May 16, 2025 • 37 mins
In this episode, Pascal is joined by Scott Wingo, a pioneer in the e-commerce space, to discuss the concept of 'liminal space' and the rapid changes brought by AI and geopolitics. Scott shares his journey from growing up with a mainframe in his home to navigating the fast-paced shifts from desktop to mobile and now AI. The conversation delves into the complexities organizations face in adapting to technological advancements, highlighting Martex’s Law and the need for a non-binary approach to change. Scott explains his 70-20-10 framework developed at ChannelAdvisor to balance core business activities with innovation. The discussion also covers the future of AI-native organizations, the rise of agents in e-commerce, and how leaders can manage these agents effectively. Scott emphasizes the importance of staying updated through podcasts and other high-signal sources while also sharing insights on potential disruptions in traditional search models by AI-driven platforms.
00:00 Introduction and Liminal Space
00:59 Technological Evolution and AI
03:15 Organizational Change and Martex Law
04:15 Framework for Managing Change
07:34 AI Native Organizations
11:31 Staying Updated in a Rapidly Changing World
17:36 Diving into Agent AI in E-commerce
18:56 The Potential of Agent AI in Everyday Life
20:52 Managing a Company with AI Agents
22:06 Building an AI-Driven Organizational Structure
28:42 The Future of AI in Business Efficiency
31:26 The Shift from Google Search to AI-Powered Tools
33:58 Conclusion and Final Thoughts
↗ Scot’s new venture: ReFiBuy.ai
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:00):
You're the boss agent and you give it all this context and I
want you to spin up a new, but what you're doing is what we're
doing is we're backing out of it, if that makes sense.
So we started with ICS and the engineering team was managing
them. And then they're like, we could
write an agent to do what we're doing because I'm babysitting
engineer one is babysitting the front of this pipeline.
And then it gets to the point where you get and you write a
(00:20):
prompt and then you're like, nowyou're managing that thing and
you're like, I could actually, man, I could write an agent for
that. So that's like how we've backed
into it. And it ends up being a very,
very clean job description. So fascinating.
It's funny because it's in some ways it goes back to the
beginnings of the Industrial revolution and.
(00:42):
It is a very fortification of yes, exactly.
We're going back to the conveyorbelt.
Yeah. Disruption.
Hey everybody, welcome back to Disrupt Disruption, where we are
going to explore what the liminal space holds for us, the
(01:06):
weird twilight zone between the old world order and the new one,
which hasn't quite emerged yet. I'm Pascal Vinette, your host,
and today I'm incredibly excitedto have a very dear friend and a
very special guest with us. Scott Wingo is a serial
entrepreneur with an exceptionaltrack record of building and
(01:26):
exiting successful tech companies.
As the CEO and cofounder of Refiby dot AI, he's leveraging AI
technologies to revolutionize ecommerce experiences.
Scott previously founded and LEDa series of successful startups,
Stingray Software, which he soldto Rogues Wave,
auctionrover.com, which was acquired by Yahoo, and Channel
(01:48):
Advisor, which he grew from startup to IPO on the New York
Stock Exchange. He also Co founded Spiffy, an on
demand car care service operating across multiple
cities. Beyond building companies, Scott
serves as a general partner of the Triangle Tweener Fund,
investing in early stage startups in North Carolina.
He Co hosts the popular Jason and Scott Show podcast, which I
(02:13):
highly recommend you check out, focused on e-commerce trends and
was a Forbes Technology Council contributor.
And probably more important thanall of this, both Jane, my wife
and I used to work for Scott when we were at Channel Advisor.
With that being said, let's dig into the conversation.
Scott, it's amazing to have you on this podcast.
(02:36):
We go way back and as a true pioneer in the e-commerce space,
I'd be incredibly curious on howyou think about the liminal
space. You have seen it play out many
times in e-commerce and the broader Internet context.
And maybe to get us kicked off, like, how do you think about the
period of time you're currently finding ourselves in with the
(02:59):
advent of AI changing in like even geopolitics being a big
topic. But I'd be curious to hear in
from you. Like where are we in this moment
in time? Yeah, it's a crazy moment in
time. Thanks for having me.
It's good. Always good to see.
See you, Pascal. Any other either of the
phonettes are always amazing people.
The I didn't know what liminal was, and I'm from South Carolina
(03:20):
so that's a big complicated wordwith lots of weird consonants
and vowels in it. So thank you for explaining to
me what it is. I would say I've been at this a
while and because I have a math orientation, it feels like it's
a great time to be alive becausewe're on this logarithmic curve.
And so when I was growing up in this small town in South
(03:43):
Carolina, I've always been a misfit toy I wear that's a
positive to me. Everyone in my town either was
like dirt poor or they worked for a nuclear facility in my
city. So it's kind of like The
Simpsons, if you will. And then my dad was an
entrepreneur and we had a vax, amainframe in my house.
It's probably the only computer in my town and it was a
mainframe. So it's very weird, right?
(04:04):
So to grow up with computer. So I've started, I've lived with
a mainframe in my house and thenI got a desktop computer and
I've been through this whole curve.
And you know what's really wild is from mainframe to PC was a
pretty big chunk of time. And then if we Fast forward, you
and I have lived through desktopto mobile.
We've even gone from build your own infrastructure to cloud,
(04:26):
from desktop to mobile. And then what I would say is the
logarithmic IT, we're on that log curve.
So now things are just happeningfaster and faster.
And the evidence we see of this is there's all these charts that
show different consumer things and how long they took to get to
100 million. Color TV took 50 years and the
microwave took 25. And now Chachi BT I think is the
fastest and you see the lines and most of them have this kind
(04:48):
of curve and now they're starting to go vertical.
So as a technologist, exciting, it's exciting and entrepreneur,
it's exciting because these things we're, we always tend to
embrace them and ride those curves because on the other side
is opportunity. And I think it's this AI 1 is
(05:10):
pretty crazy. And I've gone through my own
little 7 steps of doubt. And then the Gentex side is
where it can actually do stuff is really bonkers.
And, and that's what I'm workingon because I got, so I got into
it and I'm not one to jump on these things.
I didn't do anything in the crypto space, for example.
Not that I don't like crypto, I could just never really get my
head around. I couldn't get as excited about
(05:31):
it changing the world as what we're seeing with the AI stuff.
That's a lot. Let me pause and see if you want
to dig into any of that. Absolutely.
Be curious. So you're talking about the
increased rate of change, which I think we undoubtedly see when
we talk about this quite a bit. The flip side of that is in the
(05:51):
space of organizations, there's something called Martex law,
which basically says that organizations change
logarithmically, but mostly linearly.
So it's like widening. Yeah, widening gap between
what's possible and what you canactually absorb in the
organization. The context would be to say as a
leader, you need to be very careful to take and choose the
(06:14):
change you want to implement in your organization because
otherwise your organization justcan't absorb it worked.
