Retailgentic | Consumer Behavior & Retail Trend

In this episode of the Retailgentic Podcast, Scot talks with Stu Solomon, CEO of HUMAN Security, the company analyzing 20+ trillion digital interactions every week to detect bots, fraud, and now agentic AI behavior. Stu breaks down:
  • Why bots now exceed human traffic
  • The 1300% growth in agentic activity
  • How HUMAN identifies good vs. bad bots
  • The rise of agentic logins, navigation, and even checkouts
  • Why spoofing of ChatGPT and Perplexity traffic is exploding
  • What retailers need to know about letting AI agents crawl their sites
  • The coming world of agent-to-agent commerce
If you’re preparing for the future of retail, ad tech, AI commerce, or online security, this is essential listening.

Timestamps
02:20 — What HUMAN actually measures (bot vs human vs agentic)
03:40 — Stu’s background (military → Amazon → banking → VC → CEO)
06:40 — The logistics + retail connection
09:00 — How HUMAN works: media + enterprise security
11:00 — The digital journey from ad → login → checkout
12:11 — HUMAN at 20 trillion interactions/week
16:55 — Why HUMAN started publishing agentic data
17:50 — 1300% growth in agentic activity
18:30 — Machine traffic surpasses human traffic
19:40 — The diversification of generative AI traffic
20:30 — Agentic login + website navigation behavior
24:00 — How HUMAN distinguishes bot vs agent vs human
33:00 — The bad side: spoofed ChatGPT & Perplexity traffic
38:30 — Retailers blocking bots without realizing it
47:00 — The inevitability of agentic adoption
48:40 — The future: agent-to-agent transactions
49:30 — The end of 2FA hell?

👉 Connect with Stu: https://www.linkedin.com/in/stu-solomon-733a217/
👉 Learn more about HUMAN: https://www.humansecurity.com/

🧠 Want to stay ahead in AI commerce? Subscribe and follow along:
📰 Subscribe to the free Substack: retailgentic.ai
📺 Watch episodes on YouTube & Subscribe for updates: https://www.youtube.com/@Retailgentic

Creators and Guests

Host
Scot Wingo
CEO and Co-founder, ReFiBuy.ai
SS
Guest
Stu Solomon

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).

AI has gotten into the public conscience
very recently and in, and in, in an

increasingly fast pace because of the
amazingly rapid, um, introduction of

generative AI into our environment.

You know, I, I think I heard
the other day that OpenAI is the

fastest adopted consumer product
in the history of mankind, right?

Uh, b- and, and so it leapt,
it leapt into our, our public

conscious and understanding.

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 checkouts?

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.

Today my guest is Stu Solomon.

He is the CEO of Human Security.

Right now you're probably wondering,
"Hey, I wonder why Scott has on

someone from a cybersecurity company.

Isn't this podcast about agentic commerce?

It's right there in the
title, Retailgentic.

What's going on?"

Well, it turns out Human's focus is
measuring agentic, non-agentic bot,

and human activity, understanding their
various intents, and either blocking or

allowing their activity, not only at a
bot level, but at a very granular level If

you're like me, that sentence generates a
lot of questions, and we're gonna get into

pretty much all of them on this podcast.

I learned a massive amount about bad
agent behavior, good agent behavior,

various intent signals, and Stu shares
some really interesting data about

what they're seeing historically and
where he thinks this is all going.

I hope you enjoy the conversation
and learn as much as I did.

Welcome to this episode of Retailgentic.

I am excited to welcome Stu Solomon,
CEO of Human and fellow North

Carolinian, we just discovered.

Uh, so that's kind of fun.

Uh, I'm in Raleigh, you're in Charlotte.

Stu, welcome to the podcast.

Scott, thanks so much for having me.

I really appreciate it.

Absolutely.

Absolutely, excited to do it.

How long have you been
in the Charlotte area?

So I moved down here in 2006, and
originally with Bank of America many, many

moons ago, as most Charlotte- Charlotteans
have done, and, uh, and stayed ever since.

Yeah, we get that B- B of A influx.

Um, once you've been here for
about, uh, five years, you're

basically a North Carolinian.

I, I feel like at this point
I'm definitely a native.

We have so many incoming that
the, the OGs are swamped.

Well, cool.

Let's, uh, let's start, let's
start a little bit more about you.

So we now know you were at
Bank of America and Human.

Tell us a little bit
more, uh, about yourself.

Yeah, absolutely, and thanks.

So I, I'm actually a, uh,
a military guy by trade.

Uh, US Air Force officer, and, uh,
actually still serving in the, uh,

the Delaware Air National Guard of
all things for a North Carolinian.

That's a fun weekend trip right there.

It, it is.

A, a quick jaunt up the highway,
the old, uh, US Airways route

up to, uh, Philadelphia.

And, uh, e- enjoying, uh, a
opportunity there to, uh…

I actually command the, uh, the entire
Air National Guard for, for Delaware.

So that's kind of my,
my little side hustle.

And, uh, I Made my way, just like
most, uh, folks in Charlotte, uh,

through banking, uh, particularly on
the cybersecurity side of the world.

Ultimately, uh, got into, uh, the,
uh, venture capital space and the

cybersecurity space, and built and
sold a number of different companies

for both, uh, venture capital and
private equity over the years.

Uh, most recently part of a,
uh, a wonderful transaction,

uh, to, to MasterCard.

And, uh, came over to Human in
2004, where I've been the CEO

since January of last year.

In your Air Force days, to the extent
you can share whatever, um, I don't

want any national secrets or anything.

Which I know you would, you would
not say, but, uh, um, so we know you

have leadership, you know, 'cause
you're running a, you know, what must

be a pretty big operation in your
current, um, you know, role there.

But did you, did you go as an aviator,
or were you more like ground ops?

Were you in cybersecurity and always
kind of had a, a, a interest in

that from, from that perspective?

Yeah, it's, it's a great question.

So I actually, I, I
started out on active duty.

I graduated from the Air Force Academy
as a, uh, Russian history major.

And so what would a Russian history major
with a Bachelor of Science degree do?

I became a supply and logistics officer.

Um, I, I did get to do a little bit
of flying, but I got into a really

interesting technical expeditionary
logistics, figuring out where spare

parts needed to go all over the world
for the fighters and bombers of, of our,

uh, fleet during the, uh, the, the war
period, uh, in Afghanistan and Iraq.

