Retailgentic | Consumer Behavior & Retail Trend

What happens when more than half of online shopping traffic comes from AI agents, not humans?

In this episode of the Retailgentic Podcast, Scot Wingo sits down with Alex Rampell, General Partner at Andreessen Horowitz, to unpack the seismic shift happening at the intersection of AI, commerce, and payments.

We cover:
  • Is Google really screwed? (and why AI threatens its business model)
  • The future of agentic commerce & attribution wars
  • Lessons from TrialPay, Affirm, and affiliate marketing
  • How platforms like ChatGPT could become checkout destinations
  • Why “optimization” may be the killer feature of AI shopping agents
If you’re in e-commerce, retail, fintech, or just want to know how AI will change the way we buy and sell, this is a must-listen.

⏱️ Time Stamps
04:45 – Alex’s early days: shareware & credit card processing
06:26 – TrialPay, affiliates & the Netflix-for-shareware idea
08:45 – Founding Affirm & moving into venture at a16z
09:55 – The “apps practice” and where value accrues (infra vs. apps)
10:57 – “Is Google Screwed?” background & themes
13:02 – Commerce attribution & affiliate marketing evolution
15:04 – Impulse vs. highly considered purchases in AI era
18:27 – Optimization: why AI wins on time vs. money tradeoff
21:00 – Reviews, shill content, and trust problems
23:50 – Hosted checkout: could ChatGPT become a marketplace?
26:14 – Will Apple jump into AI commerce?
28:35 – Coupon codes, group buying, and negotiation by agents
30:04 – How commerce could be “rewired” by AI platforms
31:25 – Closing thoughts & what’s next

🧠 Want to stay ahead in AI commerce? Subscribe and follow along:
📰 Subscribe to the free Substack: retailgentic.ai
📺 Watch episodes on YouTube: youtube.com/@retailgentic
🎧 Listen wherever you get your podcasts: Spotify, Apple, etc.

Creators and Guests

Host
Scot Wingo
CEO and Co-founder, ReFiBuy.ai
AR
Guest
Alex Rampell

What is Retailgentic | Consumer Behavior & Retail Trend?

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).

One of the functions of AI is I
know exactly what I wanna buy.

Like, this is the item that I want,
and maybe I got that from AI, right?

So for, like, more of a lifestyle
purchase, it's like, "I want this

thing that's quite expensive.

Please tell me what to buy."

I spend four hours researching it, talking
to my friends, reading Reddit, watching

YouTube videos, and yes, of course,
conversing with ChatGPT or something

else, and then I decide what I want.

And once that decision has been made,
then it kinda goes into this, you know,

kind of one, one step up of impulse.

And so it's not an impulse buy,
but it's like, "I want that SKU.

This is the exact item that I want.

Find it for me, you know, cheapest
shipping, lowest cost, best warrant,

w- whatever it is that I want.

In fact, even…

Here are my five different credit cards.

I don't know who's gonna
give me the best cashback.

I don't have time to figure that out.

I value my time more than my money.

Boom, go."

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 the Retailgentic Podcast.

As we head into this long
holiday Labor Day weekend, we

have a special treat for you.

This is a perfect podcast to
listen to, uh, on a drive or

whatever you have coming up.

Maybe you'll be on a flight.

Um, I do…

Uh, I am a bit of a slower talker,
but our guest is a fast talker, so

I think you want this one on 1X.

You, you…

I wouldn't flirt with 2X on this one.

Uh, anyway, on today's pod, I
am excited to host Alex Rampell.

He is the general partner at
Andreessen Horowitz, where he heads

up their billion-dollar fund that
focuses on the application layer.

The reason I asked Alex on the pod
is to talk to us about Andreessen

Horowitz's thinking behind the
recent piece, "Is Google Screwed?

Maybe, But Not How You Think."

Uh, we featured this on last week's,
uh, Agentic Commerce Roundup in the

Substack, and Alex was kind enough to
come on and give us a little bit more

color about what inspired the piece,
some questions I had about it, and

where he sees agentic commerce going.

Alex has the bona fides
to talk about this.

He has very deep experience in e-commerce.

I first ran into him
when he was at TrialPay.

Um, he also started
this firm called Affirm.

And he has many other experiences
in the world of e-commerce,

digital payments, et cetera.