You've been on the set up side, you've brought a company onto
the New York Stock Exchange and you have worked with very large
incumbent legacy organizations as clients.
I'd be curious to hear from you.How do you think?
(06:34):
Like what do you advise your clients on the change now?
Like with your new venture, you will talk to legacy businesses
and you will want them to change, but they can only absorb
so much change. Yeah.
And I have a framework I developed at Channeladvisor,
which is the company that took public.
And so we're 800 people, we're public and we're in the
(06:54):
e-commerce space and it's just changing.
You and I, you were very early on this kind of Shopify trend if
we will. So a SAS based relatively
inexpensive. You came to me with that idea
and I was like Pascal, that's crazy and you're right that you
were. So what you have to do is you
have to because you have a largeorganization.
We were like 800 a thousand people.
To your point, you can't just say, all right, everyone, we're
(07:16):
pivoting to this thing. But I hate binaries as an
entrepreneur, I hate these decisions of do nothing or do
everything. And so I always try to split the
binary. And the way to split this binary
is to have more of a allocation matrix.
And I like 70 2010. I learned this from the Google
10% time where they allow engineers to, if you talk to
(07:36):
anyone at Google, they don't do this anymore and it's just PR.
But anyway, there it was a good story.
They would allow engineers to take 10% of their time and work
on out their projects. And that's where the lore is.
That's where way MO came from. And some of these ideas, again,
I'm not really exactly sure that's true, but anyway, so this
allocation matrix is good because you basically say, all
right, you've got your core business and just use the
(07:57):
channelvisor example. We're helping people sell on
eBay and Amazon and we can't take our eye off of that, but
we're big enough that we should take 20% of our time and do
things. I kind of use a, a, a, a dart
board, if you will. So a target.
So 70% is like right in the red zone is where we're going to put
our effort. We're not going to stop
maintaining and customers are there.
Then we're going to 1 ring out and that's the 20%.
(08:19):
So that may be in this context of channelvisor, that may be OK.
There's this marketplace, this company in China called Alibaba
and they've got this team all let's today at 0.
But we think we're going to spend some 20% time on that.
So you know, and in an organization of 800, that's like
maybe 100 people kind of working.
It's actually lower, oddly enough, because usually it's the
(08:39):
engineering piece. So 800 people, we're not going
to really put sales on it, but we have 100 engineers now.
We have 20 working on something a little bit out there.
And they don't only do one, but they do 3 or 4 projects.
It can't be this infinite list either.
So you pick has a few be on one hand under 5 things in the 20%
time. And then the 10% time is where
you're like, hey, there's this AI thing we should have of our
(09:03):
150 engineers, we probably need 10:00-ish that kind of are
tracking this thing and doing some experimenting around it
because it right now we're not sure, but it could go.
And then over time, some of the 10% things don't work out and
they die. So there's like a little
incubatory kind of thing. It's almost like managing a
portfolio of startups, if you will.
And some of them get product market fit that they pull into
(09:25):
the 20% piece. So you put more Reese's on them
and then they make it into core.So I like that framework and
that's a framework I then now because I've run it, I espouse
that when I talk to you, any retail, any clients of any kind.
And I did the same thing at a totally different company called
Spiffy and we had, I think I'm getting better at it because the
(09:45):
most of our ideas actually gained pretty good fruit and
moved from the 10 to 20 70. I love the non binary version of
this because typically what you hear is I don't know like the
20% rule which means like the hard boundary between 80 and 20.
They also seem to be hard or we pretend there are hard
boundaries. I love the gradient in this.
(10:06):
The thing you hear people talk about lots these days is this
idea from whatever you want to use as a framing, but like from
AI Infuse to AI native and everybody talks about what does
the accounting firm look like inthe future, which is AI native?
What does the e-commerce firm look like, which is AI native?
What does the apparel business look like, which is AI native?
(10:30):
I'd be curious, be curious, how do you think about AI native and
what does it even look like? Because I feel a lot of people
say AI native but they have no idea what it actually means in
real terms. Yeah, and so let's say we're
we're in this mythical thousand person company and again, this
binary thinking. And I run across this a lot
(10:50):
where, you know, I meet JP Morgan and Bank of America and
all these people and none of them, they're not tracking AI
because they have a stiff policyat work.
They have some internal AI that kind of stinks.
It's llama one on some bad rag database or something.
It's like ChatGPT, the 1st 2020 Chat 222 ChatGPT.
And that's I feel bad for them because there needs to be
(11:12):
someone in this organization that is thinking these things
through. So what I would do is I would
take 10% and I would say, all right, you're AI native.
What does that mean? Whenever you start building
something in my world of software, first of all, the
engineers are going to start with looking at the tool set.
You don't do it. A lot of this again, you don't
want the tail to wag the dog. The idea is let's try to be AI
(11:33):
1st and say, all right, let's doan analysis.
What are the tools out there? There's wins surf and cursor and
the copilot knowledge has and inthis part of the org you're
going to give them infinite flexibility, maybe even each and
because we're in this, this is 10% innovation mode.
Maybe one engineer does cursor and the other does windsurf and
another in fact, variety is actually good.
(11:54):
Whereas in the rest of your organization variety is bad.
And this like little 10% incubator thing, you want to
encourage it because you get more learnings from that.
So that's phase one. But then as you get into go to
market, so every function has anAI native kind of aspect to it.
Engineering is the easiest one because that's where there's
like the most uptake. So that that one, you probably
don't have any questions on that.
But then the second one is now what are you going to do is go
(12:16):
to market. And this is where you want to do
a lot of experimentation. But again, you don't want the
tail to wag the dog. I'm actually a little counter
the popular opinion on this one because there's all, I run into
all these startups that are using these AIBDRS and this kind
of thing. And I think, and I'm the
recipient of a lot of that. I'm sure you, you get quite a
bit as well and you can tell every time.
(12:37):
And I, I think it's going to be interesting because that may be
where an area where certain parts of go to market, the part
that doesn't touch the customer,I think can be AI native.
So you're using these Otter meeting notes and there's a huge
value in all those things. The there's no, back in the day,
we would spend half our time putting stuff in Salesforce that
can go away, all those things. That's where I think you'd be AI
(12:58):
native. But when it comes to
authenticity and talking to the customer, I think you you say,
all right, we stopped there. I think the value of human to
human interaction is going to goway up in a world where there's
all this AI noise. I don't know if I'm answering
your question, but every function I think has to evaluate
this. But you can't overdo it a lot of
(13:19):
times. I remember the first time we
implemented Scrum at my company and the Scrum Master was just
running around insane because wedidn't have the right Scrum
master. And I forget what the other one
is. There's what I would call a
product manager. They split it into two roles.
We had six products. I was like, I'm going to have to
hire like 50 freaking people forScrum.