And, uh, then I, I left
active duty in 2005 and, uh,

joined the Air National Guard.

At that point, I was just entering
into banking, and so the cybersecurity

parts of my life on the banking side
became really interesting to the

military and vice versa, and then they
just became interwoven over the years.

And is that how you,
uh, ended up at Human?

So ul- ultimately, I guess
in, in theory, that's true.

Uh, I, I knew many of the, uh, the board
members and investors at Human from

various other endeavors that I had done
over the years, and it just became a

serendipious, a serendipitous, uh, uh,
scenario where we just came together

in the right time and right place.

Yeah.

It's, uh, it's really interesting because
a lot of military folks, uh, are in the

back- back office side of the retail.

So once you get into fulfillment centers,
you, you tend to meet a lot of, uh, you

know, veterans in, in that role, and
it's because of the supply chain part.

You know, whenever the, the…

all of our various armed branches have
to be able to move so many things around

so quickly into different theaters and
everything that, you know, suddenly

standing up a distribution center in
Iowa is not, not that challenging.

It's like if this was in Beirut,
it would be a lot harder, but, you

know, over here in Iowa, in Des
Moines, it's like not, not that hard.

Well, and, and ironically enough, and
we obviously didn't talk about this

ahead of time, but my very first job
after active duty, I moved to, uh,

New Castle, Delaware, right outside of
Wilmington, Delaware, where Amazon's,

uh, largest distribution center on
the East Coast was back in 2005.

That was my first job.

So I was actually, uh, running the
night shift building out, uh, their,

their distribution center at the time.

It's on the tip of my tongue.

What's their three-letter
acronym for that one?

I don't remember off the top of my head.

It's been so many years.

I know they have like DFW one through
eight now, and, uh, the Delaware one,

ah, man, it'll have to come to me.

Well, at, at the time I
believe there were only three.

So there were, there was, uh, New
Castle, Delaware, there was Louisville,

Kentucky, and I think there was
Henderson, Nevada, and that was it.

That-

Yeah, Kentucky's CVG.

It's usually after their closest airport.

And they, now they have like CVG one
through, I think, 10 or something.

They have like a ton there 'cause that's
where their air, their air operation is.

Uh, and I had a, a co-founder that was
in the Army, and I don't know if you

guys do it, but he was always like…

he always…

he would never say supply chain.

He would always call it, uh, let's
see, uh, boots, bullets and beans.

There you

go.

'Cause you gotta, you gotta move people,
you gotta move food, 'cause once you

move people, you gotta have food.

And, uh, all the fighting branches
are not very effective if you

don't have the bullets, so.

Usually how that works.

That's right.

Cool.

So you know a little bit about
retail having worked for…

Did you work for Amazon proper

or- No, th- this was Amazon proper.

Uh, at the time there was nothing else.

Yeah.

Got it.

Wish I had kept all those stock options.

Yeah, that, that was, that was
a good, good investment there.

Um, and then when you went into Human,
uh, you, we now know you're CEO, like

what was your progression inside of Human?

Uh, so, so I came in as the CEO.

Okay.

So just like jumped right to the top.

Didn't start at the mail room, skipped the
mail room, went, went right to the top.

Skipped the mail room on this one.

That's right.

Uh, but, but over, over the years as
I, uh, after I left, uh, banking, I

moved over into the venture capital
world and kind of built up some

startups and then moved into later
stage private equity companies.

And I've kind of run that gamut of
everything from a general counsel to a

president of, of the business, to a COO
of a business, to the CTO of a business.

And so this was kind of the, uh, the
next step on that, on that journey.

So give us a, give us an idea of what
Human does and a little bit of, you

know, feel for the sc- just the scale
and scope of, of your operations there.

Yeah.

It's, it's always fun to talk about that
'cause I think that Human is probably

one of the largest and most significant
companies that most people have never

heard of, but they actually are protected
by and influenced by on a daily basis.

If, if you do, uh, any kind of, um
interaction between your digital

environment and a computer or a
phone, you probably are touched

and protected by Human in some way.

So there, there are really
materially two sides of our business.

There's a media side of our business, and
there's an enterprise security side of the

business, but both of them are predicated
on the same singular focus, which is

this historical notion that machine or
bot-based activity is used at scale to

be able to manipulate and create fraud
throughout that entire kind of customer

journey between-- from the time that a
digital ad is first conceived in any kind

of environment, in, in, on a website,
in a podcast, in a connected TV or a

connected device scenario, when, when a
advertising scenario occurs where digital

ad content is created and then it's
married to a publisher, there is all kinds

of potential, uh, fraudulent activity
that occurs both pre and post the bidding

on that space that places a digital ad
in your environment in the first place.

As that fraud begins to occur, it gets
scaled by automated or bot-based activity.

Human sits in the middle of that scenario
to be able to start to build a layer

of visibility into the potential for
manipulation or fraudulent scenarios

associated with digital ad content.

Then once that, once that digital ad
content gets placed, you now put it

in front of the demographic that you
expected to, to put it in front of,

and there's all kinds of potentially
fraudulent manipulation, um, fraud-

fraudulent clicking, fraudulent
redirection of different scenarios.

Human is sitting in the middle of
all of those scenarios as well to be

able to ensure the sanctity of that
relationship between the digital ad

content and the intended audience.

Now, the next step, let's say, um,
either a human or a machine, 'cause

it could be either, um, appropriately
accesses that digital ad content.

What does it do?

It builds some level of brand
affinity and probably takes you to a

website if you've clicked on it and
you wanna go and you wanna log in.

And now at that point that you want to
log in, Human's sitting in the middle

of that as well to make sure that
it's not, um, fraudulent, um, or, or

malicious activity that's authenticating.

You get into that website.

Next thing you wanna do is you wanna go
shopping or you wanna set up an account,

you wanna do something, we're sitting
in the middle of that as well, right?

To protect you all the way
through to your checkout.

So to-- so that entire kinda digital
journey from ad placement all the

way through transactional behavior,
we have different components of our

capabilities that help to identify,
um, potentially fraudulent or malicious

behaviors and/or more importantly, to
protect good stuff and, and to be able

to allow for that journey to continue.

