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).
We've made some progress But when
you think about any one of those
announcements, I don't think
that they fundamentally solve
the core challenges that we have
in making agentic commerce work.
Like, I think we-- Every, like, every time
and every time one of these comes out,
it's like, "Oh, we've solved the problem."
But if you really start to dig
down into the details of it, it's
like, yeah, we solved, you know,
at OpenAI working with Etsy.
Okay, that's interesting.
But the big problems of, like, how
do you connect every retailer on the
planet into this agentic commerce
infrastructure, like, we're just not
that much closer to that, uh, objective.
Welcome to the Retailgentic Podcast,
where we explore the intersection
of retail, e-commerce, and AI
agents, which we call Retailgentic.
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 this episode of Retailgentic.
My guest on this episode is Matt Nichols,
general partner at Commerce Ventures.
If Matt's name sounds familiar, it's
because this is his second appearance.
He was previously on the show
back in mid-June, and that episode
has continued to be one of our
most popular podcast of the year.
With only six months transpiring since
he was on, that may be a relatively
short time in typical, uh, industries.
But here in AI, it feels
like about ten years.
So we thought, given that it's been so
long, it would be great to have him come
on and share what he is talking about.
In that time, we've seen a huge
step function and activity around
agentic commerce, payments, and
much more, which are centered to
Commerce Venture's sweet spot.
And as a GP at the firm that sits in
the middle of this, so on one side, they
have all the innovative startups that
pitch them and, uh, ask for funding.
And then on the other, their LPs or
investors in their fund are Fortune
one thousand companies that come
from the retail world, e-commerce.
payments, uh, and financial
services, as well as, uh, brands.
So, um, they, they see both
sides of this marketplace.
On one side, they see the supply,
which is innovative startups and the
different things they're working on.
And on the other, they see the demand,
which is these Fortune 1000 companies that
are looking to partner, uh, be customers,
or invest in those different innovations.
So in this episode, we cover a ton of
topics, such as the four capabilities
merchants need to add or enhance
to prepare for agentic commerce.
This is Commerce Venture's, uh,
framework that they've come up with that
they share with those Fortune 1000s.
Why a traditional Google feed
mindset is not gonna cut it in
this agentic commerce world.
And finally, we go over Matt's
predictions for 2026 and beyond.
We thought this would be a great
end-of-the-year podcast for
you heading into the holidays.
So it acts like a bookmark, uh,
a bookend, so this is kinda gonna
be our capper for the holidays.
This is gonna be our
last episode of the year.
I, I know that's, uh, a little
sad, but it's been a great year.
This is, I think, our 33rd episode that
we've put out, so we really appreciate
everyone listening and sharing and rating
and all the things you've been doing.
We really appreciate that.
And we thought we'd
end the year on a bang.
This is a doozy.
I think you're gonna really enjoy it.
Um, one housekeeping thing.
In the podcast, we reference a
Commerce Ventures presentation
that came out in October.
You can find that on the Retailgentic
substack that, uh, announced this podcast.
It'll be front and center, uh,
and then also it's in the show
notes for the pa- podcast.
So in your favorite podcast
listening device, uh, you…
uh, and technology, you should be
able to find a link to this if you
wanna pull it up on your mobile.
So, um, we think you'll enjoy
looking at those visualizations
while you listen, uh, or afterwards.
Uh, don't do this when
you're driving, of course.
And I think it'll help explain some
of these topics that they have to boil
down, uh, at Commerce Ventures for
these larger businesses on the LP side.
So enjoy the podcast.
We appreciate your listening.
Happy holidays, merry Christmas,
and happy agentic commerce-ing.
Matt Nichols, welcome
back to Retail Agentic.
Excited to have you.
Excited to be back.
Yep.
So you were our third guest, I
think, if I'm doing the math right.
Uh, and it was way back on June 12th,
which feels like seven years ago now.
Um, I was gonna say, you haven't
aged a lot, much since I saw you
last- … but it's only been six months.
I would hope
not.
Yeah.
Um, that was our fourth episode, and,
uh, here we are six months later.
We're on episode 30.
Uh, and so a lot has happened here
in our world of agentic commerce
since we last talked back in June.
Um, on that show, you predicted a
lot of good things, so, um, many
of them have already come true.
Isn't that crazy about this, where we
used to make e-commerce predictions and
we had, like, five to 10 years to see
how they, how they gestated, but now,
now they just kinda, like, pop right out.
Um, so I thought this would be a good
episode to kinda end the year on.
Um, also, I know venture capitalists
out there in Silicon Valley tend to
have big egos, but, uh, I thought
I would feed yours a little bit.
Your episode was one of our most popular.
So, um, you know, I don't ever get
between the, the audience and what they
want, so the audience wants more Matt,
so we're gonna give them more Matt.
So I thought, I thought we could talk
about a little bit of recapping what
happened in '25, kind of- From the
lens of what you thought back in June.
I think it's really inter- interesting
to kind of hear how investors think
and, uh, you're ob- obviously big in the
space, so you also think a bit like a
retailer, so that's kind of a double win.
And then we'll look at, like, next year
and maybe even a little bit further out.
Sound like a plan?
Let's do it.
Last time we went really deep in your
background and Commerce Ventures, so
maybe do the thumbnail version of Commerce
Ventures, and then if people want more,
they can go back to episode four and, and
listen to the, the full, the full story.
Sounds good.
So Commerce Ventures, we're a venture
capital firm based in San Francisco,
been around for about 12 years, investing
out of our fifth fund, um, all focused
on commerce tech, as you can imagine
given the name, uh, which is really
retail tech and fintech or infrastructure
for retail and financial services.
Uh, my background had been in venture
for 20-ish years, but took a five-year
hiatus where I actually ran a retailer.
So I ran a direct-to-consumer e-commerce
business in the jewelry space.
Um, and, you know, I think our
thesis and our fund's focus almost
map- maps up perfectly to all of the
questions, problems that are happening
around agentic commerce, right?