He walks us through, um, the background,
uh, of, you know, how he got to Andreessen

Horowitz, uh, and then we jump into
a wide-ranging discussion of where

he thinks agentic commerce is today,
uh, why he thinks a agentic camel,

camel, camel application is the next
logical killer app, and finally, where

does he see agentic commerce going?

I hope you enjoy this conversation,
uh, and have a great long weekend.

Alex, welcome to the Retailgentic Podcast.

It's great to be here.

Thanks for having me, Scott.

Yeah.

I really appreciate you taking time.

You're super busy.

Um, let's do an abbreviated
version of your background.

Um, you and I were reminiscing in
the green room, and we've known

each other for at least 13 years.

So, um, you've been in e-commerce for
a while, and like many of the folks

at a16z, they really like partners
that have an operational background.

So tell us how you, you cut your,
your chops on, uh, e-commerce.

Yeah.

So I, uh, this is a term that will date
me, but you will probably know this term.

I used to write shareware, try
before you buy software, in the '90s.

Mm-hmm.

And that was all checks in the mail.

So you would upload a pro- like, I had
all these different products, and I'd

upload them to different bulletin boards
around the country, as well as the

AOL File Library, the CompuServe File
Library, and then if you were lucky,

PC Magazine would include a floppy
disk every now and then, and they'd

include, like, a bundle of shareware.

And all of that was, "If you like
the product, please send a check."

And then if I got the check, I
would send you a code, and then

the code would unlock the product.

And they, they were kind of all…

You know, you'd just have these
annoying reminders until you paid.

And sometimes it was just,
like, straight up honor system.

This was kind of honor system
and, uh, reminders or annoyances.

So that's actually how I got into credit
card processing, in that I figured out how

to process credit cards in the mid-'90s
on the internet, which, um, seems…

There, there was no Stripe back
then, so it, it was a little

challenging, but, uh, got into that.

Um, and then, uh, started a, a
venture-backed company called SiteAdvisor,

um, which was a web security product.

And for both that company and
for, uh, all of my shareware,

people didn't like to pay.

Um, they, they, they like getting…

They'll…

You know, you're, you're happy to pay
for a glass of wine, um, or for a,

you know, a golf ball or something
tangible and physical, but most people

don't like paying for digital goods
like my shareware or like this, you

know, $20 security software product.

So I had this idea of I'll, I'll
give it to you for free if you try

something else And that's how I really
learned about the affiliate mar- model.

Um, I had known about affiliate marketing
for a while, but, um, but basically,

you know, go sign up for Netflix and
we'll give you this product for free.

Go shop at The Gap, we'll give
you this product for free.

That was all powered by affiliates.

Um, or we became one of the biggest
affiliates in the world, and I, I started

a company called TrialPay that basically
commercialized this concept, because I

had done it for my shareware, I had done
it at this other company, SiteAdvisor.

Like why everybody should want
this, and our first client became

this company called WinZip.

Everybody would download WinZip.

Nobody wanted to pay for WinZip.

Soon they were making a million
dollars a year from this little thing.

It's like, "Ah, it's a good idea."

So I, I raised venture capital for that.

Um, ran that company
for about eight years.

Uh, sold that to Visa.

Um, and, uh, Visa had good plans for it,
but, uh, they, they have too good of a

monopoly business with for their course.

It didn't…

And unfortunately it
didn't really go that far.

Um, and then also started another
company, um, called Affirm,

the buy now, pay later company.

So, um, Max Levchin, who's the
CEO of Affirm, um, he had…

I, I was trying to get him
as a client for TrialPay.

Failed miserably, but, uh, he was the
one social gaming company that kind of

didn't use, uh, didn't use TrialPay.

That was kind of our, our big, our big
sweet spot was people that don't want

to pay for social games, guess what?

They'll sign up for Netflix,
they'll s- they'll switch to

Geico, they'll shop at The Gap.

That's what we did.

Um, so then he and I cooked up this idea
that became Affirm, and, um, and then

I joined Andreessen Horowitz about 10
years ago in 2015, um, because I brought

Affirm and, uh, two other companies that
I kind of helped start to Andreessen

Horowitz, and it's like, all right, um,
um, might, might as well kind of, uh,

take my hookie job, take my, my, like
part-time job and make it a full-time

job by, um, investing in companies.