And to give you perspective on the
scale, um, you know, it's, uh, we're,

we're seeing over 20 trillion digital
interactions on a weekly basis, so a

few trillion times a day tenths of a
millisecond, we have to make decisions of

good or bad, yes or no, right or wrong.

And historically, these things have
been predicated on this idea that it

was a very binary scenario where a
human meant good and a bot meant bad,

and that's simply not the case anymore.

Now, in, in a more modern world, bots
are actually outpacing humans on the

internet, and most of them are doing
things we want them to be doing.

So that, that's kind of the human
journey and, and where we are right now.

Uh, when you say trillions,
that's a big scale.

How many, uh, you know, again,
to the extent you're comfortable

disclosing anything, give us an
idea for how many customers you

have, what segments they're in.

L- like, one, one kind of, you
know, for dummies question, I'm,

I'm the dummy, I don't know your
business, is who pays for all this?

So, like, is DoubleClick paying you?

Is the advertiser, the publisher?

There's, like, a lot of people, you
know, there's even-- there's like 18

constituents in the digital ad world,
and enterprise is probably easier.

I'm sure the enterprise is paying you.

Uh, and like, so who pays you?

How many customers do you have,
and like, how do you think of

certain segments that they're in?

Yeah.

I, I think a, a good way to think
about it is we have many hundreds

of clients, um, across, across the
globe, and those clients will tend to

fall into two camps, as you said, on
the kinda enterprise security side.

So when you get to the point where you're
getting challenged and, and going through

an authentication at a website, it's
gonna be the, the brands, the retailers.

Um, we, we tend to, uh, to f- to
focus very heavily on the people

that would want to have a website
and transact with you in some way.

Um, and, and that could be on the services
front or it could be on the retail front.

On the media side of our business, we, we
tend to historically have built products

that would really focus on the platforms
where advertisers and, and agent- and

agencies come together with publishers,
so think, uh, DSPs as an example,

uh, but n- but not exclusively so.

We also, uh, have an, a suite of products
that are attractive to advertisers,

brands, and, and to publishers as well.

And then, um, I assume private company,
or are you more like private equity?

We are a, a later stage private
equity-held company and, um, have,

have, uh, wonderful financial sponsors.

Interestingly enough, because the
company has come together through,

uh, an amalgamation of multiple
companies br- being brought together

on a singular platform over the years,
I still have early stage venture

capitalists who are still invested
with us and much later stage private

equity invested in us as well.

Um, our primary financial
sponsor is, uh, Goldman Sachs.

Oh, good.

Yeah.

Know them well.

I've been on those boards where, uh,
I had a company, ChannelAdvisor, where

I had my series seed and then, like,
my series D people on there, and like,

you know, the D people are, like,
wanting to talk about, you know, shaving

off a point of margin or something.

Like, this, like, really esoteric thing.

And then the A guys are like, "You should
try these 800 ideas," and like, you

know, and it's like, whoa, all right.

A lot, a lot of feedback
from, from all over the place.

It's kind of, it's a fun management trick
to get them aligned on, like, what's

mat- what's important for the business.

We, we literally have, uh, people that
have been A round investors all the way

through v- late stage private equity.

The other thing that's really unique in
our environment, and actually I think

the coolest part about our business,
is that I have people that have a, a

cybersecurity lens and focus, um, and
then I have individuals who have, and,

and both investors and board members who
have more of a, a media business focus.

So I actually have this really interesting
dichotomy of not just different economic,

uh, scenarios and, and aggressiveness
as you've articulated, but also

those that just simply have different
world lenses on the economy as well.

So it's, it's a pretty
cool, uh, discussion.

So one of the reasons we reached out
to you guys to be on the pod is you've,

you've started to put out some really
cool, um, research and, um, you know, data

around what you're seeing with agents.

So, um, say a little bit about that.

So, so I, you know- Kind of reading
the tea leaves of the reports, it

seems like you guys are sitting there
as this layer, and you're seeing a

lot of activity going on, and, and
you've been doing some reporting.

Um, is it specific to, um, you
know, purchases, digital ads,

or is it all-encompassing across
your whole sphere of things?

So, so where's the data come from, I
guess, is the, the overarching question.

And, uh, give us an update of, you know,
when I first started seeing your data,

uh, the numbers were, were, were big.

And, and then they've just gotten even,
like, more, more, you know, more bigger.

Yeah.

A- and, you know, so at first
they were kinda shocking, and

now they're just, like, head, you
know, mind exploding shocking.

Uh, so, so tell us a little bit about why,
why did you guys start publishing this?

Where's the data come from, and some
of the data that, that you can share?

Yeah.

Ab- absolutely.

So, you know, as, as we talked
about a little bit earlier, this

notion of 20 trillion digital
interactions in and of itself just

sounds, like, completely ridiculous.

But what's, what's more important is when
you start to peel the onion back into

these, um, big, frankly, societal-changing
activities that we're seeing today,

particularly with agentics, what it,
what the data's starting to tell us,

and I'll go into more detail on this
in a second, is that it's actually a

today problem, not a tomorrow problem.

Uh, the, the rate of adoption and the pace
of adoption, and the change in the dynamic

of the, of what we're seeing in the data
is leading us to believe that there's a

rather significant movement, even quicker
than we thought, into agentic behaviors.

So I'll, I'll go into some specifics,
which I, which I think are very relevant.

So obviously when you look at all
that data, you know, those 20 trillion

interactions, let's say a subset of that
is going to be bot-based activity, and a

subset of that is gonna be human activity.

Within that subset of bot activity,
there's actually, um, a significantly

increasing presence of agentic bot
activity, and we'll, we'll kind

of talk about what that means as,
as we go through the conversation.

But I, I think the, the place
to start with is to, to think

about a couple of key concepts.

The, the, the first concept is we're
at an inflection point right now where,

where not just we at Human with our
rather, um, large install base, but in

general, the internet is very rapidly
hitting an inflection point where there's

more machine or, or bot-based traffic
than there is human-based traffic.

So we hit about a 50/50 split
or inflection point this year.

But, uh, there are many studies out
there that are, that are pretty clear

that it looks like u- upwards of 80%
of all internet traffic is actually

gonna be machine or bot-based activity
within, uh, the next 10 years.

So, uh, we've already hit the,
the, the inflection point, and

it's only going to increase.