It's a, it's a commerce infrastructure
problem, it's a payments problem, and
it's a, the sort of coalescing of those
two things that makes this category…
this problem, question,
opportunity really interesting.
So it fits really well with our fund.
Share a little bit about, you know, to the
extent you can, about some of your LPs.
Yeah.
So, um, as a result of having that focus,
we have some of the Largest, um, folks
across retail and, um, financial services.
So we've got, um, two
of the top 10 retailers.
About three of the top 10 retailers in
the US as LPs in our fund, uh, acro-
across a couple different categories.
Um, and then the financial services
side, we have some of the biggest
payment processing companies, credit card
companies, banks, and sort of financial
intermediaries that you can imagine.
So, um, I think the, the combination
of the insights that we get from those
LPs as well as, you know, the, the sets
of retailers, brands, and folks in the
financial services ecosystem that we
talk to outside of our LP base is sort
of where we draw our inspiration for
where we invest and where we try to use,
try to use to, to help our companies.
Yeah.
In my head, you're like
a two-sided marketplace.
So on one side you've got all these large
retailers and payment companies that,
that are, you know, trying to figure
out the world and, um, looking to you.
You know, they, they've put dollars
in your company to help them have
a front row seat of what's going
on, um, on, on the innovation side.
And on the other one, you've got all
these, um, you know, startups that,
that, that want to, uh, you know, benefit
from your, not only your capital, but
your knowledge about the space, and
then also get connected to your LPs.
So you, you get to kinda see both sides
of the equation, which is pretty neat.
Exactly, and if we're doing it right,
we're kind of being able to use
one side to help val- drive value
to the other, and, and vice versa.
So I imagine you've been very busy the
last six months because, uh, every,
uh, every retailer brand and PSP out
there is trying to figure this out.
And, uh, it went from, you know, this
is not a thing to this is a big thing.
And it's been really interesting to
see the, um, you know … Just from my
little corner of the world, it's been
quite frenetic and, and very exciting.
I, I like, I like this level.
It reminds me of the very early days
of marketplaces and e-commerce and
selling on Amazon, and everything
was just, like, very exciting
and new and lots of activity.
Yeah, it's actually, it's interesting.
I, I, I think about our focus.
Like, fintech has always been- A
board-level discussion at financial
services companies for a long time.
Like, it's that important.
And I think in the early
days, retail was, right?
In the early days of e-commerce,
if I'm a physical retailer,
what am I doing with e-commerce?
But it's become, had, up until this
agent of commerce topic, you know, it was
important in your infrastructure and who
you're using, et cetera, was important,
but it wasn't always as, like, board
primary, like, critically important.
And I think what's so interesting about
this cycle is that, again, has come
back full circle to boards asking their
CEOs and their CMOs, et cetera, about
what they're doing about this topic,
which has made it exciting for us, for
sure, and for our companies that are
providing tech into those companies.
You know, since we've talked, we've had,
uh, AP2 come out and three other payment
protocols that I can't keep track of.
Um, MCP was out when we last chatted, but
it's got a, you know, it's become this…
A lot of people are trying to
build all kinds of things on it,
maybe correctly, maybe incorrectly.
Um, and then there's these variants of it.
There's, like, one that has
a UX and people are trying
to jam a checkout in there.
All, all kinds of stuff
going on around MCP.
Um, we have, uh, let's see, uh,
ChatGPT obviously launched instant
checkout and the corresponding ACP,
which was, which was pretty exciting.
Um, it was kind of rumored back in
June, July, and then 9/30, I think
it was, was when that dropped.
That was crazy.
And then we, we have Gemini making
big moves, not only with Gemini 3,
but prior to that, they had their
shopping announcement, um, you know,
where they, they can go buy for you.
Um, and then now we had this kind of
surprise PayPal Perplexity move where
Perplexity just redid everything basically
right before Cyber Monday and came out
with a V2 of their old agentic thing.
What are … What kind of reflections
do you have on all that activity?
I think so.
I think it's fascinating because it
We've made some progress But when
you think about any one of those
announcements, I don't think
that they fundamentally solve
the core challenges that we have
in making agentic commerce work.
Like, I think we- every, like, every time
and every time one of these comes out,
it's like, "Oh, we've solved the problem."
But if you really start to dig
down into the details of it, it's
like, yeah, we solved, you know,
at OpenAI working with Etsy.
Okay, that's interesting, but the big
problems of, like, how do you connect
every retailer on the planet into
this agentic commerce infrastructure,
like, we're just not that much
closer to that, uh, objective.
I mean, I, I think…
So having some protocols is useful, right?
That's the first step, but
it's not the core big problem.
So I think one of the things I'm r-
thinking about are, like, our focus on
what the big problems are and trying
to find companies who are solving those
problems hasn't changed all that much
in spite of all of that activity- Yeah
which I think is strange, honestly.
Yeah, it kind of helps define
one side of the problem.
So prior to this, there was all these,
like, people that thought agents are
gonna talk to agents, and this, that, and
the other, and, you know, there's gonna
be all these other mechanisms in there.
Now, now at least you know one endpoint.
Uh, you know, this has got enough momentum
that if ACP is gonna be an endpoint
for most retailers and be a priority.
And I think for me, like, uh, I think
it will, I think it probably will be.
Uh, whether at that emerges as the, as
the ultimate, um, standard, I don't know.
But again, I think, I think the
standard matters less than the
infrastructure to implement the standard.
That's the hardest part.
Like, the retailer side that will
actually implement this standard.
Again, it's easy for Shopify merchants
and it's easy for, call it the top 20
retailers, but if you really wanna make
this- Work for, uh, you know, all of the
merchants or nearly all the merchants,
which I think you almost have to to
make the consumer value prop be there
or give them the confidence to buy.
That's the big challenge
that hasn't been solved.
So I think about, like, where, where
I think the big, like, opportunities
and challenges are, um, I, I still
bucket into three things we may have
talked about on our last podcast.
But, like, I think it's retailers and
brands trying to figure out how to bring
AI magic to their sites to basically
stop somewhat of the inevitable of
people moving to ChatGPT or Perplexity.