So have been here for about 10 years.

The side hustle became the full hustle.

Exactly.

Yeah, and, uh, I was poking around on the
a16z site, and it says you are leading

the firm's billion-dollar apps practice.

What is an apps practice?

Is this like, um, iPhone games?

No.

So I mean, it's, it's basically, if you
look at the history of, of technology,

uh, at least since the PC came out,
there's always, like, an infrastructure

layer and an application layer.

So if you look back early, like,
what was the infrastructure layer?

It was all the, uh, programming,
like all the compiler companies

like Borland or MS-DOS.

That was an infrastructure company, and
Microsoft was an infrastructure company.

The application companies were the
applications written on top of that.

When the internet came out, like, the
canonical application layer companies

were, like, Amazon and eBay and
Yahoo, the, these companies that got

very, very big in the, the mid-'90s.

The infrastructure layer was,
like, Cisco and Akamai, like down

to the pickax and shovel plays.

So the way that we've divided up
our venture practice is we have

the, um, the infrastructure layer.

So, you know, think what is OpenAI?

That's an infrastructure company,
and they actually do work

directly with the consumers.

But then there are lots of application
layer companies that are built on top.

And it's kind of this age-old question
of, like, where will value accrue?

Does it go to the infrastructure layer?

Like, everybody wants to be pickaxes and,
and shovels, but if there are too many

people selling pickaxes and shovels, you
actually want to be the gold miner, um,

and kind of find your, your, your gold.

So the way that we're divided up is
we have applications, which is kind

of think of it as anything B2B or B2C.

Um, if it's a B2B or B2C, like,
thing, then that's application.

If it's selling to developers,
that's a very, very different motion.

Or if you're selling to people
with, like, you know, super P…

If, if you have a mega PhD and you're
selling to other people that have mega

PhDs, it's probably less about the did you
figure out your go-to-market prowess or

d- do you have a great go-to-market plan?

Um, and this is kind of the, the, the
saying that I've used for a long time,

which is the battle between every
startup and incumbent is whether the

startup gets the distribution before
the incumbent gets the innovation.

Um, but if you have, if you've,
if you've figured out how to split

the atom- kind of doesn't matter
about your distribution strategy.

Like, you, you, like, you can,
you can just kind of commercialize

your super geniusness.

Whereas at the application
layer, it normally, that, that's

almost like a secondary thing.

It's like you have to find some real
problem, but you really have to figure out

how to sell it, and it's that latter point
that's actually more, more, um, important.

Got it.

Um, okay.

Very cool.

The, uh…

So one of the reasons I reached out
to you is you wrote this really good

article, you and, and a colleague,
and it was called, "Is Google Screwed?

Maybe, But Not How You Think."

So, so maybe like, uh, before we
get into that, like, what led you…

Is this a space you guys…

You know, obviously here we're,
we're somewhat obsessed with a

very, you know, myopic focus on this
intersection of e-commerce and agentic,

uh, you know, so, you know, agentic
commerce, for lack of a better phrase.

Um, and you know, we- we've, we've
got some theories on that, that we've

been talking about on the Substack.

But curious, like, do you guys
have a theme here or, or, or

even you personally that…

Or what, what got you interested
in this general space?

Yeah.

I mean, I think in general, um,
I mean, Google is the original…

I talk about this in the piece
obviously, but Google is the original

freemium business model where they
don't make money most of the time, they

make a lot of money some of the time.

Um, and I think there are kind of three
intersecting trends that kind of actually

come back from my, my early career.

That's why I, I thought it was
relevant to mention affiliates.

Um, how do you get paid?

Actually, there, there, there's like
this long-running battle that I'm,

you're very, very familiar with is how
do you attribute credit for a sale?

So somebody buys a car, um, you know,
BMW advertised or bought, uh, you know,

paid for product placement in the James
Bond movie that you saw 25 years ago,

and that's why you wanted the BMW.

And, like, part of the attribution for
the credit for the sale goes back then.

Part of it is also the, the,
the dealer and the, the la- or,

like, the actual sales guy at the
dealer who, like, convinced you

to buy this car and not that car.

Part of it's that Google
ad that you clicked on.

Like, there's, there's many, many,
many things that kind of, um, make

a synapse fire in your brain in a
certain way to go buy something.