Now, of that traffic, a portion of
that is obviously agentic in nature

The agentic piece is, is the thing
that we have-- we've seen some really

unique observable scenarios around.

So just since we started looking over the
last couple of months, let's call it over

the last four months, we've seen over a,
a 1300% increase in the amount of traffic

that's specifically agentic in nature.

So it's, it's proliferating
is, is the bottom line.

But, but let's go a
little bit deeper on it.

It's, it's fascinating to see that
over time, not only is the amount of

traffic that's either attributable
to a-agentic traffic or to generative

AI traffic is increasing every month,
and it's increasing rapidly, but the

mix of what makes up that traffic
that we're seeing is starting to

change and become more diversified.

So, um, generative AI traffic that
was attributable, for instance, uh,

primarily to, um, research activity going
on using ChatGPT as an example, as the

predominant amount, uh, back, uh, even
two months ago, uh, we're now starting

to see that there's more traffic and even
more of that traffic is now diversified.

We're actually seeing more
perplexity, and we're seeing more

of other generative AI browsing
scenarios that are starting to occur.

So you start to see a diversity
of that traffic that's leaping

off the page right now.

What, what's, what's really
interesting though is that even

within that, as that continues to
diversify, we're seeing an uptick in

specific observable agentic activity.

So agents going in and logging in.

3% of all the traffic that we see
right now that's agentic is actually

login and-- login activity that's
authorized in an agentic way.

Another 6% of that that we're
seeing is actually related to

clicking on pages and websites.

So these behaviors that are indicative
of leading to ultimately an outcome

that we fully expect, which is
what comes next, a transaction.

And, and all this is observable
in traffic that we're just seeing

within our install base today.

Imagine what it's gonna look like
three or, or six months from now.

If we cross this year, seems like we
should get to 80% faster than 10 years.

I don't know, I don't know how long it--
I mean, it took us, I guess, since the

inception of the internet to get here.

But I bet the, I bet the trend tells
us it's not gonna take 10 years.

Do you think that 10 years is…

I, I think you're gonna probably say
you would take the under on that.

Yeah, I, I think you would take the
under on it, but I think one, one

thing that becomes a natural limiter
to that is human desire to accept that.

So a-as, as I alluded to earlier, there's
always been this presumption that if,

if you're a bot, you're, you're bad,
and if you're a human, you're good.

Humans are the natural limiter that
add friction into their environments

and controls to stop that from
happening as quickly because they

haven't made that mental leap yet.

So, so, so I think that, I think
that there's a natural limiter there.

I, I think the other, the other component
to it is that we simply haven't built

all of the protocols to be able to
determine not just how do you establish

visibility and how do you ba-back up that
visibility with trust, but then the next

step is what, what authority and autonomy
do you give these agents to act with?

I think a-as we build more of the
protocols to be able to build that trust

layer with the a-ability to have the
authority and autonomy assigned to an

agent with verifiable credentials, it'll
continue to, to kind of exponentially

create a hockey stick of adoption.

So I, I think it'll probably be a little
bit slower over the next couple of maybe

year or two a-as things really settle in
from a technical control perspective, and

then it's just gonna leap off the page.

So it seems like you have a taxonomy that
is human and then bot, and then bot has

two sub flavors, non-agentic and agentic.

Is that, is that kind of a way of s-of
thinking about this- I, I think that,

that's a great- … taxonomy in your…

Okay.

Yeah.

And then so, so we're at 50/50 or
we've passed it, um, human and then

the totality of the two types of bots.

Within the bot category How do you
tell if it's agentic or not agentic?

I don't want any secrets, but like, yeah.

So, so, you know, just I'll caveat that so
I don't have to always say it, but yeah.

Not at all.

I appreciate that.

I w- I would go back really quickly
and say, and even with human or

bot, it could be good or bad.

It…

And-

Yeah, there's like a whole another

It could be a good human or a bad human.

It could be the right
human or the wrong human.

It could be a good bot or a bad bot.

It could be the right
bot or the wrong bot.

And my dog, uh, is making
a guest appearance, so-

Guests are welcome.

Yeah, you, you can't stop a-

Is that a Weimaraner?

It…

She is a Great Dane.

Ah, okay.

And she's larger than I am,
so we'll let her keep going.

I have no, uh, no, uh…

I can't tell her size.

I- if I saw her legs, I
would know the difference.

150 pounds of love.

So, um, but, but, but as, as you
kind of contemplate those notions,

there is an element of intent
that becomes really important in

the conversation and important to
establishing that trust as well.

Is the bot d- acting or is the human
acting to create an outcome that

you expect or want it to create?

And what is it trying to do becomes
part of that, that, um, matrix

as well as you think through it.

W- uh, w- with that in mind, there's
a couple of ways to start to think

about an agent acting in the right way.

Um, a- as I mentioned a moment ago, an
a- a- an autonomous or an agent actually

is an autonomous activity with some level
of verifiable credential, with some level

of authorization, with some level of
accountability, and with some level of,

um, of learning over time that starts to
emulate behaviors and anticipate behaviors

that you would want it to do, that you've
authorized it to do, that you expect it to

do, and that you can have accountability
on the outcome once it's created.

So all of, all of those different
elements have both technical aspects

to them and behavioral aspects to them
that help you to be able to triangulate

and create a, a position of trust at
the point in time at which it starts to

act, and then in the persistent trust as
it acts, and then trust in the outcome.

So in my world, we'll just pick on
ChatGPT, but Perplexity and all the

others have similar kind of setups, right?

So, so in my world, you've got ChatGPT.

They, they have a, a bot, GPT
bot that is used, a crawler that

is used to crawl sites, um, for
training and other th- purposes.

They have another one that
builds their web index.

They have yet another one that
kind of like goes and checks prices

at retailers in near real time.

And then they have the thing that
was previously called Operator

that is now called Agent.

Um, and it, it, it has two
modes where it can operate in a

virtual machine, server-based,
or it can live in your browser.

Um- Yeah.

Are all those agent or bot?

I, I guess they're all bot, but w-
but can you distinguish on your side,

or you're just more looking at intent
for, for how you look at things?

No, no.

So we, we, we can distinguish on our
side and, and you made a, a really

important distinction as well in, in,
in your question, which is there's gonna

be a browser-based scenario, and then
there's also an API or MCP scenario.