Like, I think ultimately if you can keep
someone on your site, that's better.
So how do you bring AI magic
to your site to kind of obviate
the need for all of this stuff?
But then, you know, ack- acknowledging
that some amount of your traffic is
gonna go be shopping for you, through
you, using these AI platforms, it's
really how do you get aggressive about
using catalogs to get your product out
there and visible by the AI platforms?
And then ultimately, how
do you enable this, these
transactions to actually occur?
And I think we haven't, like,
solved those last two yet.
So you guys put out a really
good, um, you know, uh, I think
you called it a newsletter, but
it's a … How many pages is it?
It's a 20-page PDF, so it's a, a
light … It's not a mega deck.
It's a, a tight, tight deck that
just, you know, very straightforward.
Uh, and I love the title.
It's pretty simple.
It basically says Agentic Commerce,
so it's just Yeah, this is, we're
gonna, we're just gonna say what it is.
Um, and two, two slides I thought
were really kind of topical to
those, uh, were slide 11 and 12.
And, uh, for those of you listening,
I'll talk you through this.
Um, and if you're watching, uh, this
is incentive for you to go over to
YouTube if you wanna see the slides.
Uh, I found these…
I think you guys put it on all your
socials there for Commerce Centers.
It was-- I remember seeing it on
LinkedIn and then, uh, I believe
someone saw it on X, but I'm,
I'm sure it's on all the places.
And if anybody wants to get a copy of
it, just hit me up at matt@commerce.vc.
Happy to send it to you.
So, uh, you know, the title of
the slide is From Static Feeds to
Intelligent Transaction-Ready Data.
New feeds must help AI
understand, recommend, and
complete transactions seamlessly.
And there's kind of two columns.
There's the legacy, uh, Google
shopping feed column, and then
there's the AI-centric feed.
So walk us through some of the concepts
on this and, and, you know, what is
the, what's the thinking o- behind your
point of view of what's going on here?
I would start with, like, what's the
point of these two things, right?
You know, the point of the legacy
shopping feed was effectively to
populate a product detail page.
Like, that was, for the most part, what
the goal was, or it matched up pretty
precisely with that, which is honestly
a limited amount of information that
a consumer can consume on one page.
Um, which is, was, which resulted
in a relatively narrow feed.
Um, I think what's so fundamentally
different about this, we call
it the AI-centric feed or feed
that you might send to ChatGPT or
Perplexity, is that one, you're no
longer constrained by that, right?
You're no longer constrained by a single
PDP and what could be shown on it.
But probably more importantly
is that in almost every case
for them, more is better, right?
More information about how your product
could be used, more information about
how the product was produced, um,
more, you know, any information that
is available about the product can
be delivered to s- answer a question
that a consumer might have that you
would never show to every consumer.
So I think that's, there's a fundamentally
different, um, need there, and I think
that will result in data feeds that are…
I don't know what the number is.
Curious what you think, Scott.
But, I mean, this could be one to
three orders of magnitude more data
points than the original shopping feed.
Like, it's just gonna be so much bigger.
And I think there's the
size that's interesting.
There's the, you know, what types
of data you could include in it.
There's also this idea of, like, maybe
even share things that are somewhat
proprietary that you would never show
to a consumer- That might help these
AI agents make a recommendation.
For instance, "I'm Nordstrom, how many
of these sweaters did I sell last month?"
Right?
And if you find a way that you can
keep that confidential, et cetera,
there's some interesting data points
that would give the AI the, the sort
of confidence to recommend that product
versus one of your other products or
a product of your competitor that I
think is totally different than you
would see in a Google Shopping feed.
I kind of liken it to Google
Shopping, uh, feed is a dirt road.
Uh, I'm from here in the South,
so we have a lot of dirt roads.
And then the, you know, what ChatGPT
wants is a 20-lane kind of dual path
highway, especially if you're gonna
do the instant checkout, because now
you've got a whole another set of, you
know, bi- bidirectional things going on.
So, so it is, like, significantly
orders of magnitude more complicated.
And I think you also…
I mean, I think there's, there's
cases for some retailers that you
don't want to share everything, right?
It's proprietary, and I think which
really makes the need for this even
bi- even greater because I think that
retailers are going to effectively
shut down the scraping of their site,
or should shut down the scraping of
their sites and replace it with a,
you know, a feed that they control.
Because trying to control what
you s- what you block is way
harder than creating this feed.
So I think you'll see this move towards
really blocking everything, but replacing
it with something where you control the
levers pretty precisely about who you're
willing to give information to and when,
and you kind of need that feed to be
able to block it without going dark.
Yeah, two good examples are, um, ChatGPT
and the feed spec really wants a Q&A.
So they want you to do some Q&A, um,
and they want it to be real, you know.
So, so m- most sites don't have
that, so we, we've been helping a
lot of people figure out how to make
that, that, the, the real stuff.
Um, but then they also want some reviews.
Um, and that's where we're seeing a
lot of retailers say, "No, I think
I'm gonna draw the line there.
Go, go get reviews some- somewhere else."
But, you know, uh, but it, you know,
the downside of that is if there's some,
some, you know, rando on Reddit from 50
years ago that's really mad about, you
know, your jewelry company, we added this
exact example where someone on Reddit,
we were dealing with a large online, you
know, jewelry company, and someone on
Reddit had said, "This is fake gold."
Um, and they just, you know,
because they're on Reddit, there's
no transaction verification,
and they can just say anything.
And they had, they ha- they went and
responded, and it kinda helped, um,
so at least when it was surfaced, if
someone did the research, you could
figure out that it was fake news.
But who's gonna, w-
who's gonna go do that?
So the downside of not providing
reviews is you don't, you can't
tell your side of the story, but
that's like, that's a, that's a
decision some retailers are making.
They kind of view that review
data as pretty proprietary,
and they don't wanna share it.
A- and I think, uh, ultimately, I think,
I, I also think there, uh, right now it's
just, "Let's just send the information."
I think that there will also be an
economic arrangement around feeds,
that we haven't gotten there yet.