Um, and what's interesting about Google
is, I mean, I, I was really, really, um…

I found it fascinating when, uh, Google
stock had, like, a, I think it was, like,

a 5% to 7% drop in one day on the heels
of Eddie Cue's testimony that actually

they started seeing search volume drop.

But then fast-forward a couple months,
like, and, and he wasn't lying about that.

It's like it dropped.

Um, but fast-forward a couple months,
Google's revenues and profits were

stronger than ever, and it's not like,
you know, it, it, the one part of revenue

was up and the other part was down.

The search profits are up.

So how, how do you square that?

And it kind of makes sense, which
is, I mean, w- what we lay out

in the piece, which is, you know,
the first thing to fall is these

non-commercial intent, um, uh, searches.

And then meanwhile, like, OpenAI and
others are making a big deal about how

they are going to hopefully bring agentic
AI to commerce, and then that actually

goes back to affiliate marketing, right?

It's like, well, if you're
going to recommend…

Like, you can't just say, um, you're,
you're not going to have, like,

little tracking pixels and cookies,
which is how affiliate commerce

has been adjudicated for 30 years.

Like, how is OpenAI going to get paid?

And then let's design an ontology,
if you will, for different,

for commerce writ large.

And what I mean by that- Mm-hmm … is,
like, buying something is not buying

something, and that's where we try to lay
out there's the highly, um, impulse buy,

and that's the, like, I'm in line at the
supermarket and I end up buying a Snickers

bar even though I know I shouldn't.

And they place those things strategically.

These are called end caps in retail
terms, but, you know, even just the

checkout line itself is filled with
things that you should never buy.

You weren't even planning on buying it.

They're usually at a kid level
too, so if you go with your

kids- Yes, yes … they can grab

the-

Exactly.

The, the, the sugar is down
there- … and, like, the, the, the

magazines are at the top or something.

It's like they know their
audience very, very well.

And then on the other end of that,
you have things, uh, that I call…

Not I call, but I think
are commonly referred to as

highly considered purchases.

So nobody's ever going to buy a, buy a $10
million house on a Google AdWords like,

ooh, click on this ad and then buy house.

These highly considered purchases, um,
are probably, you know, AI will, will

take a, a role potentially on those two
extremes of that, but I think de minimis.

Because almost tautologically, if it's
something that you aren't looking for,

well, you're not gonna go to ChatGPT
to buy something you aren't looking

for because you're not looking for
it, so you're not going to go there.

Um, and then if it's highly, highly,
highly considered, you might have

a, you might have an input in
terms of like, where should I live?

Let me ask ChatGPT or let me
ask one of these AI agents.

But when it comes to actually, you know,
adjudicating the, uh, the, the credit

for the sale, as I mentioned, it's
probably not going to be like I'm going

to click and buy this right within, um,
an AI chat or use an AI agent for that.

It's gonna be part of the equation, but
certainly will not be the last click.

So that kind of leaves three
different forms of commerce that I

would call roughly in the middle.

And, um, I think I mentioned this
in the piece as well, like the, the,

the site that My job as a venture
capitalist is actually in more respects

to kind of, um, articulate the present
more so than predict the future.

It's like, "Oh, here, here's what's
actually happening right now," and, um,

the entrepreneurs are predicting the
future and hopefully building the future.

Let's bet on something that already…

that consumers already want.

And there are some of these things that
consumers have shown an appreciation

for, and this is why I talk about
CamelCamelCamel as my favorite site in

the universe, because CamelCamelCamel
is like the poor man's AI, pre-AI.

It's like, "Hey, I will buy
things when they are cheap.

Please let me know when they are cheap."

Okay.

W- that, that's what CamelCamelCamel does.

It's like, "I want, you know, Coke Zero,
but I don't want to pay 70 cents a can.

When it goes down to 50 cents a can
on Amazon, send me an email alert."

I'm gonna click on that email alert.

What am I going to do with it?

I'm going to buy.

So you have this process where,
like, you the human are the agent.

Like, clearly that is something
that if I just say, "Buy it for me,"

like people already do steps one
through nine in that process, they

will do step 10 of purchase as well.

So I think it's trying to figure out,
like take commerce, which is enormous.

