribution.
It's like this endless battle of, you
know, why did last click win, blah, blah.
Yeah, it solves it.
Yeah.
Whoever got the order.
O- also another, uh, you know,
something closer to your channel
advisor background, right?
Like, endless aisles and things like that,
and in the past you had to have all of
those, like, nailed up connections with
like, you know, the, the, the brand, you
know, the suppliers and stuff like that.
Well, in the future,
maybe not so much, right?
Maybe that can be agentic
execution on the back end, right?
To, um, you know, display those, uh, that
inventory and then, you know, when the
order comes, you know, placing the order
to the right supplier, all of that stuff.
You know, that, you know, there
could be a world that, you know, that
got, gets unlocked there as well.
And feel free not to answer any
of these questions, but have
you raised venture capital yet?
Or where are you in that
part of the life cycle?
Not yet.
I mean, I'm c- you know,
cutting the checks right now,
like so fully bootstrapped.
But yes, about to start the process of,
you know, sorta, you know, r- uh, getting
some cash into the war chest, right?
Uh
I found the longer you write
checks yourself, the more
open you tend to be to VC, so.
Yeah.
So, uh, yeah, I, I'm about to start
the process on, you know, uh, getting
something going on that front.
Cool.
And you obviously did a good, you know,
little stint there in the grocery world.
Um, any focus on that,
uh, in particular or?
Uh, it's, it's definitely, you know,
something that we would consider
in the future because it, like,
unlocks a lotta, um Uh, you know,
uh, unlocks a lot of doors, right?
In grocery, if you think of all
those, uh, recipe apps, right?
There were some-- Even before the rise
of AI, there were some, like, good
companies like, you know, doing some
really cool stuff on recipes, but the
translation of a recipe, and maybe
that's, I don't know, 20 SKUs into a
cart at a merchant where you could check-
Always fails
very hard to do, right?
Un-unless, unless, you know, uh, you
know, you had some hard integration or
eventually the grocers themselves got into
the recipe business or either acquired
or, you know, had their own sort of
like, uh, embedded, uh, utility, right?
Now, this can open that up, right?
Now, you know, recipe, you know,
uh, utilities or, you know, or AI
base can live somewhere else and can
easily translate into a cart at any
grocer, right, if you will, right?
Uh, I think, I think that's kinda huge.
I think for CPGs, uh, you know, having
a checkout on their website where your,
your, your, uh, you know, uh, your local
retailer who carries that fulfills, uh,
you know, is, is, you know, now real.
You can make that happen, like, very
easily now with agentic checkout, right?
Uh, so I know a lot of CPGs, you know,
also have a D2C side, and I don't know
how much those have taken off in general
in the industry, uh, you know, because,
uh, it's kinda ha- you know, you, you, you
have a, a, a bulk of your business focused
on distribution and maintaining, you
know, your distribution network, and then
you also have a D2C side and you gotta
fill, you know, figure out warehousing,
logistics, all of that for it.
Well, now your own retailers can, uh,
you know, they can deliver your orders
and you can, you can actually add more.
You know, that's the whole concept of
earlier maybe you were an affiliate and
you were sending traffic to your, uh,
retail partners, now you can be a channel.
You can actually send orders.
Uh, isn't that huge for CPG?
I mean, you know
Yeah.
Could, we used to have, we used to
call it where to buy, and now it could
be a, uh, you know, go ahead and buy.
Yeah.
I don't know.
For, for inspiration.
Or, or buy
from me, you know?
Yeah.
Seems a real popular phrase.
Yeah, go ahead and buy it for me.
Yeah.
And then, um, the…
All these, what we learned at
ChannelAdvisor is, uh, the big
social companies and search companies
wanted to do this kind of thing,
and then they realized, you know,
transactionally commerce gets a little
bit hard in there 'cause there's
a fair number of error conditions.
It seems like, uh, in, in, uh, when
you're talking about grocery recipes,
the, one of the error conditions is you
would have this recipe site, and it would
have encoded in it the products, and
they would try to do this cart injection.
Well, there was…
It was so static, there was no way to
look and see if the things I'm about
to add to the cart are even available.
And, and that, that was like always a
f- and invariably Murphy's Law kicks in
in retail and, you know, that, like six
of the four, you know, four of the six
ingredients are out of stock and you're
like, "Well, this is kind of a mess.
Now I'm gonna have to spend all this
time figuring out a substitute."
H- how do you guys handle
that, like out of stocks?
There's, there's a lot of error conditions
and then even post-transaction stuff.