But like, there are certain, there are
certain tiny little merch- retailers
that might pay to have their feed
ingested by ChatGPT, and there's
probably really, really, really big
merchants that could extract dollars
from ChatGPT to, to get their feed.
So like I don't- it's not clear which
way it goes, but I think there will be…
Because in every case you're not getting
traffic back, I think there may be also
an economic arrangement that spills out
of what information you share, um, that
is still, still too early for that.
But I think that's a,
a distinct possibility.
Yeah.
So for example, maybe ChatGPT says,
"Well, you're currently at our, our rack
rate is 5%," whatever it is, X percent.
"And if you share some idea of sales
rank, we'll knock a point or two
off of that," because, you know,
the algorithm will get better.
It'll make everything
better and yeah, okay.
And then I think there's also this,
uh, this idea of, you know, it'll be,
there's, there's the generic feed,
"Here's all my products," but now here's
the feed- Ranked, sorted, et cetera,
for females between 30 and 45, right?
So there's a bunch of other things
you can start to do which go above
and beyond just the characteristics of
the products that I think I s- assume,
assume ChatGPT would love as well.
And if you can find the right way to
do it, it, it could be really powerful.
And you don't really capture it here.
The, the biggest area we have found
that has the highest impact is you,
you wanna beef up the content, but
it's the context that matters the most.
And, um, that's where these personas
kinda start to matter, right?
Because a good example is we had a, a
beauty company we were working with,
and our system, um, got very aggressive
about y- almost haranguing them.
It kinda s- kept saying, "High priority,
high priority," and we dug into it
and, and could see why, um, that,
uh, uh, they needed scent profiles.
So apparently, uh, women shop by, um,
if anything has a scent, they have a
set of scents they like and, and, and
are kind of like in their boathouse,
and then some they really don't like.
So for example, these are
things like lavender, citrus.
Some people are very strongly no scent.
They have a bad reaction to scents.
Um, and so on and so forth.
So, um, so it was really interesting,
um, to watch that play out because
there had been a proponent, a couple
proponents in the company that thought
this should be on the PDP, and they
kept getting vetoed, so the, you know,
they, they were very excited by this.
And a lot- most of our customers
take these things and just send it to
ChatGPT, but this one was interesting
'cause there became this internal really
interesting discussion of should we
put this use case on the PDP, which
is this kinda hallowed ground and, you
know, harder to get information on.
But, but what's happened is, um,
because we've streamlined the PDPs
for this keyword world we live in and,
and the fact consumers have such a
short attention span, it's created the
opposite of what the LMS want and need.
So n- none of that use case
stuff is really on there anymore.
Many times you'll have an ingredient
list or, like, some really
kinda specs kinda thing in a tab
somewhere, which the LM likes.
But…
And then you'll have some content,
you know, like this sweater's blue
and made of leather or whatever.
And, uh, but then you're
missing that, the usage of it.
And sometimes it will be captured
in images, so they'll someone…
they'll show someone wearing this outfit.
It's athleisure, so they'll show them at
the gym or playing tennis or something.
But the listing doesn't say that, and
the LM's not processing those images.
They, it, it just, it could, but
they don't use tokens on that.
Um, so, so there's this context piece
missing which is really interesting.
So, you know, we're, we're trying to
help folks figure out how do you, like,
really beef it up and talk about that.
Yep, exactly.
And to your point, y- yeah, it's an,
uh- A- a- and, and people used to
think about this PDP as your point
about, uh, the, the scarcity of it.
Like, you could only put so many data
points, and you gotta just, like, change
that mentality, whereas, like, there
isn't necessarily a scarcity in this
Yeah.
You could send, you know, in
this feed it's actually infinite.
You could send 80 pages
of stuff if you wanted to.
Uh, and why not?
Uh, okay.
Uh, and then here on slide 12,
so you guys have kind of four key
capabilities needed for AI-enabled
transactions, and they are…
I'll start over at the right, uh,
and work our way around clockwise.
So API integration, agent
authentication, payment infrastructure,
and SKU level identifiers.
Uh, and full disclosure, you have ReFi
buy on there, so we, we appreciate that.
You guys, uh, I should have
mentioned this at the top, you guys
are an investor, a small investor,
which we appreciate as well.
Yeah, so where are we?
Give us an update on where we are on
these four buckets, and, and what are
you seeing that's interesting there?
So, a- and I usually start
with the SKU level identifiers.
I think it's kind of s- like, in terms
of my thinking of, like, the, the
full wheel, I usually start there.
I think this is an area that I
think is very much underappreciated.
Um, you know, today there's a bunch of
players that are trying to agentically
check out using- An agent through a
consumer facing e-commerce site, right?
Add to cart, create it, create this
order, et cetera, match it with payments.
And there is a crazy high rate of
failure for a bunch of different reasons.
Most people have thought about bot
detection and fraud, but there's
also just an incredibly high
failure of actually being able to
add the right product to the cart.
And the reason for that is that even
if you know that I want a, you know,
specific Nike shoe in size 12 in blue, the
ability to actually successfully add that
to cart by traversing Nike's website is
not that precise or not that definitive.
And so I think our view is that what the
world really needs to make the checkout
part of the agentic commerce experience
work is some, like, SKU, call it a
global SKU, that identifies what this
product is across Nordstrom and nike.com
and all of these different sites.
The same shoe across multiple sites.
So that will have a specific SKU
that then, you know, ChatGPT can
coalesce all the information about
that shoe to recommend that shoe
versus an Adidas shoe, for instance.
But what's also necessary, which people
aren't talking about, I think, is
mapping that SKU to the individual SKU
identifier that Nordstrom thinks about
in their backend, or Nike thinks about
in their backend, that would actually
systematically 100% of the time allow
you to, to, to add the exact right
SKU in the right color and the right
permutation to cart without that--
some, like, agent making some decisions.
And I think we're missing that.
And I think w- w- part of the value
of sort of the SKU, um, uh, sort of
s- sending, sending these, like, data
feeds out there is in addition to
making, being used for discovery, also
being a, an identifier in that feed
that enables a transaction once that
consumer ultimately decides to buy
that shoe or that, or that perfume.