I think 16% of commerce is
e-commerce, so still regular com…

offline commerce much bigger
than, than e-commerce.

But what are the different parts that
potentially are addressable by AI today?

Which are the part…

What, what's missing?

I think this is a really, really big part.

And the thing to me that's missing the
most is actually, I don't think this

is, "Oh, we need AI to get smarter."

The, the data is actually quite weak
because if I want to know what product

XYZ to buy, well, the problem for the
last 25 years of the web is it's been

polluted with these affiliate websites,
and I was guilty of this as well because

I did a lot of stuff in affiliate.

But the, the search engine optimization
and all the content that's been developed

around, uh, commerce and around affiliate
marketing is not necessarily what I

would call Consumer Reports in the 1990s.

And if you know Consumer
Reports, this was…

this is a famous magazine
that would not…

It's a nonprofit, doesn't take
advertising, just wants to do…

It's like the Ralph Nader
of consumer reviews.

Only wants to show the best
product, say, "Don't buy this.

Do buy this," and you can
trust the recommendations.

Can you really trust the recommendations
if every recommender is a paid shill?

And the reason why is not necessarily…

It's like, oh, may- maybe they
are recommending the best products

and getting credit, but the issue
is the dog that didn't bark.

They don't even show the products
that don't pay affiliate commissions.

So half of the reviews that therefore
the AI is going to summarize for you

are this So if I wanna buy the best
blender, if I wanna buy the best socks,

if I wanna buy the best this and I ask
AI, AI hasn't tried these things on.

It can only summarize what is out there.

And what is out there, unfortunately,
a lot of it is, is, um, is

conflicted, to take the, the least
pejorative way of putting it.

So, so I think there are
a lot of, you know…

The, the, the goal of the
piece was we think that, um,

AI commerce is gonna be huge.

Absolutely.

This is, you know, Google
makes all their money this way.

It's a search engine.

ChatGPT is a search engine that
makes none of its money this way.

Like, do they covet what Google has?

Of course they do.

Is Google gonna lose this tomorrow?

Like, no.

Um, partially because there are
many, many different things.

Commerce is not just one, one catch-all.

There are different forms of commerce.

Instagram generates tons of commerce
through these impulse buys or, you

know, retargeting that they do.

You were looking for something for
a very, very long time, they might

as well steal that last click.

But there's a big area in the middle,
and that's where like CamelCamelCamel

to me is kind of like patient zero
of, like this is already a thing.

It's going to be a bigger thing.

And I'll, I'll stop talking in a second
here, but maybe the last thing that I'll

say is I kind of think of the world in a
binary of there are people, like when it

comes to, to shopping, there are people
that value their money more than their

time, and then there are people that
value their time more than their money.

And obviously there are more
types of people than those two.

But if you take that extreme, there
are a lot of these things on the

internet today that are designed for
people that value money more than time.

Um, and these are all the
cashback sites on the internet.

And it kind of requires a
technically astute, um, cheapskate

buyer, um, and I'm one of them.

So I, I, I know this market well.

Um, you know, and then but there are a lot
of people that just don't do those things.

They don't try to save money.

They don't try to use coupons.

They don't try to do any of the…

They would love to, but it just
takes too much time, and their time

is more valuable than their money.

And this is the thing that I think
AI will actually unlock, is that now

if it takes no time and it just saves
you money, it's more of an IQ test.

And I think more people will kind
of gravitate towards these tools

that have already been defined and
in use for a very, very long time

that appeal to people that value
their money more than their time.

And now they're applicable to a much,
much larger part of the audience.

And that's one of the many things that
AI in commerce is going to unlock.

Yeah.

So you have your five.

The three in the middle are called,
um, routine essentials, lifestyle

purchase and, and functional.

Which of those three do you
think are the first to go?

Like they'll, they'll start to get
eaten up by a- agentic commerce?

Well, I, I think it's, it's kind of, um…

Or maybe another way of putting it is
there is the, I know what SKU I want

Um, so or anything that has a UPC.

Like, the reason why Wayfair
has been successful post-Amazon

is they sell bar stools.

They sell all sorts of things that
don't have a, you know, universal

product code on the bottom.

So across those three…

I mean, I, I think it is those three
writ large, but the way that I would

think about it is one of the functions
of AI is I know exactly what I wanna buy.