So there's automated connections
that, that are programmed, and then

there's also the behavior within
your browser environment as well.

Both of those are two different flavors
of ways that you could start to understand

and triangulate on, on that trust
factor that I mentioned a moment ago.

But those are, those are two separate
scenarios based on the way that

you expect it or want it to act.

So whe-whether it's pulling in
pricing information, or whether it's

automating a workflow, or whether it's
connecting the dots between a, a, a

clear text prompt and an outcome, or
increasingly, if it's one agent talking

to another agent, both representing
humans that aren't even in the middle

of the transaction at this point.

So I imagine in your world, you're
seeing all this traffic and other

signals, these fact patterns, and
you're having to, in, in near real time,

kind of come up with a, a decision of,
okay, you use all that information.

You're, you're trying to put together
the stream of that to get an intent,

and that's how you're kind of judging
these things as, as things go along.

So, so you, you don't have, like, enough--
You don't have perfect information.

You're having to constantly
figure that stuff out.

Is that kind of like the
scale of the problem?

At, at the scale of 20, uh, trillions
you're, you're, you're, you're doing that.

Uh, yeah.

I would actually look at it
as a multi-phased approach.

So you add different elements of
friction at different components to it.

So the…

It's really a triangulation of intent,
behavior, infrastructure, and then

I would also add that verifiable
credential component to it as well.

Um, so, so, so maybe it's more of a box
than a, than a triangulation, but it's

taking all of these different signals,
wrapping them together, and going down

the path of ensuring that you can trust
that thing to do that thing at that time.

And then just to-- let's just repeat
this number 'cause it, it is the big one.

So 1,300% increase.

Um, and can you- Can you look at
that and, and segment it in some

way to attribute it to something?

You, you mentioned a
couple of them, right?

You, you talked about a
couple, but say more about that

At, at the highest level, and,
um, we, we could start to break it

down into generative AI kind of use
cases, particularly around research.

Uh, but then there's also s- um, indexing,
authorized scraping as an example.

But I think that's almost a negative
connotation and not meant to be.

It's, it's authorized discovery of
information and taking that information

and feeding it into these models
or feeding it into these outputs.

Then the n- the next piece, uh,
after that, that you would start

to see is, um, auto- automation
workflow management scenarios

that leading to, um, an ability…

We're seeing much smaller, but,
but increasing amounts of traffic

specific to automated login activity
on behalf of a human via an agent

Then the next component after that is
website navigation, uh, via an agent.

So imagine a scenario of go
to this box, click here to be

able to create a dropdown menu.

When you get to the dropdown
menu, select a thing.

Those are automated steps that a human
would normally have to navigate to

be able to get to a buying decision.

The human's probably navigating
a workflow of some kind.

Yeah.

Be it enterprise or a consumer
purchase or something.

That's right.

But now the agent is doing that for them,
and we're actually seeing that, that

behavior in increasing amounts leading
to ultimately we're actually starting

to see actual agent check, checkouts.

That's all we talk about in my world.

Um, and let's talk
about that a little bit.

So, you know, in my world, uh, and
I'm coming from, you know, the user,

and you're, you're kind of this, this
layer in there, um, which is important.

But, um, you know, in my world,
this-- it's been a very interesting

year because we've had all these
standards announced, right?

So we had this agent-to-agent
protocol, um, and, and, uh, Visa

and MasterCard have their own.

There's the payment people
have like a bazillion of these.

Uh, and then we had Google come out,
and they did AP2, and then now we have

the new one from, uh, OpenAI, ChatGPT,
which is ACP, Agent Commerce Protocol.

Uh, and, you know, it, it's
kind of married with the

checkout that they rolled out.

Um, do you guys interact at that layer?

Uh, you know, because it's
almost like how the data…

You, you probably have to understand
these standards 'cause s- you

know, signals coming in with…

Like, I remember the, the Google
one has this little manifest

associated with it, every little
thing, and, and that they all…

Are you allowed-- Are you in
this layer that can see that?

And that's like the manifest, you know,
you, you say, "Show me your manifest."

It's like a passport control
kind of thing, and it does, and

you're like, "Okay, you're good."

Do you guys get involved in
kind of that protocol layer

of these agentic frameworks?

Yeah.

I'm, I'm so happy you asked the question.

I, I, you know, if, a-as you well
know, because you're familiar with

the protocols, I was basically, uh,
quoting Google a moment ago with

verifiable credentials going into
authentic, authorized, and accountable.

Um, look, choose your protocol.

It doesn't really matter, but the point
is, is all of them go back to that notion

of verifiable and authorized, uh, before
you can even get to the accountability.

We play in those first two steps.

So re-regardless of the protocol, by, by
understanding and identifying the agent,

creating a trust layer on that agent, and
being able to pull back to that intent

component, you could start to get into
the authorized and, and, um, uh, the,

the authorized components in particular.

Let's talk about bad bots.

Um, the, uh, in my world, you know, the,
the, the things we typically are looking

for are Uh, you know, credit, bad credit
cards, you know, people using us to,

to, you know, inject bad credit cards.

Uh, and then another layer would be de-
denial of service ti- kind of attacks.

Uh, then we have some, you
know, various script spoofing.

Now we have prompt spoofing
or whatever that's called,

uh, and those types of things.

So, so that's, that's the bad
behaviors we're most tuned into.

And then there's always the someone
trying to get even underneath all that

and get to, you know, some underlying
employee data or, or, you know, God

forbid, customer credit card, which
is the big honeypot, which is largely

protected by the PCI side, which is
kind of old school now, and now we have

these tokenized card systems that, that
seem to be very robust on that side.

Uh, it's always interesting, most
of the big things that are happening

now come from the point of sale,
the, the older point of sale

systems, but that's a digression.

But I, I'm smiling because that's where I
actually got my start in the, uh, cyber-

B of A, you know all this
stuff better than I do.

Um, and, uh, so w- is that the kind of
bad bot behavior, or this is totally

different bad bot behavior we're seeing?

You nailed it.

It's exactly.

So when, when I talk about the enterprise
side of our business, um, starting really

with helping to identify and potentially,
uh, prevent malicious redirect scenarios

and getting into client-side protections
at the browser level, leading to, you

know, things like CAPTCHA scenarios.

So how do you look at CAPTCHA or caption,
CAPTCHA lists ability to make sure

that there's not a bot manipulation
at the point of authentication?