And I think people aren't talking
about that because they don't
realize that it just doesn't
work for, you know, well enough.
So, like, that's where, like, my
head is on this sort of SKU level
identifies that I think we need
to make the transactions work.
You know, we, we saw this at,
in the ChannelAdvisor days.
These new systems are not
good at canonicalization.
And, and part of canonicalization
is taking all this data coming in,
canonicalizing it, mapping it, matching
it, and then taking the union of the…
It creates a com- If you have a
two-dimensional variant structure,
um, it creates a really complicated
matrix of what's available in, across
those two, you know, some call it
size, color are the two most popular.
Um, and ChatGPT is struggling
with that right now.
You, you can kinda…
They've turned it on
and off, and they're…
We can watch them fiddle with it on
the user ex- interface, and, and we
kinda can infer what's going on there.
Um, so it, it's been really
interesting to watch that.
And then some of the partners, like
Walmart tends to have a lot of the
products turned off that, that have
variants, because I don't think
they, they like the experience.
Again, I'm reading tea leaves on all this.
So that hasn't even gotten to
the part you're talking about.
Like, then once you, once you solve
that, then we've gotta, then you've gotta
get it downstream in the order correct.
And, and so they're still working
on that real time while this thing,
this thing's in the air and they're,
they're changing wings and stuff, right?
Exactly.
And that l- actually fits really well
under the API integration, right?
So we have a strong view that if
you want this agentic commerce
transaction to happen, i.e.
a purchase that you make for a pair of
shoes on ChatGPT, and, um, the merchant
of record is, you know, Nordstrom or Ni-
or Nike, we have a strong point of view
that that transaction needs to be done
with an API and not some agent trying
to traverse a consumer website, right?
For all the reasons I just discussed,
I think it's just too hard to have the
speed and the pres- precision necessary
To do that agentic version versus an API.
The challenge is we don't really have
the infrastructure today to take a
merchant level SKU and just add it to
cart and make that work 100% of the time.
So we're missing this kind of res- like
receptacle, uh, or, or thing that sits
next to, on top of, however you describe
it, the e-commerce system of record
that can take an agentic transaction.
And so I think, again, Shopify works
pretty well 'cause every Shopify
cart is the same, but for everybody
else, even, you know, various
different versions of Commerce Cloud
checkouts are gonna be different.
I think you need a new
piece of technology.
And as we think about the ecosystem
of startups that are going after
this, I would say, like, there's
sort of two fundamental different
approaches people have taken.
One is, like, called the API or bust.
Like, we're just gonna build the API,
and we're gonna go sell this thing 'cause
we're clear that's the end solution.
And then there's, like, the fake it till
you make it, which is a bunch of players
that are saying, "Look, yes, we either
believe or don't believe the API is the
right solution, but to get there and to
have a broad catalog of merchants, we've
decided to use this agentic flow," like
just, like, hack it and try to use agents
to flow through a consumer website to have
a broad set of merchants because you don't
effectively need the merchant to set,
sign on to saying yes to that approach.
So I think the jury's out on what the
right app- if even if you believe that
the API is the end solution, should you
start with a fake till you make it, make
a bunch of merchants available on your
platform through this thing that works
70% of the time, or do you just need
to go door to door and sell it as API?
I think it's to be determined, but I think
it's a really interesting category that
needs to be solved for the transaction
piece of this ecosystem to work.
Are you doing this for fraud or…?
There's a couple different things.
So like, so today- It, agent
authentication is important for
what I just described, right?
You have to, like, identify that
there's an agent that's trying
to use your consumer-facing site
to purchase a transaction, and
somehow, like, allow that to happen.
So, like, even in the fake
to make approach, you…
Like, there needs to be some
technology, whether it's working
with the, the fraud prevention,
you know, something like Forter, et
cetera, um, to enable that to happen.
So there, like, we'll
have to figure that out.
Um, there's also agent identification
for the scanning of your catalog.
Who should you let scan
your catalog or not?
Um, but I think even in the long term,
even if you have an API to take the
order, you've got to determine is this a
legitimate agent and a legitimate consumer
to be able to take the order from?
So, and which relates to charge,
like, chargeback and who owns it.
So I think there's a bunch
of necessary tech there.
We're probably spending a little
bit less time here in that, like,
some of these fundamental building
blocks need to be built first before
this becomes super important, but it
will become important at some point.
And you're not including in
here, so I've talked to, um…
We had Human, uh, on the pod and
then, uh, I know the For- Forter guys.
If, yeah, Forter, so we're actually,
we're, we're invested in Forter, so yeah.
Yes, we, we, we've talked to
them a lot about it, yeah.
They're more, like,
infrastructure layer though.
Um, and then this is more, like,
maybe baked into the agent, some, some
encryption agent to agent encryption,
which is, like, not in their wheelhouse.
Or are you including that
in the whole stack of this-
We, we, we actually had a little
bit of a debate about, you know,
should we put Forter in this, in
this, in this category or not.
Um, uh, because I think what, what
Forter does is it, like, identifies
the real human making a purchase and
that they are, like, it's not fraud.
And this breaks some of that.
So, you know, they are partnering
with and working with some other
folks to go figure out how you change
that, um, change that relationship.
Um, but you'll need to because,
like, ultimately most of the fraud,
fraud players were designed to
block these sorts of transactions.
So we just need this full-scale change
or, like, opening up whitelisting,
et cetera, to make this possible.
And then the last is like, you know,
payment, the payment infrastructure.
Um I think our view is that weirdly,
a lot of the payment infrastructure
tech is already out there.
We already have tokens.
Like, the fintech side of things
is farther along than I think the
commerce infrastructure pieces are.
We have, we have- we're trying to figure
out how we assemble them, but we have
the fundamental building blocks there.
You know, I think the payment processing
isn't gonna change fundamentally.
Do I think we're gonna use
stable coins or something?
I'm not totally sure we will.