We're going to do it this way. And they're like, you violated
all the rules of Scrum. And I was like, OK, then they
(13:41):
quit on the spy. You can't be such a purist that
you make the wrong decision for the business.
And you have to infuse these things into your DNA in the way
that is the best thing for the customer and for your
organization. And then again, if you're in
this 10% model, I like a lot of variation in there so you can
see what's working. Yeah, love it.
Curious in this context, how do you, it's probably a two-part
(14:02):
question. One is as a leader, how do you
stay on top of the massive amount of change we you talked
about like this idea or this exponential curve kind of goes
vertical at the moment in terms of adoption and developments,
etcetera. Maybe even you personally, how
do you stay on top of all of that stuff?
You're in the AI space where, you know, I do a decent amount
(14:23):
of reading in the AI space and Ifeel like, I feel it's dizzying
as in new model comes out like every three days and it like
tops the old model and like all this kind of stuff.
And this is the second part of the question.
How do you deal with that in terms of making sure your
organization doesn't go crazy, your people don't go crazy, and
at the same time they stay up todate?
(14:45):
Yeah, as the person that sets the strategy, I've never really
had a chief strategy officer. I always want to do the
strategy. I've never understood chief of
staff or this may be controversial, this may be a hot
take. I've never understood chief of
staff or chief strategy officer because that's basically my job.
And at the end of the day, if the founder isn't doing that, I
have questions about the whole thing anyway.
(15:05):
So my job is to ingest all this information, process it.
And my immediate thing I work onis a mental framework.
I really like this idea of I need to framework.
So to your example, you've got all this information coming in
and my current framework is all right, we've got the frontier
models. And I think we all agree now
that those are commoditizing. There's going to be some
(15:28):
interesting things going on in there and we'll keep track of
it, like the new llama for and this kind of thing.
But that that's in this quadrantof it's stabilizing and we know
what's going to happen. The the cost per token is going
to continue to drift down. And then the interesting areas
are for us is some of the because we're focused on this
one area where like hyper, any piece of information that comes
(15:50):
about this intersection of e-commerce, retail and agentic
where agents are shopping for you.
So open AI, this is very timely because this the news is
happening like logarithmically, right?
I had this idea in October, November of last year and now
every day a new news release is coming out.
When I started Channelvisor, I had an idea that there'd be a
(16:11):
lot of marketplaces and we supported eBay and it took seven
years for Amazon to come out andthen another five years for the
next marketplace. And then suddenly they, so it's
that logarithmic curve. It's crazy.
But then now I have a framework for that and it's actually in
our name and it's kind of which of these things are helping the
consumer research, find things and buy and that I will sort all
(16:31):
that and then share it into the org and that framework.
And then a lot of times what I'll do is I'll send a signal
that says this is just information.
So as a, the other thing I've learned as a CEO, what you say
and distribute, sometimes I just, I like to make sure
everyone sees everything. We're overwhelmed to buy it.
So I'll now scale it and say this is just an interesting
news. No, just doesn't really change
anything for us. It's but it validates our idea
(16:52):
here. Open AI is coming out with
something in this space. I'm like, this is pretty
important. We need to think a little bit
more about this or do that. For example, Open AI just came
out with this ability. The retailer can they have these
cards in the chat and the retailer can send a different
data feed to have that. So that's that impacts us.
So that one, I'm like in my matrix of consideration, that's
one I need to share into the organd make sure that everyone
(17:14):
understands. So I tried to build a framework
on what it is that's going on and the ultimate framework.
Is either this going to change our strategy?
If no, then share it or don't share it if it's high quality.
At the Super top of this, the framework is where is the best
signal to noise ratio. And I'm sure you spend a lot of
time on this too. So as people that consume a lot
of information that's and it's been interesting over time, it's
(17:37):
really changed. We used to read magazines and
books and then we've had podcasts, but they weren't as
good as they are now. And then we went through a phase
of you could social media, you get a lot of info, but now the
noise ratio there is pretty high.
You have to figure out where to go get the best information to,
which is another challenge in today's world.
How do you stay up to date like very concretely like what is
(17:58):
your like information diet, so to say?
And also I'd be curious just to as a general of like
observation, are you spending more time now or significantly
more time now keeping up to datethan you did 10 years ago, 15
years ago when you are in the Channel Advisor world, for
example? Yeah, I would say it's the same
time. The density is like 10X, OK.
(18:21):
The throughput of the amount andquality and throughput of
information because reading a book is interesting, but it just
is not a fast way to consume information.
I would say podcasts are definitely where I get a lot of
information. The other, the other part of the
signal, the noise. So a really good example is the
All In podcast. And this is a controversial 1
because it has politics, but it matters because I'm a venture
(18:44):
backed. My side hustle is, I also have
AVC fund, but I'm also a venturebacked startup.
If you're any venture backed startup, you need to get out of
that mindset that oh, this is political or whatever.
Because every VC listens to thatpodcast and they form their
opinions from there. So if I don't listen to that,
I'm going to be behind the curve.
And it also has a very high signal to noise ratio.
It is very, I don't care about politics, that part's boring.
(19:05):
But when they don't talk about politics, this concept that
Chamath talks about of 8090 and he said that one weekend and
then within three days or founder mode or all this stuff
gets amplified there. So in my framework, that's a
must listen thing. And then I have other ones like
Lex Friedman tends to have really good interviews that have
a big impact and so on and so forth.
(19:27):
I love my favorite piece of content over decades of
stratechory with Ben Thompson that I highly if anyone takes
any one thing from this podcast,subscribe to that is worth every
penny. It used to be a long read, which
was hard to get to every day, but the podcast version is what
I've pivoted to. I really enjoyed that.
I was just earlier listening on my way into work.
I was listening to an interview he did with Mark Zuckerberg
about llama stuff, and it was excellent.
(19:48):
Yeah, I find that I get a lot more information out of the
podcast, and there's very specific ones you have to
navigate to to get really good content.
And of course, your podcast is the best.
All in podcast is number 1 is number 2.
No, fair enough. Thank you for the shout out
here. Let me increase the signal to
noise ratio on this podcast and double click on the area you are
(20:11):
currently working on and spend like most of your brain power on
which is AI, particularly agentic AI in this e-commerce.
Broadly speaking e-commerce space genetic AI very clearly is
like the topic at the moment. So we have pivoted from just AI
to everybody needs to put a genetic in their in their
frameworks and names. I'd be curious to hear from your
(20:33):
perspective, where are we on this journey?
Are you feeling we're like it's good enough already?
We're like seeing actual real productive use cases.
It's very early. How do you see this?
And then related to that, as we're talking about this like
weird in between space, the Twilight Zone is agentic, this
kind of piece which unlocks thisTwilight Zone, which gets us
(20:53):
from like the old way of doing things to the new way.