Therefore, we've spent a ton of time
really working on things like account

takeover or credential stuffing
scenarios that would lead to that

fraudulent, um, login in the first place.

That's where we put a significant
amount of, of our technical

friction into the process, and
that's where we first introduce it.

And, and we don't like CAPTCHAs because
it, uh, decreases signups and logins.

So, so we like security, but we don't
want anything to slow the money flow.

So there's, so you're a nice solution.

Ex- ex- exactly right.

And then when you get to the other side
of it, are they using a, uh, stolen or

compromised credential or, or credit card?

Um, are, are they, uh, trans…

Once you've authenticated, are you
demonstrating behaviors that are

not indicative of a well-intended
human who has the, all the rights

after authentication occurs,
uh, leading to a transaction?

These are all the things both pre-
and post-authentication that we

would do from a, a behavioral anomaly
perspective or from a technical

perspective before that authentication.

Got it.

And are you seeing agents do these things?

That's an interesting question.

I-

Agents are on the, the s-
the stu- n- naughty list.

What I would say is, uh, uh, uh,
we're, we're, we're seeing more of

the abuse right now, and I think it's
just indicative of, of the volume

of traffic we're seeing, is we're
seeing a significant amount of very

concerted effort to spoof, um, where
you're coming from, particularly

in the generative, uh, space.

So, uh, uh, you know, very concerted
efforts to really focus on looking like

you are a, an appropriate generative
AI capability that you're interfacing

with and presenting yourself that way
to websites or to individuals, and

you're simply not Um, you know, o-of
our observables, I think o-over a, a

small period of time, we saw, like,
one in 18, um, scenarios that were

specifically generative or purported to
be generative AI just simply weren't.

And the numbers were outrageously worse,
uh, for the larger, uh, functions.

Like one, one in six ChatGPT observances
that we had were purported to be ChatGPT,

were actually spoofed by, uh, by bad guys.

So you're seeing it a lot on
the, on the generative side.

Um, when you get into the agentic side,
I think that, um, it's inevitable.

Uh, look, auto-automated workflow is,
is inevitable in these scenarios, but

we just haven't seen a ton of that yet.

So I kind of wonder-- So my guess
of what's going on here is…

This actually is a question I was
gonna ask anyway, so I'll just ask the

question, then I'll tie these together.

Um, you know, we, we frequently--
Again, we're involved, uh, in this other

side, so we, we help people optimize.

W- you know, people that wanna be
involved in agentic commerce, and

they're kinda like, "I want ChatGPT and
Perplexity to have my whole catalog.

I'm, I'm-- You know,
Google already has it.

I am totally fine with that.

I'm not gonna give them any of my,
you know, uh, uh, you know, my, my

customer data or anything like that
or whatever, um, a-and certainly

not sales rank or anything, but
I want them to have my catalog."

The catalog kinda like almost
wants to be free in my world

'cause it generates sales.

So catalog out, sales in, so
everyone's kind of incented to do that.

And then, but then, you know, they'll
come to us, and we have this whole

framework, and then we'll start
going through the framework, and then

we'll look, and they're actually just
blocking the, the all the suite of

bots that are gonna get it in there.

So we're like, "Let's start there."

So that, that's kinda
where we always start.

Like, let's look at your infrastructure
and make sure you're not just

blocking one of these bots.

'Cause it is a long list now because it
used to be Google Bot, you're good, right?

And then Bing Bot or
whatever everyone's called.

And now, like, each of them
may have three or four.

Like Perplexity has three or
four, and so on and so forth.

Um, so I imagine what's happening is
there's some, you know, let's give it

the benefit of the doubt, and it's,
like, some entrepreneur that's trying to

build a shoe pricing engine or-- which
is like one of these really aggressive

categories, um, you know, because of
these drops they do and everything.

And, you know, I'll, I'll pick
on Nike and, and, you know, Nike

is, you know, got a white list.

They're not on the white list, and
therefore they're, they're trying to say

to you, "Hey, Sudhu, I'm, I'm ChatGPT.

I'm on the white list."

And, and, uh- Human is, is kind of saying,
"Mm, you know, I, I don't think you are."

Is, is that kinda like the situation
that we're, we're seeing here?

Yeah.

I, I think there, there are a couple
of subsets to that point, uh, for sure,

which is, yeah, may- maybe you wanna
get discovered or indexed in, in a

particular LLM because you want, uh,
your pricing to be available and your

product to, to surface through the model.

But then you also wanna make sure that
there's a connection between your data

and your pricing and what's presented.

So the sanctity of, of that
relationship actually has to maintain

from a source sanctity perspective.

The, the, the next one is that, uh, you
know, maybe you wanna be discoverable

from, um, you know, the, the, the
next bastion of, of SEO, right?

So from search engine optimization,
as I move into agentic discovery,

you want to be discoverable, but you
have to index the right way with the

right kind of keywords and the right
discoverability to even be found in

the first place so that you do surface.

But then on the flip side of it, that,
that there are plenty of other companies

that we're seeing that, um, their data is
their value And so on, on the flip side,

you have to monetize that in a controlled
fashion, whether it's intellectual

property, which is a whole separate
conversation, or it's just simply the

rates that are published, um, on your
website for plane tickets or, um, mortgage

rates or things that, uh, competition
would traditionally nefariously

scrape and then be able to undercut.

Well, it's just, it's still
an agentic problem as well.

It's not just a, uh, old internet problem.

So there's also the notion of how
do you protect the, the data that

you want to monetize, even if it
doesn't need to be discoverable.

And we see that as a specific
use case as well that we're

helping clients with today.

So it's kind of across
that whole spectrum.

So, so let me read that
back so I understand.

So you're not just like saying,
"Okay, you're ChatGPT bot, come on

in, you can have the whole site."

You're getting very granular
in saying you can…

There's parts of the site you
could cordon off if you wanted to,

and maybe even parts of a page?

That's right.

The, the, the parts of page
and, and specific components.

And, and that's where, where,
where that really leads to is

you build this layer of trust.

It starts with visibility.

Who's touching me?

Going back to that intent, what
are they doing when they touch me,

and what are they trying to create?

And then the next piece is,
great, now that I have the

visibility, how do I control it?