I think we'll, it'll be, it'll look
a little bit more like the current
payment processing ecosystem.
We will have to figure
out a couple things.
One is, like, the chargeback
and liability, who's on
the, who's on the hook?
It's not really a technical
problem, it's more of a business
problem, but who's gonna…
W- if there is fraud, um, or
issues with the transaction,
who's gonna be on the hook for it?
Um, and then I think there's also
this, like, battle over the wallet.
Who is gonna own the consumer's
wallet for these agentic transactions?
Obviously, ChatGPT, Perplexity, they
all wanna create their own wallets.
They're us- using folks like,
like, like, they, they're using big
players to b- to build these wallets.
I think all of the merchants probably
don't want these transactions
to be used by ChatGPT's wallet.
They'd rather use it, their own.
So I think we'll have this battle
about who gets to own the wallet,
and ultimately who owns the customer.
But the technology is
actually pretty far along.
Companies like BGS are
doing some of this stuff.
But it's, and it's more of a
business decision question than it is
brand-new fundamental infrastructure.
Any other slides you wanted
to hit in the presentation?
I think those are
the two.
No.
No, those, those are the two that,
that I think worth chatting here.
We got a bunch more, but happy to chat
with other folks offline, but yeah, um.
You, you have all these
meetings with payment companies
and retailers and brands.
W- what's top of mind with them?
I mean, obviously these, I think
this presentation's a reflection
of some of that, but anything above
and beyond that, that we haven't
talked about?
No, I think, I think I k- I mean,
talked about it a little bit earlier,
which is, like, there's the three
things I talked about, right?
Which is bring AI goodness to your site.
So, like, we strongly think people need
to do that, and there's some really cool
companies that are doing, like, building
that, bringing some of that, like, AI
conversational stuff to their sites.
The second is just getting a feed
up and running to give you the
levers you need to be able to send
this data out to the AI platforms.
Almost regardless of how much you, you
wanna send a lot or a little, you gotta
have these levers, so that's probably,
those two are, like, top of mind.
Obviously we're a little bit less, like,
I think the retailers probably have to
think less about this transactional thing.
I think they're gonna be
recipients of how this plays out.
It's so early that I think focusing on
that today from a retail perspective
is probably a little early, but just
being aware of what's out there.
So it's probably the first two ca-
two of those three categories that I
mentioned earlier are where we've told,
we've, we've recommended that our sort of
retailer friends really focus their time.
What do you think about the whole
Google versus OpenAI battle?
You're kind of, um…
I went on a trip, uh, actually
I saw you when I was there.
Uh, it was way, it was a while ago.
It was like, what was that, August?
And it- Every visible surface
had an AI promotion on it.
It was like, yeah, it made me…
I, I didn't think we were in a bubble
until I went to, out to Silicon Valley.
I'm like, "We're in a bubble."
Um, and there was even, like,
billboards for companies
saying, "Are you a YC company?
We have this infrastructure
layer for you."
And it was like, whoa, we're,
you know, we're, we're like, the
bubbles are nested six layers down.
Um, and, um, you know, y-
maybe let's start there.
Do you think we're in a bubble?
I mean, I think we always go through
the, you know, the cycle, right, where
everybody gets very excited about it.
Um, and i- there's bust boom,
and I think we will have some
of that, which is a lot of stuff
that we hope would work out won't.
Um, but I think this is
fundamentally real, right?
I think you look at the, the
consumer activity is there.
I think the promise of what this agentic
commerce stuff could look like is there.
I think I'm probably more on
the, the bullish side of things.
I mean, a lot of people are saying,
"Hey, this is great, but it'll work
for, you know, reordering your toilet
paper, but not for buying a sweater."
I'm actually more optimistic than that.
I think people…
I think we will move really quickly
into this agentic commerce experience
feeling much more emotional, and
less just pure chat-based, which
I think opens up the market.
Um, but I think a lot of this, like,
there's so many players going after
this market, the opportunity is so
big, it will burst a little bit.
But I think it's, I think it's real.
I think it's a long-term trend
that is worth investing against.
Yeah, there's no GPU sitting
idle right now, unlike Dark
Fiber that was obviously dark.
It's in the name.
That's always Jensen's, uh, answer,
which I think is a pretty good one.
It's definitely, if it's a bubble,
it's not like the internet bubble.
It's definitely different.
It's a, it's a…
It would be a different bubble.
It could be.
Yeah.
But it's definitely
not the same dynamic as
the- I think the-
… nine, 2000
… I guess one definition of a bubble
they always talk about on CNBC
is you don't know you're in it.
So, s- so you almost have to say it's not
a bubble, 'cause, well, I guess if you say
it is a bubble, then you're not in one.
I don't know.
I don't understand the…
There's some philo- I wa-
not a philosophy major.
The logic breaks my head on that one.
Um, what surprised you the most
this year since we've chatted?
Like, what, what's something you're
like, "Holy cow, didn't see that coming"?
Honestly, it was what
I talked about earlier.
It's, it's honestly the, the, the, it's
the, it's not the thing that happened,
but it's the reaction to the thing.
Like, how people have continued to see
these new announcements and think that
they are fundamentally moving us forward.
Like, e- every, every announcement is
huge, and everybody gets, "Oh, like,
why we should bother focusing on this
thing and going to invest in this company
because so-and-so has solved the problem?"
And just, like, it keeps
happening, and I just surprising
to me that that keeps happening.
It's probably my biggest, um- The thing
that has surprised me the most, honestly
And you think that's because they don't
understand the back-end part of our world,
that there's all this complexity that,
that you've surfaced on that one slide?
I think that's, I think that is the case.
I think that there, um, if you haven't
lived in e-commerce or built an
e-commerce site, particularly from
scratch, I think you just assume
some of the stuff works a lot better
than it actually does, or that it's a
lot simpler than it actually do- is.
Um, and it's like part of why I
think people are like, "Oh, is
ACP gonna solve the problem?"
Like, no, it won't solve the problem.
It's a part of the solution, or
could be part of the solution.