I think maybe robotics because you worked for the singularity,
so you can go way deeper on thisthan I can.
I'm just trying to consume wherewe are on the curve and you have
the ability to look further out than I do.
So in my little view, let's lookat a 18, two years, 18 months, 2
years. I do think Agentics going to
unlock a lot of people that a lot of you know, the Aunt Jenny,
(21:16):
normal people that aren't as deep in this.
The chat thing is just not, it doesn't do anything right.
It's a way to get information and whatnot.
But at the end of the day, it's not like accomplishing anything
other than that. And what Agentic does is it
gives you have a little LM kind of a little brain and it has,
it's consumed all the Internet. So it's very wide and deep.
So it's a square of knowledge. And in some ways it's almost
(21:38):
like too wide and too deep. And then what most of them do is
you take, you want to narrow that to solve a specific
problem. So you use a retrieval augmented
generative database. And, and my layman's way of
explaining that is you take thishuge brain that can see
everything and you focus it through a lens.
And you say, I really only want you to think about this.
And by the way, here's some moreinformation about this that you
may not have because it's not out on the web.
(21:59):
And then you give it some tools to go do stuff.
And that's where it gets interesting in, in my little
sliver. It could go shopping for you.
That's going to give people a lot more time back.
Another framework I love is there's a value oriented
consumer and a convenience oriented consumer.
I started a whole company aroundthe convenience oriented
consumer. And what I'll tell you is people
will pay a lot to save time. It's split evenly amongst the
(22:20):
population, but the more affluent tend to be in the
convenience oriented bucket. So it's 80% of the wallet out
there. They value their time more than
things and money. And so these agents are going to
have a really big impact on our lives because you can spend more
time with your kids, you can go golf or play pickle, whatever it
is you want to do. You will have more control of
your life because these agents will go book your travel shop
(22:42):
for you, manage your task list. I think there's, I'm fascinated
by the service economy. It's super analog, but now these
digital agents are going to be pushing into that and they're
going to be managing projects for you.
Like right now we had a water leak in my house and we got 10
people in there spackling and drywalling and all that kind of
jazz. Eventually an agent will project
manage that kind of stuff for you.
So I think that's what it's going to be like.
(23:03):
I think then when we get some robotics going, I do think AI,
all indications are it's going to, and I think this is more
like 3 or 4 years out, it's going to really juice self
driving cars. So things will change more in
the physical world in that way. But I'm more dialed in on this
kind of, I do think the digital stuff we do is going to
accelerate very dramatically. That'll be the fastest to go.
(23:25):
So one of the leaders we interviewed for our work with
the AICPA, the American Institute for Charter
Professional Accountants is mentioned to me or brought up
the question of so they're in the software space building
agents. And he said one of the
challenges I believe people haven't really thought about is
for the CEO or executive in a company.
(23:47):
How do you think and manage a company where you will have a
million agents working for you? How does the world of
organizational design and management change once we have
access to an abundance of these agents which will do things for
us on our behalf? It's really weird because we are
pretty deep into this in my company and we're only like, I'm
(24:10):
not a, we're learning it as we go.
So this is just hot off the presses.
When's your when this is going to be released?
In about a week or so. Okay, all right, two weeks.
I'll give you breaking news, butby the time this comes out,
we'll announce it. So we built our infrastructure
all based on agents, and then wewanted to see how it scaled up.
So we threw 1000 skews at it andit just adjusted it instantly.
And then we're like, all right, 10,000.
(24:31):
So then we just got to the pointwhere we just gave it a million
skews and it just processed it very relatively quickly.
So we built this pipeline of agents and the way we did it is
we're in some ways we're replacing a function.
Most of these agents are doing things a human could do.
OK, so then what you need to do is just give it the instructions
to go do that and the tools to achieve it.
(24:52):
And then what we found is initially it was very flat and
then we had to give it some structure and we basically did
it as if we ended up using just typical organizational
management things. So we have a, we have a super
boss. I should use manager.
So if that's a southeastern, we call him the boss, but we have a
super manager, the CEO agent andit's overseeing this whole
(25:13):
pipeline we have. And then the pipeline is divided
into let's say 10 different things.
The pipeline is data excuse comein and get processed and come
out the end and there's 10 specialized different agent
teams in there, sometimes with different specialization, but
they're individual contributors.And then each part of the
pipeline has a manager. And then the front of the
(25:34):
pipeline actually has another intermediate layer.
And then so we basically built an org chart and then we talked
to the boss and he tells the team what to do.
And they actually have pretty defined roles and an org
structure. And we almost you, when I talk
to the developers, they talk as if they're humans, not in a
weird way, but they're like, this guy's going to go do that,
(25:55):
this guy, but he doesn't do thisbecause that guy's going to and
they have names for him. They're like pipeline Agent 1 is
going to do this. And so I can't sure.
The answer to your question, to the best of my knowledge, is you
manage them. Like when you say it, what do
you do with a million agents? Everyone's Oh my God, they're
going to overwhelm me. But what you're going to do is
you're going to the manager agent, though a human manager
agent, a human can manage 10 people is a common thing.
(26:17):
But the CEO of NVIDIA, everyone reports to him or he has like an
insane number of reports. So it's going to look more like
that. And it's going to be the manager
can almost manage infinitely individual contributors, but you
want them to have a very specific role.
Have you ever been in a company where you're spending all this
(26:37):
time on job description? You're like, this is ridiculous.
And it's in the human world, it's usually because an employee
doesn't want to do something right.
So it's not the tail is wagging the dogs.
They're like wanting, is this inmy job description?
Because I want to be able to tell Sally I don't want to do
it. So it ends up being like a total
waste of exercise. But in this world, the job
description is important becauseif you can nail it and
(26:59):
individual contributors, there could be 100 of them working for
one boss. And then the boss layer is
pretty thin. So you've got this very thin
middle management layer of agents that are very
specifically doing something. And then you have a lot of ICS.
So it's in a million, it's probably something like 990,000
ICS and then some layer of management in there.
(27:22):
And then you'll have a as CEO, you'll have a go to market agent
that is the CMO and you talk to them just like you would ACML
and then they have a team that does stuff.
That is so interesting because it, you know, what came up for
me is for, I don't know, like 1020 years now, we keep telling
people that you have to move away from the org chart lines
(27:44):
and boxes and the idea that you can like linearly manage an
organization towards something which is much more fluid, much
more dynamic and agile. And this here sounds to me that
at least on the agent side, you get to like fairly rigid lines
and boxes because they are machines you need to like and
the machine intelligence. Yeah.
(28:05):
And it ties into a part we don'thave in the human world where
there's usually a little bit of DevOps in there too, right.
So the boss agent is saying, allright, I've got this.
You have tasked me with this task.
I'm going to spin up this many ICS and I'm going to know to
stop because you've given me some constraint.