And so putting kind of group-based
policies in place categorically,

I don't wanna do this or that, I
don't wanna see this or that, but

then in a more fine-grained way,
maybe it gets down to the specific

component of a page or a data field.

So go- going group policy, fine grain
policy, uh, leading to trust that

ultimately starts with overall visibility.

How do I know what my traffic is?

Yeah.

Is that usually how you deploy it?

You like s- you, you go through a
monitor reporting phase where you're

like, "Look, 8,000 people are just
hammering this one page to get the price."

That's right.

That's right.

And, and here's like some, some,
a pie chart of who they are, and

then the customer says, "Okay,
I wanna turn off this, that, the

other, and this, that, and…"

Is that kinda how it works?

They, they, they kind of get a read on
it, and then they start pruning it down?

Yeah.

Quite, quite literally, it's a traditional
user interface where you've got a

characterization of all your traffic.

It's broken down by here's what
bots look like, here's what

agentic looks like, here's…

And now let's start to
apply policies against it.

This is kind of an in, in the weeds
question, but we get in the weeds here.

So, uh, so, so, you know, Google's
confusing because they kinda

have a lot going on, right?

So you've got the traditional
Google bot, you've got, uh, Google

Shopping, which uses the Google bot,
but you can also send it a feed.

Um, and then, you know, then
now they have AI Overviews.

So, so Gemini powers all this.

Um, and I actually don't know, does it…

I doubt it has a separate bot.

It probably just uses…

Do you know, does it use just standard
Google bot for the training data?

And, you know, can, can you-- Like, let's
say I said to you I don't want any of

the Google AI stuff to touch my site,
but, but core Google is fi- like, you

know, traditional Google search is fine.

Is that-- I don't even know
if that's, like, split outable

on the Google side or not.

I, I honestly don't know.

Uh- Yeah.

Uh, me neither.

Okay.

Yeah, yeah.

I'm not sure why anyone would even
wanna do that, but I was just always

curious, like, you know, the other
ones are so clear on, you know, we

have a bot for these things, and then
th-therefore, at the, at the, you know,

the meta le-level, you can punch out
the, you know, de-whitelist or not

whitelist the ones you don't want.

But, but it seems like Google's
like, "We've got this giant

straw, and everything's going
through this one straw."

So, interesting.

All right.

Uh, do you guys get involved in…

W-what other, you know…

So, so I've read a bunch of your content
that talks about agentic readiness, and,

and when we talk about that, we're talking
about your product catalog and, you

know, the amount of content and context.

What do you guys mean
by agentic readiness?

So I, it starts, I think, with visibility.

Um, as, as I've talked about, understand
your surface area of risk, understand

your surface area of opportunity.

Uh, when-once you can catalog that,
then it's build your profile around

what you need to establish for trust.

And, and by the way, we're, we're
talking about this a lot in, at the

latter stages of, um, kind of an agentic
transactional relationship in e-commerce.

But this is exactly the same paradigm
that we apply to, um, to looking at

eight, at, um, at ad traffic as well.

So be-being able to very clearly
articulate that, uh, your expected

behaviors in your ad traffic are,
are verifiable, uh, is, is another

application that we do here as well.

So it's not just on the, uh,
transactional side of it.

In, in the ad world, do you get
involved with the buyer of the

ad where you say, "Look, this
percentage of views or clicks were…

you know, we think were not true.

You need to go back to the…"

Do you get involved in that discussion?

Quite literally.

So, so we'll, we'll, we'll get
in-involved in, historically,

particularly pre- and post-bid invalid
traffic determination, and then we

also do viewability, brand safety,
other verification kinds of scenarios.

And then we also do click defense,
uh, scenarios to, to start to

contemplate, uh, actual clicking
behaviors and determine the validity

of those, those clicks as well.

Okay.

Uh, I think I took a little side
journey there when we were talking about

the bad bots and some of the things.

Were there any more that you were
gonna list off that I interrupted?

I, I think we've, we've
circled through most of it.

This may be something you, you can't
answer, but like, uh, in my mind, it

probably starts as just like, you know,
kinda shady actors, like some entrepreneur

trying to do something to get some data,
and then it starts to, to progressively

get more aggressive to nation states.

You know, like a lot of the cybersecurity
people, I see the, I see the penetration

people talk a lot on CNBC that, you know,
some high percentage of what they see is

other nation states trying to impact our
infrastructure and just be disruptive.

Do you guys get involved in that
whole world, given that you're Air

Force, you know, maybe that's…

And, and if you don't even wanna
talk about that, I understand.

No, uh, look, I, I, we, we don't, we don't
directly focus on that part of the world.

But I mean, sufi- suffice to say,
it's, it's a really simple notion

that, uh, all the bot activity
that we're discussing here, um, it,

it's just simply automation, right?

It's a- it's automation of connections,
and those connections get a certain

amount of scale when you take
the human out of the loop, so-

Even for bad actors.

So if you're a manual bad
actor, you're soon gonna be an

automated manual bad actor, so.

I, I think that's a huge part of it.

I think the other part of it is that,
um, there's an increasing, uh, in-

increasing reduction in the barriers to
entry to be able to do this because of

the automation component and because of
professionalization of the availability

of these kinds of things to, to be
readily, um, used by an individual.

A, a nascent skilled individual can
create an outsized impact, and that

outside in- outsized impact is then just
scaled exponentially when you start to

talk about the use of, of these tools.

So what, what it, what it really means
is that these basic notions that we

have around, uh, nation states are
sophisticated, individuals are not

sop- as sophisticated, and they have a
different volume of activity that could be

created and impact that could be created.

That's just all going away.

Um, the, the, those barriers to
entries just don't exist any longer.

And, um, I, I'll go a step further.

It really boils down to intent.

You know, a bad guy might wanna monetize
something, a bad nation state might wanna

steal something, get some information.

Um, some- somebody who has a social
cause that they're driving may just

wanna be able to proliferate a message.

Um, so, so the intent changes the attack
profile, but what doesn't change is the

ability to scale through automation.

This whole intent thing
is very fascinating to me.

Must be, uh, it's like hard to, hard
to Sherlock Holmes all that out, but

I'm sure you guys have figured it out.

The, uh, in our last five
minutes or so, let's…

So we've, we've kind of
talked about today backwards.

Let's talk about today forwards.

Do you spend time thinking about, um, you
know, what does this future li- look like?