Is there like a series of APIs and
communication protocols that you and
the, the customer have to kind of walk
through through this whole mechanism?
Yeah.
So, um, so the catalog layer, you
know, that's not instant, right?
There's, uh, some, uh, periodic re-
refresh rate, you know, to that, right?
Your catalog may be a day old,
could be 30 minutes old, you know,
but it's not real time, right?
So the real time angle gets in
when we get on there, right?
When we are trying to go and, you know,
do the execution, you know, whether
it's browser agentic or, you know,
based on, you know, some API connection.
That's when we actually find out
what's available, uh, uh, or not.
And, and, and we also do things like,
uh, multi SKU single merchant, where
you're buying more than one thing.
We also do stuff like multi
SKU multi merchant, right?
Your cart may not be
limited to one merchant.
Like you could buy two things from
one and, you know, uh, another thing
from a, a, a different merchant.
So we have different flags in our API
that says like, "Hey, what happens
if I can't find one thing," right?
"Do I just fail the whole thing
or should I just go ahead if
I can find the other two?"
Uh, you know, a- and,
and, and stuff like that.
Or we can come back and say things like,
um, "Hey, that shirt is, you wanted black,
not available in that size, but I do see
that it's available in red and blue."
Uh, and we can give this
information back, right?
Now, our customer has to build
that experience, you know, uh,
with their end consumers on how
do they wanna show that, right?
Obviously this is not instant
checkout anymore, right?
Sorta, right?
There's some interaction and back
and forth and, you know, so that's a
little bit of a challenge on how they
figure out that piece of checkout.
But yeah, but we, we do
pass that information back.
Yeah.
Are you enjoying variations?
Oh, variants are you know, uh,
People that are new to e-commerce are
like, "Tell me more about these things."
And they're like, "Ah, it sounds easy."
And then they're like, "Wait a minute."
This would always blow the mind of
all the search and, and social people.
They, they were like, "Can
we just do it without that?"
And I was like, "No."
You know, so initially when we used
to test on retail websites, I always
go to the accessory section because
that's like no variations mostly, right?
But as I get into like, you
know, I see stuff like three
deep, hey, what style of pants?
And then what color?
And then like, what,
uh, you know, what size?
And tho- it gets like
progressively kinda harder, right?
That's-
The matrix gets, uh…
Yeah, the square of number of
variations, the complexity, it
goes up to the square of that.
Um, how about, uh, post-transaction?
So the, the checkout does an
order complete, it frequently will
generate a, you know, an order
number and maybe some other message.
Um, it's not gonna have, um, you
know, a, a tracking number at
that point 'cause there's a delay.
But do you then articulate that
back up and it's up to your customer
on how they wanna surface that?
Yeah, yeah.
Uh, so when we place an order and
we're successful, right, on that, uh,
the immediate post-checkout screen,
right, we capture that information
around the order numbers and, you
know, stuff like that and, uh, you
know, relay that back up, right?
Uh, even that's like, you know,
there's so many variations in that.
I didn't know if you knew, but the
number that Shopify gives you at
checkout is very different from the
number you get on your email Right?
So when you check out on Shopify,
you get some hex number, right?
And then when you get the email, it's
a totally different, you know, like
a proper like alpha nu- you know,
uh, a numeric number, and, you know.
So, uh, so there are different
variations within that.
But yeah, we, we get that stuff out.
In terms of like the, all the other
post-purchase stuff, we feel like,
look, the theory is omnichannel's
like pretty mature now, right?
Everybody's got decent customer
service, decent returns, right?
Decent post-purchase and
tracking and, and all of that.
So, uh, as soon as we place the
order on your behalf, you know, that
chain of emails from the retailer
starts coming automatically, right?
"Oh, Scott, here's your order."
And, you know, and then it's like
you place any other order, right?
So we, we let that process be, uh,
in terms of, you know, how it's like,
uh, progressing and things like that.
I know Amazon quoted like a, an email
to intercept that and, you know-
They, they do man in the
middle kind of intercepts-
Yeah … that stuff, yeah.
Uh, we haven't gone that, down that route.
I, I feel like it's best to, you know,
hey, you place the order, you tell
them the order's placed, and then,
you know, retail systems are so mature
now that they do a decent job at like
communicating back to the customer, right?
As to how their order's progressing.
Yeah.
How about some of the
dynamic nature of checkout?
So sometimes you'll put an address
in, and then it'll, you know, it'll
figure out your tax and shipping,
and then it'll relay back shipping
options and c- cost and, and all that.