Maybe it's dollars or a compute or something where the IC
(28:26):
doesn't need to know that. So this is where the job
description becomes important. The way we the way we've done it
is we just built the ICS and we managed them.
And then we're like, we did it enough where we're like, OK, we
could just, we know the and the job description somewhat becomes
almost a prompt. It ends up being code, but you
could almost write it in a prompt where you're like, you're
the boss agent and you give it all this context and I want you
(28:48):
to spin up a new, but what you're doing is what we're doing
is we're backing out of it, if that makes sense.
So we started with ICS and the engineering team was managing
them. And then they're like, we could
write an agent to do what we're doing because I'm babysitting
engineer one is babysitting the front of this pipeline.
And then it gets to the point where you get and you write a
prompt and then you're like, nowyou're managing that thing and
you're like, I could actually, man, I could write an agent for
(29:10):
that. So that's like how we've backed
into it, if that makes sense. And it ends up being a very,
very clean job description. It's so fascinating, truly a
different. It's funny because it's in some
ways it goes back to the beginnings of the industrial
revolution and like. It is a fortification of yes.
(29:31):
Exactly. We're going back to the conveyor
belt. Yeah, yeah.
So to use your CPA thing, there's at the end of the day, I
don't know as much about CPA world.
I've had big finance teams. At the end of the day, there's
data entry people, right, Because you've got to take this
world of data receipts and digital things and Docusigns and
all this noise and you've got tofigure out how to entry.
(29:51):
So that's clearly like an IC level data entry agent, but you
don't want that agent to be thinking about Rev rec or
something. Then there's going to be like
something, maybe there's like a bookkeeper agent and the
bookkeeper thinks about a lot layer up like charts of accounts
and things and starts to see, Ohmy gosh, there's so many
expenses in going that the data entry people are putting into
(30:12):
miscellaneous. I'm going to look in there.
Oh, I need to actually have a new entry because I don't know,
our company is consuming a lot of AWS and I need to put that
into a new so on. So you could actually whiteboard
out this org chart, I think if you have to take the human part
out of it, which is very hard atfirst for me, the engineers,
they're very good at it because they deal with computers and
(30:33):
they did it by backing out of the role, if that makes sense.
So that's like a weird learning that we've had as humans do it
first and figure out how to put themselves out of that job and
then go layer up. And so far we're still the Super
boss. But at some point, like when do
we get replaced is kind of like the because some everyone
crosses this moment in this kindof singularity ish where you're
like, I'm sitting here consumingall this information.
(30:56):
We talked about this and then I'm building frameworks and but
in a way, am I pattern recognition rising?
Could I train an LLM to do that?So is your vision.
Could I back out on my job? Yeah, absolutely.
So that's like where your head starts to explode.
Right. But you don't start there when
you're talking to people outside.
That's a Pascal, Scott, kind of.Yeah, that's fair.
(31:17):
But I'd be curious, in the midterm, would you expect
companies to become leaner as asin physical staff people?
Yeah, absolutely. I'm a big believer in this.
You're starting to see it. It's a little unfair.
So you start to see these chartswhere I think Carta has one they
talked about on the All In podcast.
This is why you got to listen tothese really like big signal
(31:39):
kind of things out there. And if you dig into that and
they just showed AI companies, they get to 5 million ARR twice
as fast and with half as many people or something like that.
If you dig into that data, it's because they're looking at
cursor and windsurf. It's like there's two data
points in there. So cursor is 300 million an AR
and has 20 people or something ridiculous.
I don't think it's going to looklike that.
I think that's a very specific use case.
(32:02):
I do think in our world, let's use Channel Advisor.
When you start a SAS business, you have maybe you're losing
money because of the way SAS works.
You get paid in the future. So each employee cost you like a
hundred, 150 and they're making fifty.
I think revenue per employee is a good metric.
So you're in a SAS business, you're 0 to 5 million.
You got 50K per employee. If you're lucky, then you get to
(32:23):
a kind of 100K and then you could get to really good ones,
get to 202 fifty I think. I think that we'll be all that's
going to shift down. So I think AAI native company
will get to to 150, zero to fiveand then two 53105 to 10.
And then I think you got a shot to get to 305 hundred.
(32:44):
Fascinating part. Of me is an entrepreneur.
The interesting question is should you take that extra and
invest more? Because if you see enough
opportunity, you should probablynot harvest that hard.
Like I would argue cursor may bedoing the wrong thing.
So I don't know. Time will tell.
If they had another 50 people, could they, could they have a
(33:05):
dev OPS product? Could they have a, should they
do some cloud hosting or something?
I think, yeah. Each entrepreneur has to
navigate that complex matrix. Yeah, reminds me to my early
days at eBay, way back in whatever in 1999 where we were
crushing it in Germany and made,I don't know, like $0.70 on
every or made a dollar on every,whatever $0.60 we spent or
(33:28):
something. And we took all of the money and
poured it back into the marketplace just because that's
how you get to dominance, right?That's how.
And it worked. Today's eBay would not do them.
They would cash out every penny and then they'd be shocked when
it stopped growing. Totally.
Hey, I have a last question for you as a bit of an oddball, but
I would love to hear your opinion on it because you're an
(33:50):
expert and I see you and our dear friend Jason talk about
this on LinkedIn. I follow the two of you quite a
bit in this kind of liminal space, this in between moving
from an old world order to a newworld order, the old world order
being Google search being the dominant way for us to refi and
buy. Is this changing?
(34:11):
Is it the end of Google? Part of the fascinating part of
the liminal space is the innovator's dilemma.
I'm sure you actually know that book better than I do.
And I never, I read the book andit was like academic.
And then I went to meet the CEO of Borders at Channelvisor and
he was telling me how the Internet was a fad and people
love coming into the coffee and they would gladly pay more for
books because they could have a cup of coffee.
(34:32):
And I was like, wow, this is innovator's dilemma right here.
This guy is the living embodiment.
So I think it's going to, I don't know the answer, but I
think it's going to be fascinating.
You have two companies, Google, maybe 3, Google, Meta and
Amazon, who have basically explicitly said we're going to
beat the innovator's dilemma andit's going to be interesting to
see if they do or not. Amazon did one time with Kindle,
(34:55):
coming out with Kindle and self eating.
The book market was genius at the time.
It seems crazy, but Netflix is another one that has had a
history of doing this. So maybe we're in this new
generation clearly defining the innovator's dilemma.
People can get out of it. So Google's doing everything
they can. I don't think they're going to
because their innovator's dilemma is they've got the best
business model in the world where money just rains out of
(35:16):
the sky. It's going to be very hard to
get out of that individual enough.
But yeah, the they're doing someof the right things.
But is that enough? I don't know.