So you're, you're already starting to
see these, these crazy numbers, 1300%.

Is that gonna be like 5000% next year?

Or you, you kind of hinted there's
like, you gotta take the human friction

element into it and, and, and, you
know, and, and folks like yourself are,

are part of the, the tool set to, to,
you know, put some friction in here.

Do you, do you see this actually…

Maybe I'm wrong.

Maybe you, maybe you say, "Nope,
we're gonna clamp down on this, and

it's gonna go back the other way.

It'll, it'll be 90%
human traffic over time."

So what, what's your prominent, you know,
any forecasting or, or trends you, you

can think are gonna happen going forward?

Yeah.

A, a, a couple of things.

One, there's no way this,
this train gets derailed.

It's, it's gonna continue rolling
down the tracks, and it's gonna

pick up s- uh, speed as it goes.

I think it's an inevitability
at this point, but it's an

inevitability for a slightly
different reason than we think about.

You know, I, I think that Um, AI has
gotten into the public conscience

very recently and in, and in, in an
increasingly fast pace because of the

amazingly rapid, um, introduction of
generative AI into our environment.

You know, I, I think I heard
the other day that OpenAI is the

fastest adopted consumer product
in the history of mankind, right?

Uh, but-- and, and so it leapt,
it leapt into our, our public

conscious and understanding.

Uh, but the notion of generative AI
wasn't even a thing in January of

this year, and now it's something
that everybody in, in your audience

is talking about and planning for.

They haven't quite figured out what
budgets they have to apply to it or

what problem it's going to solve or
create, but they know they have to do it.

But y- but the reality is that,
um, this, this AI movement

didn't start with generative AI.

It goes well before that.

I, I would argue it goes back decades
into machine learning and, and to data

science, and these other things that all
led to this idea that machines, it turns

out, can actually move faster than the
human brain and make connections that

are more complex than the human brain.

Now, how do we optimize our data?

How do we organize our data?

How do we process our data to be
able to enable that to happen?

Where, where things are really
taking off now after decades of,

of learning and building, is that
anybody can interface with it.

So you don't need to be a data scientist
to be able to go in and manipulate 20

trillion digital interactions anymore.

Uh, but what you do need to do is be
able to have the logic to be able to

store, query, and use that data to
create a three-dimensional picture

that makes sense, and then to apply
it to an automation that you have

a level of trust associated with.

Those, those notions of the trust
and the protocols that, you know,

as you picked up on, I think that's
where the friction comes right now.

But those are rapidly gonna be broken
down, uh, because everybody in the

world, in the technology world is looking
at this and trying to figure it out.

And as soon as those walls come down, it's
gonna, it's gonna be adopted even faster.

I think it's an inevitability.

Um, the, the second point that I would, I
would, you know, make sure to think about,

I think the, uh, you know, G- Google has
been doing a great job of talking about

this as they opened up their new protocol,
is that the entire internet is predicated

on this idea, or the control environment
in the internet is predicated on this idea

that there's a human being on the other
side of a computer screen Well, in, in

the, in, in the new agentic world, that's
just simply not going to be the case.

It's really gonna be one agent talking
to another agent, both with some

level of authority and autonomy that's
been granted to them by a human,

but that human is gonna be at arm's
length from that relationship now.

So wh- whereas you and I know who we are
and we have a true identity, and that true

identity takes on verifiable credentials
in a digital form, that's gonna very

rapidly have a layer of abstraction,
um, that, that we don't have today.

So all of, all of the, the
paradigms that we understand

about the control environment are
simply going to have to shift.

And as they shift, that's gonna be
the other natural friction point

that will slowly be removed, which
will accelerate the adoption here.

Yeah.

I just hope I can get out of this
world where I have to do 18 two-factor

authentications a day and, and look at
some authenticator app 50 other times.

So, so here's my agents, I, I
trust them, and then I can just,

like, not have to log in again.

We'll, we'll see.

We're counting on…

Come back in a year.

I want you…

Y- you gotta solve that for us, Stu.

We're counting on you.

I'm on it.

All over it.

Think, think of the GDP loss from all
the two-factor authentication going on.

Exactly.

It's gotta be 5% or something.

Yeah.

Um, uh, thanks a lot, Stu.

I, I learned a ton from this conversation.

Uh, I'm gonna have to go take a nap now.

My head hurts, uh, from all the
information that we've jammed in there.

Uh, if folks wanna…

You guys put out a lot of great content.

Where, where can they go to find
the content that, that either

you or the company produces?

Yeah.

Our, our company is rather
prolific in social media.

Please, uh, please find us
particularly on, on LinkedIn.

You can follow us there.

We, we publish a couple
of things a day, frankly.

Also, humansecurity.com

is our, our website, and, uh, it,
it's got a wonderful section with

all of the different content that
we develop, everything from webinars

to white papers to research papers.

Um, I'd also be remiss if I didn't say
that we have a, uh, a sub-brand in our

organization called Satori, and Satori
is our, um, our intelligence team that's

constantly looking at and, and developing
data based on observables and linking

that back to all kinds of unique campaigns
and activities that are out there.

Some really, really amazing
work that that team does.

Yeah.

And that's, like, for more,
like, security nerds or-

No, it, it runs the gamut.

In fact- Okay … uh, they, they've
been part of some really interesting

and prolific takedowns of, um, supply
chain compromises of Android devices,

as an example, under something that
we call Bad Box and Bad Box II.

There's more to come on that, by the way.

Um, a- and, and we recently did a,
a thing around Slop Ads, which was

really looking at, uh, compromises
inside of, um, of application stores

that were being compromised right up
front to be able to, uh, to trick, um,

individuals into downloading and then
compromising from there, um, all the way

to more traditional security scenarios.

So some, some really, really
amazing work that that team

does that's worth looking into.

Very cool.

Well, thanks for coming on the pod.

Congrats to you and the team at Human
for, you know, seems like, uh, you're in

the perfect place at the perfect time.

So, so congrats on that.

Uh, yeah.

It's like, wow, I, I wish someone
had software for helping me figure

out what all this agent traffic
is, and bloop, you're, you're

right there at the right time.

Congrats on that.

Thank you so much.

Really appreciate it, and, and thrilled
to be able to represent the amazing

folks at my company, so thank you.

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

to the future of AI shopping agents.