It's the standard, but it doesn't solve
the plumbing issue, like which is the hard
stuff about making all this, like making
ACP actually work on the merchant side.
So I think it's just people who are
like, if you don't know the complexity
of how merchants work, I think
you underestimate or overestimate
how these new announcements can
solve the fundamental problem.
Let's say you, uh, you decided that
this VC gig was getting old, all the fun
meetings and free lunches and everything.
Whatever, whatever benefits you enjoy.
The, uh, um, and you decide to
go operate, uh, a jewelry store.
Like you get back in the biz.
What, what would you be doing?
Like as an operator, like put
on your operator hat, like what-
Yeah.
I, I mean, I would be getting very
aggressive on feed management, right?
Just like, how do I test every single
possible thing that I know about
my products, questions that I could
have the question to and the answer,
and just, like, flood the internet.
Like, I was…
My jewelry business was the benefit,
honestly was almost, like, exclusive-
not exclusively, was largely
driven off of SEO, 'cause we own
the long tail of jewelry search.
Like, we just owned it because we
created every single permutation
and combination, because we had
this crazy e-commerce infrastructure
that allowed us to, we built it
ourselves, that allowed us to do that.
And I think that served us for many,
many years to grow, to build this
online footprint in the old SEO world.
And I think there's this opportunity
today to build that same online
footprint in the GEO world, specific-
specifically to individual products.
So like, this is not the kind of GEO, I'm
just gonna make my SAS company show up
for a bunch of queries that are relevant.
It's like, how do you make a catalog
at scale, each of the individual
products in the catalog show up as
relevant and interesting to ChatGPT?
I think that's where, like,
the biggest opportunity is.
You know, I was talking, we
end up talking to a lot of CPG
companies, even though it's, like,
not really our, our target zone.
And, you know, my advice is flood the
zone and, like, just what you just said.
And they're like, "Yeah, but we're, you
know, Walmart and Amazon will do that for
us, or our retail partners and Target."
Let's just say the big three for
that, and maybe Kroger or something.
And I was like, "Amazon and
Kroger are blocking all the bots.
Uh, Target is just doing this
app thing, and they're…
You don't want one of those…
They may be great retail partners and
everything, but they're not going to take
the effort for every product they have and
add all these important elements, and they
don't know your product as well as you do.
And you, you need to control your own
product, you know, description destiny
by, by having, you know, every site,
every one of these CPGs has a website,
and it may not have transactions,
but it's being scraped by ChatGPT.
Use that.
You've, you've got this almost
ready fire hose to put in there,
and you can even send a feed.
Like, ChatGPT doesn't have to
have transactional data with this.
They just want product data."
A- a- and I think, and, and flood,
I mean, honestly, go back to my
jewelry days, we did that too.
We were, at the beginning,
we were a very small brand.
But because no one had content or
products across this long tail of
thing, like, we were the answer.
We were the only one left.
We were the only one there.
So I think the same
thing is available here.
If, like, if you can put out content
that is answering the question that some
user is asking on ChatGPT, if no one else
has the answer, they're gonna probably
use yours, regardless of who you are.
So…
And if you're a incumbent, a
challenger's gonna do it Because I bet
you displaced Cajewelry or something
Exactly.
There's no, there's no fundamental
advantage you have as a, as the,
as the incumbent here yet, right?
Like, 'cause if the, if the white-
if it's all white space, they're
gonna take it from anybody, and
they don't care whether it's P&G or
some brand no one's ever heard of.
Like, just you need to find that
real estate or occupy that real
estate regardless of who you are.
All right.
Let- that's a good recap
of all the 2025 things.
Let's look forward, and I
know it's hard to predict.
We, we, you know, just the
last six months were so crazy.
Let's look at next year.
Do you have any predictions?
What's Matt's crystal ball telling him?
I think we will start to s- if you ask
a retailer today how much traffic or,
like, influence they're getting from
ChatGPT or, you know, AI in general,
I think the answer that I have heard
is, it's small, kind of more than
I thought, but like still small.
I think we're gonna see
that change dramatically.
I think whether it's actual traffic
that is delivered by ChatGPT, or
influenced through, or purchases that are
happening, like I think we're just gonna
see that number spike way faster than
even those CMOs think it will, right?
Regardless of whether it turns into
actual like agentic transactions,
that'll take some time.
But I think the pace at which this happens
is going to surprise people, honestly.
Yeah.
And, and there's a, there's a
pull, which is like here's this big
audience set using these things.
But then the push is con- as,
as Google reacts to all this,
um, they're sending less clicks
downstream, so everyone's SEO traffic.
Like, if you were running the
jewelry business, this would,
you would be running around with
your hair on fire because yeah.
People that have lived on SEO are
dying on SEO right now, and it,
it is very scary to watch that go
down and to the right very quickly.
And I think part of it is like bec- it,
it's even, it's exacerbated because Google
has shown their willingness to cannibalize
their core search business, right?
If they were, if they were holding their
ground and be like, "Oh, this whole
AI thing p- probably won't play out,"
it would actually slow the thing down.
But the fact that they're leaning
into it smartly, I think, like it…
They're just gonna accelerate the
thing, 'cause they're not forcing
users to change even the destination
they go to to switch from kind of
blue link searching to AI searching.
It's just happening.
That's a good prompt, because there's
four companies that are sitting it
out, and the biggest notice, note-
notable one is Apple not doing anything.
And then you've got Groq, Anthropic.
Anthropic's kind of really focused on
coding and, and really into that vertical
and, and maybe even B2B enterprise agents.
Um, and then we have Meta, who's
like lost in the jungle, kind
of, you know, spending their Zuck
Bucks on, I haven't been fortunate
enough to have any rain on me, so.
Uh, raining Zuck Bucks on
people that can build models.
Do, do you think, um, in '26 we see
one of those guys kind of get into
the agentic commerce ring, or you
think they're gonna do other stuff?
I, I don't know if '26 is the right year.
It could be.
I mean, I think the obvious one is Apple.
I mean, one because they just
have, I mean, they've got Siri,
they've got the endpoints.