What I can tell you is personally, I've switched
perplexity and ChatGPT because the user experience is so much
better. You get to the answer faster,
but now you start to use some ofthese deep research models, you
get to not only the answer, but you get to the next 12 answers.
(35:38):
It's pretty crazy. And this is what makes it an
exciting time to be in where we are.
I found this interesting. I saw you posting on LinkedIn
that you switched essentially most of nearly all of your
searches to perplexity, which was roughly the same time.
I did the same thing. And I don't know how you're
dealing with this at the moment,but I found myself using ChatGPT
(35:59):
with the search feature significantly more in perplexity
these days because the results in ChatGPT I find actually are
better presented, typically tendto be better.
Yeah. So it's weird and interesting to
see how fast we switched from like we go from Google search to
perplexity and now we're just already like on to the next one.
Yeah, that's the liminal space. Yeah.
(36:21):
I also have a softness in my heart for the underdog and I
really have been the CEO of the perplexities interviews and
things. So I haven't switched yet, but
it's maybe just a little bit of loyalty there.
I agree with you. The ChatGPT has surpassed it at
this point. Scott this was an absolutely
phenomenal conversation. I loved literally every single
bit of it. I learned so much.
(36:41):
We will put links into the show note for people to follow the
exciting annulments which will come out of your company I'm
100% sure knowing your track record.
Also talking about good podcaststo follow, we'll put a link to
the Jason and Scott show which is an absolute must listen to,
particularly if you're in the e-commerce space.
But also even if you're not in e-commerce and just want to hear
(37:04):
2 very smart people talk about the future and banter quite a
bit, which I love. Highly recommend that.
Jason's the grumpy, the guy in the liminal space.
I hate change. So it ends up he's the guys on
The Muppets, the grumpy guys on the balcony.
There you go. He's the cremation of the two of
us. Scott, thank you so much.
This was amazing. Thanks Pascal.
Thanks for introducing me to theliminal space.
I'm going to go take a deep diveand swim around in it for the
(37:27):
rest of the day. Perfect.
Thanks. Disrupt disruption.
You're the boss agent and you give it all this context and I
want you to spin up a new, but what you're doing is what we're
doing is we're backing out of it, if that makes sense.
So we started with ICS and the engineering team was managing
them. And then they're like, we could
write an agent to do what we're doing because I'm babysitting
engineer one is babysitting the front of this pipeline.
And then it gets to the point where you get and you write a
(00:20):
prompt and then you're like, nowyou're managing that thing and
you're like, I could actually, man, I could write an agent for
that. So that's like how we've backed
into it. And it ends up being a very,
very clean job description. So fascinating.
It's funny because it's in some ways it goes back to the
beginnings of the Industrial revolution and.
(00:42):
It is a very fortification of yes, exactly.
We're going back to the conveyorbelt.
Yeah. Disruption.
Hey everybody, welcome back to Disrupt Disruption, where we are
going to explore what the liminal space holds for us, the
(01:06):
weird twilight zone between the old world order and the new one,
which hasn't quite emerged yet. I'm Pascal Vinette, your host,
and today I'm incredibly excitedto have a very dear friend and a
very special guest with us. Scott Wingo is a serial
entrepreneur with an exceptionaltrack record of building and
(01:26):
exiting successful tech companies.
As the CEO and cofounder of Refiby dot AI, he's leveraging AI
technologies to revolutionize ecommerce experiences.
Scott previously founded and LEDa series of successful startups,
Stingray Software, which he soldto Rogues Wave,
auctionrover.com, which was acquired by Yahoo, and Channel
(01:48):
Advisor, which he grew from startup to IPO on the New York
Stock Exchange. He also Co founded Spiffy, an on
demand car care service operating across multiple
cities. Beyond building companies, Scott
serves as a general partner of the Triangle Tweener Fund,
investing in early stage startups in North Carolina.
He Co hosts the popular Jason and Scott Show podcast, which I
(02:13):
highly recommend you check out, focused on e-commerce trends and
was a Forbes Technology Council contributor.
And probably more important thanall of this, both Jane, my wife
and I used to work for Scott when we were at Channel Advisor.
With that being said, let's dig into the conversation.
Scott, it's amazing to have you on this podcast.
(02:36):
We go way back and as a true pioneer in the e-commerce space,
I'd be incredibly curious on howyou think about the liminal
space. You have seen it play out many
times in e-commerce and the broader Internet context.
And maybe to get us kicked off, like, how do you think about the
period of time you're currently finding ourselves in with the
(02:59):
advent of AI changing in like even geopolitics being a big
topic. But I'd be curious to hear in
from you. Like where are we in this moment
in time? Yeah, it's a crazy moment in
time. Thanks for having me.
It's good. Always good to see.
See you, Pascal. Any other either of the
phonettes are always amazing people.
The I didn't know what liminal was, and I'm from South Carolina
(03:20):
so that's a big complicated wordwith lots of weird consonants
and vowels in it. So thank you for explaining to
me what it is. I would say I've been at this a
while and because I have a math orientation, it feels like it's
a great time to be alive becausewe're on this logarithmic curve.
And so when I was growing up in this small town in South
(03:43):
Carolina, I've always been a misfit toy I wear that's a
positive to me. Everyone in my town either was
like dirt poor or they worked for a nuclear facility in my
city. So it's kind of like The
Simpsons, if you will. And then my dad was an
entrepreneur and we had a vax, amainframe in my house.
It's probably the only computer in my town and it was a
mainframe. So it's very weird, right?
(04:04):
So to grow up with computer. So I've started, I've lived with
a mainframe in my house and thenI got a desktop computer and
I've been through this whole curve.
And you know what's really wild is from mainframe to PC was a
pretty big chunk of time. And then if we Fast forward, you
and I have lived through desktopto mobile.
We've even gone from build your own infrastructure to cloud,
(04:26):
from desktop to mobile. And then what I would say is the
logarithmic IT, we're on that log curve.
So now things are just happeningfaster and faster.
And the evidence we see of this is there's all these charts that
show different consumer things and how long they took to get to
100 million. Color TV took 50 years and the
microwave took 25. And now Chachi BT I think is the
fastest and you see the lines and most of them have this kind
(04:48):
of curve and now they're starting to go vertical.
So as a technologist, exciting, it's exciting and entrepreneur,
it's exciting because these things we're, we always tend to
embrace them and ride those curves because on the other side
is opportunity. And I think it's this AI 1 is
(05:10):
pretty crazy. And I've gone through my own
little 7 steps of doubt. And then the Gentex side is
where it can actually do stuff is really bonkers.
And, and that's what I'm workingon because I got, so I got into
it and I'm not one to jump on these things.
I didn't do anything in the crypto space, for example.
Not that I don't like crypto, I could just never really get my
head around. I couldn't get as excited about
(05:31):
it changing the world as what we're seeing with the AI stuff.