Like, this is the item that I want,
and maybe I got that from AI, right?

So for, like, more of a lifestyle
purchase, it's like, "I want this

thing that's quite expensive.

Please tell me what to buy."

I spend four hours researching it, talking
to my friends, reading Reddit, watching

YouTube videos, and yes, of course,
conversing with ChatGPT or something

else, and then I decide what I want.

And once that decision has been made,
then it kinda goes into this, you know,

kind of one, one step up of impulse.

It's, it's not an impulse buy,
but it's like, "I want that SKU.

This is the exact item that I want.

Find it for me, you know, cheapest
shipping, lowest cost, best warrant.

W- whatever it is that I want.

In fact, even…

Here are my five different credit cards.

I don't know who's gonna
give me the best cash back.

I don't have time to figure that out.

I value my time more than my money.

Boom.

Go."

And it's actually that part that
I think is going to be across

those three, where AI becomes a,
a much, a much better co-pilot.

So you like the optimization
piece regardless of kind

of the- those experience.

Yeah, it works in every experience, yeah.

Who wouldn't want it
to be optimized, right?

Yes, exactly.

Who wouldn't w- and actually there
are a lot of people that don't

want it to be optimized because
like, who has time for this?

Yeah.

Like today, like I can go
to- If it does it for you.

Hmm.

Yeah.

I mean, I'll go to, you know, Rakuten
bought this company, Ebates, that flips

the affiliate commission into cash back.

Oh, maybe I should go to, not
Ebates, but like, you know, Mr.

Cash Back.

I mean, there are so many of these
sites that are out there and so many

coupons and like you could spend hours
doing this, and some people love it.

There's a whole site called slickdeals.net

which is kind of like filled
with these people, and I would

actually look to them first.

It's like watch what they're
doing, and whatever they're doing

everybody else is gonna start
doing once it becomes easy enough.

Um, but then for the…

You know, if you take like lifestyle
purchase as an example or anything

where reviews are going to be a key
part of the consideration process,

there are some things where you
have a magical experience with AI.

And I call it a magical experience because
I have a question, I get an answer, and

it's like holy moly, like that would've
taken me like two days to figure out.

Um, you know, my, my younger son, his
holy moly moment was he realized he

could use this to cheat on his homework.

We were not very happy
about that as parents.

We found out about that
several months later.

But it's like that's
such a good experience.

If you wanna cheat, it's
a great cheating tool.

Um, but if I, if I wanna cheat
shopping, like, so to kind of take

that, that, uh, unfortunate metaphor.

If I wanna cheat and shop, well, if
the reviews that I'm getting are not

magical at all, but are just, again,
kind of like still junk, I don't think

AI actually gives you a good experience.

Um, and then if I'm like,
"Okay, I got my shield junk.

It's saying I should buy this product,"
well, the optimization of buying

that product will still be good, but
that might not be the product that I

should actually want to begin with.

Um, and of course you could try to,
just like there's page rank, and we

wrote another, uh, piece about this,
how like SEO might become GEO, right?

Like search engine optimization.

You know, is there a an- is
there an analog for the AI era?

And there probably will
be, but that'll also…

Then, then it's this kind of cat and
mouse game between like Google knows

that people have reversed engineered
page rank and are buying, you know, page

rank 10 links and all these different
things to try to get themselves…

Like if you're selling tennis rackets,
there's so much value to being organic

search result number one, and it's
not based on the wonderfulness of

your product or experience or site,
like a lot of people game that.

That's probably gonna
happen in AI as well.

And that's-- But, you know,
th-this is a nascent industry.

Um, and I, I think at the, at the
end of it, for me, it's like the

optimization piece will see the,
the, the most traction there.

Um, it's the, you know, these more--
the more considered you get on the

curve from impulse to considered, um,
I would expect when, at least when

people do attribution and say, like,
"How much do I owe AI for this sale?"

The more considered that it, it is, I
think it's just the, the, the, the less.

Because of many reasons.

Like, there are many, many
different reasons why you

buy that BMW, as I mentioned.

Um, but also sometimes the reviews
that you're gonna get from AI might

not even be that helpful to begin with.

Um, and then the price optimization thing
for, like, buying certain, like, more, you

know, highly, highly considered purchases,
it's very hard to close the loop and

actually purchase those online end-to-end.