Do, do you send that back upstream,
and then they have to have a
picker, and then you go back and…
Yep.
Uh, uh, we have two flavors, right?
Uh, and it all goes down to what
experience your, you know, your shoe app
wants to give to the customer, right?
Um, you know, um, and it also depends
on, uh, the style of checkout, right?
Whether it is, you know, um, an
instant like API-based checkout, or
it's a browser-based checkout, right?
I call it like buy now
and buy for me, okay?
Two different concepts.
Buy now is, is basically saying, "Look,
you know, downstream, I'm, you know,
hooked up to an instant API where I
can do things like instantly, right?"
Like any e-commerce checkout, right?
So in a, in a buy now scenario
where I'm connected via API, I
can surface tax shipping like
instantly, pretty much, right?
Like, like any e-commerce
checkout, right, um, you will.
Buy for me is a concept where that nail
down connection isn't there, and I'm doing
like browser agentic and stuff like that.
That takes a bit of time, right?
For me to, you know, go and
browse and stuff like that.
So there we have two options
where, um, you know, we say
one-shot or a confirmation method.
The one-shot method is basically
saying, "Hey, uh, you tell us that go
checkout and, uh, give me the cheapest
shipping option possible," right?
"And, um, and give me a range
on the tax or whatever," right?
Like, you know, an approx.
But you're not going back to the user
for a second confirmation, right?
You're just saying, "Just go
and give me the cheapest and,
you know, make it happen."
And the confirmation method is, "Go
there, give me all the different shipping
flavors that are available and their cost.
Give me the exact tax from the retailer's
website, and I will go back to my
customer for a double confirmation or…"
You know what I mean?
So, uh, that could be the other, uh-
sorta, you know, uh, so depending
on what experience you wanna code,
you know, we, we kinda support
sorta both fla- flavors, right?
Um, um, you know, up at the beginning we
can give you a decent idea of shipping
and, you know, there are so many APIs you
can get, like tax information, you can be
pretty accurate if you wanna do one-shot.
Uh, but yeah, we support
sorta like both the flavors.
Cool.
Well, I can't poke holes in it, so you're
ready for, you're ready to go pitch VCs.
Cool.
Well, um, so thanks for sharing
kind of a lot of details about
what you guys have built.
I just obviously find this
endlessly fascinating.
Um, let's, let's kinda pop
back up a layer or two.
Um, and you kind of previewed,
you know, obviously we're, we're
recording this, you know, less than
two weeks after ChatGPT dropped their
big news, and it was kinda funny.
So the week before that, the, the AP2
came out with Google and PayPal and
this whole army of people, and it-- I
read that one and it was so complicated.
I, it kinda like, it felt
like a committee had built it.
And then, um, you know, the, the, the
ACP seems more tight And people that
kinda like understood the mission a
little better and, and it, it has a
narrower mission, I'll just say that.
So, so the, the AP2 is kinda like any kind
of agentic use case, you know, very, you
know, if, if we were to kinda like take
the, the union of all use cases and build
an API for that, and then this one was
just like, "We wanna help sell stuff."
Physical products in, in a, you know,
kinda like where you and I are talking.
W- what do you make of it and,
and what do you think about that?
Is that gonna become the default standard?
What, what's gonna-
It, it's, I mean, it
wa- yeah, it was huge.
Uh, you know, everybody expected it.
You were calling it for a
long time, obviously, right?
Like, "Hey, it's gonna happen,"
and, you know, uh, whatnot.
Um, so I think the good news is, hey, it's
great for the industry, number one, right?
Uh, you know, uh, it, you know,
it, it gets people used to placing
orders on off, like merchant
properties and stuff like that.
So in general, it gets
users comfortable, right?
In, in doing, in transacting
commerce that way, right?
So it moves the needle
for like all of us, right?
Whoever's in the agentic sorta
like commerce industry, right?
Uh, now if you break it down on what, what
ACP is and what they're doing, I think
the discovery is agentic for sure, right?
Uh, the way, oh, you know, ChatGPT figures
it out and, you know, things like that.
Uh, and, and, and I think
that's, that's huge, uh, there
on, on the discovery level.
The checkout is not really agentic, right?
Like checkout, I mean, it's not
agentic commerce in the sense that,
you know, they have put out a spec.
It's an API basically, right?
So if you wanna do the checkout, you
as a merchant have to code to that API.
And, you know, so, so, you know, in
the, in the figure I showed earlier, it
is that integrated checkout basically.
It's, it's, it's via, via protocol.
Yeah.