Like, and honestly, in almost
every category they've been super
happy just waiting it out for
someone to try some newfangled
thing and then they do it right.
So I, I would be surprised if they
don't figure this thing out, honestly,
and build a really compelling sort of-
A, a, a con-, like a much more elegant
consumer-friendly version of ChatGPT,
whether it's using Siri or using
something else, that gets a whole new
set of customers in the door as it…
Because it integrated into their phones.
So I think they're the probably the
ones that have the biggest opportunity
to do something, and I think will.
I just don't know how
long they're gonna wait.
Since we talked, I think when we
originally talked, there was, like, thr-
three or four vertical agents, which are,
are ones usually focused on one category.
Um, the most famous is Daydream,
and then now there's Gensmo,
I think I'm saying that right.
Um, and a couple, um, I think
you probably know, is it Arcade?
What's…
Is that the jewelry one where
you can do custom jewelry?
That's probably near
and dear to your heart.
I don't know if you've
played with that one or not.
Yeah.
What, what do you think
about that category?
Is that, um…
You guys don't really invest in
consumer-facing stuff, right?
So- We
don't, we, we don't
invest in consumer ones.
Like, I would say in this case,
I've had some pretty strong
opinions over this conversation.
In this one, I have a modestly
high conviction, m- modestly high,
high conviction, uh, opinion.
I, I am more optimistic on the horizontal
models winning than the vertical models.
Um, reason being a couple things.
One, I think just the vol- the, the
amount of dollars they are spending
to get better at this is just, like,
dwarfs all of the vertical guys.
The second is I think they have been
successful at getting you and me and all
of us to put all of our queries through
one platform, and the ability to mine
that data to make vertically specific
queries really relevant, I think will be
high, will be, will increase over time.
So they have this, like, data advantage.
And the third thing is, like, if
you think back about why vertical,
one of the reasons why vertical
search and e-commerce existed was it
required specific inputs to get at
the thing that you were looking for.
Travel is a great example, right?
The interface for travel sites is
like, when's the date you're leaving?
When's the date you're, you know, arri-
you know, arriving and le- leaving?
Um, um, how many people in your party?
All these different things
which requires sort of a very
industry specific set of inputs.
In a world where the inputs are more
textual based, you kinda don't need that.
Maybe you want it, but
you don't really need it.
So I think there's this opportunity
to create, to not be forced down
the, the, uh, you know, a path
of creating a very vertically
specific set of inputs or results.
Um, so I think in the long term,
I'm bullish on the horizontal guys.
It's possible in the short term that the,
that the vertical guys have a better,
you know, more exciting user interface
And I think they, they are today.
They're, you know, like, but I think
ultimately, you think about shopping
for sweaters on ChatGPT in three years,
I think it's gonna feel a lot less like
a chatbot and text in, text out, and a
lot more like talking to a Shopify site.
Right?
Or multi-merchant Shopify site is
probably the way to think about it.
So I think it's just gonna
get better in ways that people
aren't totally thinking about.
All right, let's cast your magic
crystal ball out a little further.
Um, so PayPal CEO has said that
by 2030, 25% of e-commerce will
be agentic commerce, and, um,
I've seen, uh, Morgan Stanley said
20% kind of on their bull case.
Um, and then another way to look at
it is, um, the, there's a set of, uh,
Wall Street people, mostly Bernstein
and then Morgan Stanley also opined
on this, where e-commerce is growing
five to 7%, depending on which numbers
and which category you're looking
at, but, but, but basically a point
or two above retail, uh, overall.
Um, and they're basically, their bull case
is that by 2030 that really accelerates
maybe, like, three to, you know, three
to 4% or, which would be, what's that,
thousands of basis points I think it is?
Yep, so three to 4,000 basis points.
Would any of those predictions
feel right, wrong, low to you?
I would say on the PayPal one, so
the 25% of transactions being agentic
commerce, um, I think if you were talking
about agentically influenced commerce,
so inclusive of people finding some
sweater or some car part on ChatGPT and
then navigating to the car part seller
or, you know, the e-commerce merchant
and buying in the traditional way,
then that, that number feels right.
It's still up, still, it's a huge
number, but I think that could be right.
If you're really talking about purchases
made on ChatGPT where the consumer never
made it to the end merchant, seems high.
As, as bullish as I am on that, like,
that would just be incredibly high.
Um, so, but it- it's still…
Even if you get to 10%, it'd
still be huge, so I don't
think you need to be that big.
But, so I think, yeah, influenced,
probably be about right.
And then the, the, the Morgan Stanley
number, I think it's optimistic.
I, I think- If you assume that it's
basically you're assuming either overall
retail is gonna grow faster, right?
Which seems hard to believe that
there's that much unfulfilled demand
that, like, just literally changing
the way that people buy will, will
influence the growth rate that much.
Or it's a shift from offline retail
to online retail for the same reason.
Either way, I think it's … I, I
think the fundamental shift is not
gonna be between agent of commerce
driving more transactions overall
or agent of commerce pulling
transactions away from physical retail.
I think it's agent of commerce
pulling away share from other
ways that you buy online.
So I, I don't think it, I don't
think it fundamentally changes the
overall growth rate of e-commerce.
I think what it does is that more
and more of e-commerce will be
agentic, if that makes sense.
Which is still a huge shift, I just don't
think it changes the, the, the growth
rate of e-commerce volume all that much.
All right.
We have you on the record.
We'll check back next year
and, uh, see how you did.
Perfect.
Yeah.
Let's do it.
Uh, appreciate the time.
Uh, I wanna make sure you have
time to get to your next meeting,
and really appreciate you coming
back on to share your thoughts.
It's always, uh, you sit at this epicenter
of things, and it's always great to
hear what you're seeing out there.
Yeah.
Uh, love to be on, love the
conversation, great questions,
and always fun to riff on stuff.
So thanks, Scott.
All right.
Have a great holiday, and
we'll talk to you next year.
Follow, rate, and share to stay up
to date with Retailgenic, your guide
to the future of AI shopping agents.