That's a lot. Let me pause and see if you want
to dig into any of that. Absolutely.
Be curious. So you're talking about the
increased rate of change, which I think we undoubtedly see when
we talk about this quite a bit. The flip side of that is in the
(05:51):
space of organizations, there's something called Martex law,
which basically says that organizations change
logarithmically, but mostly linearly.
So it's like widening. Yeah, widening gap between
what's possible and what you canactually absorb in the
organization. The context would be to say as a
leader, you need to be very careful to take and choose the
(06:14):
change you want to implement in your organization because
otherwise your organization justcan't absorb it worked.
You've been on the set up side, you've brought a company onto
the New York Stock Exchange and you have worked with very large
incumbent legacy organizations as clients.
I'd be curious to hear from you.How do you think?
(06:34):
Like what do you advise your clients on the change now?
Like with your new venture, you will talk to legacy businesses
and you will want them to change, but they can only absorb
so much change. Yeah.
And I have a framework I developed at Channeladvisor,
which is the company that took public.
And so we're 800 people, we're public and we're in the
(06:54):
e-commerce space and it's just changing.
You and I, you were very early on this kind of Shopify trend if
we will. So a SAS based relatively
inexpensive. You came to me with that idea
and I was like Pascal, that's crazy and you're right that you
were. So what you have to do is you
have to because you have a largeorganization.
We were like 800 a thousand people.
To your point, you can't just say, all right, everyone, we're
(07:16):
pivoting to this thing. But I hate binaries as an
entrepreneur, I hate these decisions of do nothing or do
everything. And so I always try to split the
binary. And the way to split this binary
is to have more of a allocation matrix.
And I like 70 2010. I learned this from the Google
10% time where they allow engineers to, if you talk to
(07:36):
anyone at Google, they don't do this anymore and it's just PR.
But anyway, there it was a good story.
They would allow engineers to take 10% of their time and work
on out their projects. And that's where the lore is.
That's where way MO came from. And some of these ideas, again,
I'm not really exactly sure that's true, but anyway, so this
allocation matrix is good because you basically say, all
right, you've got your core business and just use the
(07:57):
channelvisor example. We're helping people sell on
eBay and Amazon and we can't take our eye off of that, but
we're big enough that we should take 20% of our time and do
things. I kind of use a, a, a, a dart
board, if you will. So a target.
So 70% is like right in the red zone is where we're going to put
our effort. We're not going to stop
maintaining and customers are there.
Then we're going to 1 ring out and that's the 20%.
(08:19):
So that may be in this context of channelvisor, that may be OK.
There's this marketplace, this company in China called Alibaba
and they've got this team all let's today at 0.
But we think we're going to spend some 20% time on that.
So you know, and in an organization of 800, that's like
maybe 100 people kind of working.
It's actually lower, oddly enough, because usually it's the
(08:39):
engineering piece. So 800 people, we're not going
to really put sales on it, but we have 100 engineers now.
We have 20 working on something a little bit out there.
And they don't only do one, but they do 3 or 4 projects.
It can't be this infinite list either.
So you pick has a few be on one hand under 5 things in the 20%
time. And then the 10% time is where
you're like, hey, there's this AI thing we should have of our
(09:03):
150 engineers, we probably need 10:00-ish that kind of are
tracking this thing and doing some experimenting around it
because it right now we're not sure, but it could go.
And then over time, some of the 10% things don't work out and
they die. So there's like a little
incubatory kind of thing. It's almost like managing a
portfolio of startups, if you will.
And some of them get product market fit that they pull into
(09:25):
the 20% piece. So you put more Reese's on them
and then they make it into core.So I like that framework and
that's a framework I then now because I've run it, I espouse
that when I talk to you, any retail, any clients of any kind.
And I did the same thing at a totally different company called
Spiffy and we had, I think I'm getting better at it because the
(09:45):
most of our ideas actually gained pretty good fruit and
moved from the 10 to 20 70. I love the non binary version of
this because typically what you hear is I don't know like the
20% rule which means like the hard boundary between 80 and 20.
They also seem to be hard or we pretend there are hard
boundaries. I love the gradient in this.
(10:06):
The thing you hear people talk about lots these days is this
idea from whatever you want to use as a framing, but like from
AI Infuse to AI native and everybody talks about what does
the accounting firm look like inthe future, which is AI native?
What does the e-commerce firm look like, which is AI native?
What does the apparel business look like, which is AI native?
(10:30):
I'd be curious, be curious, how do you think about AI native and
what does it even look like? Because I feel a lot of people
say AI native but they have no idea what it actually means in
real terms. Yeah, and so let's say we're
we're in this mythical thousand person company and again, this
binary thinking. And I run across this a lot
(10:50):
where, you know, I meet JP Morgan and Bank of America and
all these people and none of them, they're not tracking AI
because they have a stiff policyat work.
They have some internal AI that kind of stinks.
It's llama one on some bad rag database or something.
It's like ChatGPT, the 1st 2020 Chat 222 ChatGPT.
And that's I feel bad for them because there needs to be
(11:12):
someone in this organization that is thinking these things
through. So what I would do is I would
take 10% and I would say, all right, you're AI native.
What does that mean? Whenever you start building
something in my world of software, first of all, the
engineers are going to start with looking at the tool set.
You don't do it. A lot of this again, you don't
want the tail to wag the dog. The idea is let's try to be AI
(11:33):
1st and say, all right, let's doan analysis.
What are the tools out there? There's wins surf and cursor and
the copilot knowledge has and inthis part of the org you're
going to give them infinite flexibility, maybe even each and
because we're in this, this is 10% innovation mode.
Maybe one engineer does cursor and the other does windsurf and
another in fact, variety is actually good.
(11:54):
Whereas in the rest of your organization variety is bad.
And this like little 10% incubator thing, you want to
encourage it because you get more learnings from that.
So that's phase one. But then as you get into go to
market, so every function has anAI native kind of aspect to it.
Engineering is the easiest one because that's where there's
like the most uptake. So that that one, you probably
don't have any questions on that.
But then the second one is now what are you going to do is go
(12:16):
to market. And this is where you want to do
a lot of experimentation. But again, you don't want the
tail to wag the dog. I'm actually a little counter
the popular opinion on this one because there's all, I run into
all these startups that are using these AIBDRS and this kind
of thing. And I think, and I'm the
recipient of a lot of that. I'm sure you, you get quite a
bit as well and you can tell every time.
(12:37):
And I, I think it's going to be interesting because that may be
where an area where certain parts of go to market, the part
that doesn't touch the customer,I think can be AI native.