I'm- I've had this argument several…
Their argument, and I'm not defending
it, I'm just, and you probably
know this, but I'm sharing it for
listeners, is, um, the agentic part of
ACP is the, the little orchestrator.
They kinda view that as the agent
that kinda like ties the, you know…
So you've got a merchant, you've got
the payment, and, and, and then you've
got, uh, ChatGPT in the middle there.
So you've got basically kinda
like, you know, f- three to four
modules talking to each other.
It, it i- it is a little
complex, uh, until you kinda
like get your head around it.
That, that's what they would say.
But, you know- I think a purist
would say it's non, not agentic.
I agree.
I agree with the, the, the
relatively pure vi- version.
It's sort of like, uh, you know,
you need net- network externalities.
It's like a fax machine.
You have to have the merchant also code
to that protocol and things like that.
But, um, you know, uh, I
think a lot of merchants and
retailers wanna get on ChatGPT.
So, you know, that's a big draw, right?
Like, what, what was it, Scott?
800 million users, uh, MAUs or whatever.
So, you know, that's a big draw,
and maybe that could force some mass
adoption where, you know, uh, Shopify
coded into ACP, maybe Woo, Magento,
everybody else follow suit, and,
you know, all the retailers follow.
It could be the de facto
standard in the future.
That's, you know, it's possible, right?
But, you know, maybe, you know, there
are two, three other competing protocols
that may come up as well, right?
Uh, that, you know, people
may have subscribed to.
Uh, there could be individual
retailers who say, "Uh, we're not
doing any of that, but we'll expose
our stuff using our MCP servers."
So there could be like a whole bunch
of those, uh, that are out there.
And then for everybody who's
left, there's always, you know,
browser agentic, uh, right?
Uh, you know.
Yeah.
It seems like Amazon's gonna
be out there for a while.
I don't know if you work with Amazon,
but they, they seem like they're,
they're kinda saying, "Nope, we're
gonna be over here all on our own."
So it's gonna be a process, and I wanna
give you an analogy from my, you know,
food and, uh, you know, food delivery days
in my, uh, in my, uh, first, uh, startup.
So the, you know, uh, and not
only us, but even the biggest
players were doing this, right?
Uber, DoorDash.
So you enter a new market, right?
You want 1,000 restaurants
on your app, right?
It's impossible and super hard at the
beginning to go talk to 1,000 restaurants
and a thou- you know, and 1,000 locations
and get their menus and, you know, uh,
have them to buy into, uh, your, uh,
check, you know, uh, channel, right?
Yeah, all of that stuff.
So what do you do at the beginning?
You use existing rails.
That's what we all did, right?
We scraped the menu, put
it on the app, right?
And the rails already existing
for placing a to-go order, right?
Anybody can call a restaurant, place to…
So when the order came into the app,
we had our drivers call the restaurant,
place the order as a to-go order, and
they, and they paid using these, uh,
cards that we issued to them, uh, that
were API programmatic and whatnot, so
we control how much money was in there.
And then they would walk into
the restaurant and say, "Hey,
where's my to-go order?"
And they would pick that up and deliver
that, you know, to the customer, right?
That's kinda like how the food
delivery industry sorta, you know,
started at the beginning, right?
And we all did that.
And the whole point I bring that
up is we were using the rails that
existed back then to start, right?
Uh, which was the r- you know, which
is the basically the to-go order rails.
Same thing on agentic commerce, right?
The rails exist today for humans to browse
a website and place an order, right?
That's where the browser and merchants
and retailers have-- don't have to
do anything, uh, on their end, right?
The agents will go and do it,
and those rails exist, right?
So, uh, th- that's where the
browser agentic sorta has a place.
And as the market evolves and
protocols evolve and people start
adopting protocols, and then you
have, like, certain deals with, like,
certain platforms and retailers,
stuff will start shifting, uh, you
know, in- into that world, right?
But the whole thing is
gonna take some time, right?
And, and, and I think that's what
our mission is, is it doesn't
matter what happens, right?
Like, you know, we'll give a
standardized layer to, to our
customers so th- they don't have to
sort of figure this world out, right?
They have a hard enough job doing
their vertical SaaS stuff, right?
So-
One other big picture question I
think you have a unique view of is
I, I spent a fair amount of time
talking about the horizontal players.
You know, so, so ChatGPT,
Perplexity and so on.
And they're, they're horizontal in all
the-- they're trying to be in all the
categories or most of the categories.
Um, and then I imagine the
customers you're talking to,
uh, are very verticalized.