Like, try buying a non-Tesla car online.

It's very, very hard.

Um, you know, try buying a house online.

Like, I mean, you could kinda
do that in that you go meet the

broker, but no, you really can't.

So the attribution will
always be somewhat weak there.

How about, um, what about the
rumors that ChatGPT is gonna do

kind of a hosted, uh, checkout?

And in my mind, um, that makes it more
of a marketplace where if they, if they

have the checkout, then you're gonna have
to send them not only a, a data feed,

but like an inventory feed, just like
selling on eBay, Amazon, et cetera, and

then you're gonna have to get orders back.

Do you think ultimately they care enough
about commerce to go that direction?

Oh, I think absolutely.

I mean, I think it would
be crazy for them not to.

If you have 700 million active
users that are-- that have commerce

intent, that have already-- like,
many of whom have paid you, uh, and,

like, these searches are going to…

I mean, my, my-- I don't know,
obviously, uh, OpenAI's search volume

or what, what people are searching for.

They don't disclose this.

But I bet people are searching for
a lot of these commerce things.

They probably end up at Google at the end.

Um, and then, and this is one
of the reasons why, it's like

there are fewer searches.

If, if you need to do 10 searches to buy
something, and Google only gets one search

and that search is, like, the absolute
monetizable attrib- you know, the, the

commerce attributing search, that's
probably what's happening right now.

And the prior nine, some of those
might be happening in ChatGPT.

So it would be insane for them
not to go consume that 10th

search and keep you captive.

And maybe the hook, I don't know what
their hook is going to be, but maybe

the hook is the optimization piece.

Um, because you wanna make sure that
you don't do all, you don't do all of

the work and get none of the credit,
um, which is a problem that the

internet has had for a very long time.

Lots of people do all the work,
and then, like, the coupon site

at the end takes all the credit.

And this last click
attribution we know is flawed.

So, um, you know, AI has not changed that.

We haven't, like, somehow discovered
how to do attribution better.

Yeah.

And I didn't say it earlier, but shout
out for the CamelCamelCamel reference.

I was reading the article and I
was like, "Ho, that's a deep cut."

And I know everyone's gonna be like, "What
is he talking about with CamelCamelCamel?"

It's funny.

That's a…

Like, I use Cam- like, my wife u-
like it's, it has millions of u-

this pro- I think it's actually
Amazon's biggest affiliate.

It's, like, a real thing.

Yeah.

And- Yeah, it's got, like, the
cool chart, and you can kind

of figure out what's going on.

Yeah, it's awesome.

Do you think, do you think
Apple gets in the game?

It's kind of like when you start
thinking about all the personalization

that's gonna have to happen, they
know so much by being on device and,

and Apple Pay and all the things.

It, it feels like they're gonna have
to jump into this at some point.

I think they should.

I mean, I think the, uh, the fact that
they own the wallet And then there's

the second layer of optimization.

Oh, I'm buying a very expensive item.

Where should I…

What card should I put that on?

Uh, I have all this debt on my
Capital One card paying 24%.

Oh, I should just refinance
that to Bank of America.

They have a special rate at
16%, but it's too much work.

A lot of these things where consumers
are held back by inertia, and if

you control the wallet, well, you,
you've reduced the inertia problem

hopefully to, to close to zero.

But I, I, I would assume that
Apple, Apple has their own issues

around, like, they're, they always
kind of like to be the laggard.

But I think in this case they would
not like to be the laggard around,

like, how bad their AI stuff is.

Like, the fact that Siri can't figure out
how to turn on the flashlight or answer

anything correctly, and meanwhile I've
got this other thing called ChatGPT.

I bet auto accidents would go down
dramatically if Siri worked, right?

I mean, like, they should get that stuff
working first, and once they get that

working, of course it's a no-brainer.

Like, if you are the, um, the,
the, the search window, um…

And, and one of the ways that
this might happen, by the way,

is if it turns out that OpenAI
monetizes search better than Google.

And right now they, they definitely
don't, because they are the…

This is the point.

It's like they are monetizing with
a subscription, and all of the

searches they don't make any money
from, on an incremental search.

They lose money.

It actually costs them money.

It costs them electricity and water
and all these other things to go serve

up and do, you know, um, inference.

Um, but, uh, if it turns out, like, g-
uh, Apple gets paid tens of billions

of dollars a year from Google to be the
default in that little Safari search icon.

Like, if you don't type a URL and
you type a, a word, guess what?

That goes to Google, and then
Google shares a big chunk of

their revenue from the monetizing
search queries back with Apple.

Um, if it turned out that OpenAI could
do that, um, wow, like, they would switch

from Google to, uh, to OpenAI, no-brainer.

But, like, right now it's just, it's
too nascent for them to do that.

So to answer your question,
I think no-brainer, yes.

But it kind of…

I think the order of operations where,
like, make the AI stuff actually

functional and then get into the
hardcore monetization or kind of the, the

no-brainer, uh, uh, monetization, which
to me is, is a lot of the commerce stuff.

Last question.

Do you think the agents…

Like, maybe they skip some of this coupon
code 'cause it's, like, so inefficient,

and maybe they just negotiate.

There's a buyer and seller,
and they kind of negotiate.

And then we've been predicting, and I
think A16 or, or Mark made an investment

in a group buy, like 15, 20 years ago.

There's been group
buying things all along.

I could see agents, like, start to
kind of say, "Hey, you and Alex are

both looking for an Eight Sleep.

You know, what if you, you're connected
through this, you know, social graph.

What if you went together and said,
'Hey, if you buy two of these,' and we

glom on some friends and suddenly we, we
negotiate a 20% discount or something."

Yeah.

I, I, I can see that coming.

I mean, I think the thing that's
exciting now is that when I was running

TrialPay, I mean, we, we were pretty big.

We had billions of dollars
of GMV go, go through us.

Like, um, probably 15 million consumers
did something with us every month.

It was, it was a, uh,
interesting business.

But convincing merchants to do
things that made sense for them.

Like, I hated these coupon things.

It's like we were actually
accretive to a merchant.

We were sending net new
customers, and meanwhile,

RetailMeNot, I hate that company.

They don't even exist anymore.

But like RetailMeNot would
steal the last click.

The site's dead, yeah.

I mean, and, and you'd have
these dumb merchants that were

like, "Oh, I love RetailMeNot.

They're sending me so many
customers," because they were confused

between correlation and causation.

Like, they didn't realize that like, no,
they weren't sending you any customers.

You already had the customers,
and you're just paying them to

send you your own customers.

The c- the thing that's actually
very, very interesting to me about,

uh, ChatGPT going from like zero
to 700 million we- weekly active

users or whatever crazy number.

It's like one tenth of the
Earth uses this product now.

And every merchant is like, "Holy moly,
I have to, I have to do something here."

So they will be able to dictate the,
like what should the interface be?

What should the attribution be?

I think you'll actually be able to
rewrite some of these things, not by

happenstance, but kind of like, you
know how like New York has Central Park

because it was centrally planned, and
like you go to Boston and the streets are

all like, you know, confused like that.

I love Boston, but it's like
there's a certain benefit to

order, um, and orderliness.

Because navigating streets that have
names that make sense and like a

central area, like that could be better.

So I, m- my, my estimation
would be there are kind of three

things that are gonna happen.

You're gonna have the big head of
commerce, and those guys will play

ball instantly because, you know, if
you're h- if Home Depot is, is second

to being on ChatGPT and Lowe's is
first, like Home Depot CEO's in trouble.

So the big head of commerce, they
will get behind this right away.

Then there's the kind of like middle
part, which is all wrapped up under

the, you know, Shopifys and the Magentos
and the BigCommerces of the world.

Well, they have the platform
to lean on, so they're gonna

integrate with this right away.

And then that kind of
leaves the long tail.

Um, and the long tail will be the one
that's a little bit more idiosyncratic.

But I, I think things will get
rewired pretty quickly just because

you have this giant gorilla, or
even set of gorillas, and so much

potential commerce up for grabs.

It's going to be fascinating to watch.

I agree.

And I promised you we'd be out on
time, so I'm gonna call it there.

Um, leave them wanting more.

That was excellent, and hope to get
you back on, uh, after you write your

next piece to get you to talk about it.

All right.

Really appreciate it, Scott.

Appreciate it, Alex.

Follow, rate, and share to stay up
to date with RetailGenic, your guide

to the future of AI shopping agents.