And it's interesting, and I, I, I've
always had it on my list of things
I wanna write more about and, uh,
there's like new ones every day
and I try to keep track of them.
Um, so I'm gonna do a
little bit more on that.
But, but, you know, what do
you think about the horizontal
versus vertical kind of argument?
Yeah, I think we are, uh, a, a lot
of what I'm seeing at least for, you
know, our potential customers and stuff
like that are sorta vertical engines
that, uh, you know, um, that have an
expertise in a, in a certain area, right?
Like I see a lot of, uh, these
fashion apps out there, right?
That, uh, that I'm talking to where, you
know, they have their own unique taste
on like giving like fashion advice and
recommendation and, you know, help me
choose a dress or help me do this or, you
know, I wanna look like this designer or,
hey, take my picture and, you know, tell
me, you know, all of that stuff, right?
So, uh, you know, and, and then somebody
like us could then like, you know,
enable the purchase layer, you know, uh,
for, for folks like this and, you know.
But they could be- You know, there
are other ones in other industries,
you know, that are also like…
You know, because look at
recommendation right now, right?
ChatGPT, LLM-based, it's okay, a
little bit generic, if you will, right?
It's agentic and cool, but gen- you know.
But there could be experts
in a particular area, right?
"Hey, I'm B2B tools expert," right?
Like, you know, and, and, and people will
go there and listen to that and, you know.
Uh, and, you know, there would be
like, you know, even discovery will
get into some expert channels, right?
Like, uh, you know, and
that'll probably happen, right?
ChatGPT is not gonna
solve everything, right?
There, there's gonna be a rise of, um,
you know, pods of like sort of, you
know, expertise and things like that.
And there, there's always gonna be
influencers who, who, you know, who are
experts in a particular area that, that
are gonna have a lot of eyeballs, right?
Uh, and, and, and things like that.
So this is all the vertical areas sort
of we aspire to, uh, to play in, right?
But the thing that excites me the most,
Scott, is You know, if this technology
became ubiquitous, imagine, um, enabling
new services for commerce, right?
Uh, you know, so today, you know, we
can help, you know, where you can see
and can't buy, for example, right?
The Pinterest, the Google Shopping.
Well, now you can put buy now
buttons with this technology, right?
But can you imagine like, "Hey,
I work in this office and, you
know, I spend all day on Slack.
Wouldn't it be cool if I can just like
order my groceries off of Slack too?"
Right?
Well, now, you know, now you can make
these apps that had nothing to do
with commerce, but you know, people
spend a lot of time on these apps.
You can make them more convenient and
make them commerce enabled, right?
Now, the technology's
coming up to do that, right?
And that's exciting, right?
Yeah.
You could have, have, you know, "Hey,
@HEB, um, I need these kind of groceries."
Boom.
Amazing.
Cool.
If, um, if folks want to learn more,
uh, so y'all's website is cartai.ai,
and then, um, and then
you're, you're on LinkedIn.
Is any- anywhere…
Oh, tell us about your Substack.
So where's the best place
to find your Substack?
Yeah.
It's, uh, @cartai, uh,
you know, is our Substack.
And, uh, website is cartai.ai,
'cause cart.ai
was very expensive.
So, you know, so we went with cartai.ai.
Thought, you know, got to live.
Yeah.
This is the benefit of not
being a consumer customer.
You can…
B2Bs will find you even if it's
a little, little different.
Yep.
And then, uh, you know, we're, um,
available on, uh, LinkedIn as well.
And, uh, you know, to anybody who wants
to chat, like agentic commerce and
checkout, you know, whether it's from the
vertical app side or the merchant side,
stuff like that, always, you know, ready
to geek out, you know, on that topic.
Well, Anil, appreciate you taking time.
Uh, we're recording this on a Friday,
so that's, I know it's always hard
in front of a weekend to, to, to
bang out, have the energy to do this.
But you seem like you're not tired
at all, so, uh, uh, kudos for that.
And good luck raising funding if you
decide to that go path and, that path.
And we'll be keeping a close eye and maybe
have you back in six to nine months to get
a little update on, on what you're seeing.
With gratitude, Scott.
Thank you.
Uh, thank you for having me on here.
And you know, I love what you
do and, uh, you know, I…
As soon as you drop your
stuff, I'm always watching.
So, you know, thank you for bringing
the, shining a light on this area,
you know, for all of us, right?
You're, you're a beacon in that, so.
Appreciate it.
Tell all your friends.
Tell all your agentic friends.
Follow, rate, and share to stay up
to date with Retailgenic, your guide
to the future of AI shopping agents.