So you're using these Otter meeting notes and there's a huge
value in all those things. The there's no, back in the day,
we would spend half our time putting stuff in Salesforce that
can go away, all those things. That's where I think you'd be AI
(12:58):
native. But when it comes to
authenticity and talking to the customer, I think you you say,
all right, we stopped there. I think the value of human to
human interaction is going to goway up in a world where there's
all this AI noise. I don't know if I'm answering
your question, but every function I think has to evaluate
this. But you can't overdo it a lot of
(13:19):
times. I remember the first time we
implemented Scrum at my company and the Scrum Master was just
running around insane because wedidn't have the right Scrum
master. And I forget what the other one
is. There's what I would call a
product manager. They split it into two roles.
We had six products. I was like, I'm going to have to
hire like 50 freaking people forScrum.
We're going to do it this way. And they're like, you violated
all the rules of Scrum. And I was like, OK, then they
(13:41):
quit on the spy. You can't be such a purist that
you make the wrong decision for the business.
And you have to infuse these things into your DNA in the way
that is the best thing for the customer and for your
organization. And then again, if you're in
this 10% model, I like a lot of variation in there so you can
see what's working. Yeah, love it.
Curious in this context, how do you, it's probably a two-part
(14:02):
question. One is as a leader, how do you
stay on top of the massive amount of change we you talked
about like this idea or this exponential curve kind of goes
vertical at the moment in terms of adoption and developments,
etcetera. Maybe even you personally, how
do you stay on top of all of that stuff?
You're in the AI space where, you know, I do a decent amount
(14:23):
of reading in the AI space and Ifeel like, I feel it's dizzying
as in new model comes out like every three days and it like
tops the old model and like all this kind of stuff.
And this is the second part of the question.
How do you deal with that in terms of making sure your
organization doesn't go crazy, your people don't go crazy, and
at the same time they stay up todate?
(14:45):
Yeah, as the person that sets the strategy, I've never really
had a chief strategy officer. I always want to do the
strategy. I've never understood chief of
staff or this may be controversial, this may be a hot
take. I've never understood chief of
staff or chief strategy officer because that's basically my job.
And at the end of the day, if the founder isn't doing that, I
have questions about the whole thing anyway.
(15:05):
So my job is to ingest all this information, process it.
And my immediate thing I work onis a mental framework.
I really like this idea of I need to framework.
So to your example, you've got all this information coming in
and my current framework is all right, we've got the frontier
models. And I think we all agree now
that those are commoditizing. There's going to be some
(15:28):
interesting things going on in there and we'll keep track of
it, like the new llama for and this kind of thing.
But that that's in this quadrantof it's stabilizing and we know
what's going to happen. The the cost per token is going
to continue to drift down. And then the interesting areas
are for us is some of the because we're focused on this
one area where like hyper, any piece of information that comes
(15:50):
about this intersection of e-commerce, retail and agentic
where agents are shopping for you.
So open AI, this is very timely because this the news is
happening like logarithmically, right?
I had this idea in October, November of last year and now
every day a new news release is coming out.
When I started Channelvisor, I had an idea that there'd be a
(16:11):
lot of marketplaces and we supported eBay and it took seven
years for Amazon to come out andthen another five years for the
next marketplace. And then suddenly they, so it's
that logarithmic curve. It's crazy.
But then now I have a framework for that and it's actually in
our name and it's kind of which of these things are helping the
consumer research, find things and buy and that I will sort all
(16:31):
that and then share it into the org and that framework.
And then a lot of times what I'll do is I'll send a signal
that says this is just information.
So as a, the other thing I've learned as a CEO, what you say
and distribute, sometimes I just, I like to make sure
everyone sees everything. We're overwhelmed to buy it.
So I'll now scale it and say this is just an interesting
news. No, just doesn't really change
anything for us. It's but it validates our idea
(16:52):
here. Open AI is coming out with
something in this space. I'm like, this is pretty
important. We need to think a little bit
more about this or do that. For example, Open AI just came
out with this ability. The retailer can they have these
cards in the chat and the retailer can send a different
data feed to have that. So that's that impacts us.
So that one, I'm like in my matrix of consideration, that's
one I need to share into the organd make sure that everyone
(17:14):
understands. So I tried to build a framework
on what it is that's going on and the ultimate framework.
Is either this going to change our strategy?
If no, then share it or don't share it if it's high quality.
At the Super top of this, the framework is where is the best
signal to noise ratio. And I'm sure you spend a lot of
time on this too. So as people that consume a lot
of information that's and it's been interesting over time, it's
(17:37):
really changed. We used to read magazines and
books and then we've had podcasts, but they weren't as
good as they are now. And then we went through a phase
of you could social media, you get a lot of info, but now the
noise ratio there is pretty high.
You have to figure out where to go get the best information to,
which is another challenge in today's world.
How do you stay up to date like very concretely like what is
(17:58):
your like information diet, so to say?
And also I'd be curious just to as a general of like
observation, are you spending more time now or significantly
more time now keeping up to datethan you did 10 years ago, 15
years ago when you are in the Channel Advisor world, for
example? Yeah, I would say it's the same
time. The density is like 10X, OK.
(18:21):
The throughput of the amount andquality and throughput of
information because reading a book is interesting, but it just
is not a fast way to consume information.
I would say podcasts are definitely where I get a lot of
information. The other, the other part of the
signal, the noise. So a really good example is the
All In podcast. And this is a controversial 1
because it has politics, but it matters because I'm a venture
(18:44):
backed. My side hustle is, I also have
AVC fund, but I'm also a venturebacked startup.
If you're any venture backed startup, you need to get out of
that mindset that oh, this is political or whatever.
Because every VC listens to thatpodcast and they form their
opinions from there. So if I don't listen to that,
I'm going to be behind the curve.
And it also has a very high signal to noise ratio.
It is very, I don't care about politics, that part's boring.
(19:05):
But when they don't talk about politics, this concept that
Chamath talks about of 8090 and he said that one weekend and
then within three days or founder mode or all this stuff
gets amplified there. So in my framework, that's a
must listen thing. And then I have other ones like
Lex Friedman tends to have really good interviews that have
a big impact and so on and so forth.
(19:27):
I love my favorite piece of content over decades of
stratechory with Ben Thompson that I highly if anyone takes
any one thing from this podcast,subscribe to that is worth every
penny. It used to be a long read, which
was hard to get to every day, but the podcast version is what
I've pivoted to. I really enjoyed that.
I was just earlier listening on my way into work.
I was listening to an interview he did with Mark Zuckerberg
about llama stuff, and it was excellent.
(19:48):
Yeah, I find that I get a lot more information out of the
podcast, and there's very specific ones you have to
navigate to to get really good content.
And of course, your podcast is the best.
All in podcast is number 1 is number 2.
No, fair enough. Thank you for the shout out
here. Let me increase the signal to
noise ratio on this podcast and double click on the area you are
(20:11):
currently working on and spend like most of your brain power on
which is AI, particularly agentic AI in this e-commerce.
Broadly speaking e-commerce space genetic AI very clearly is
like the topic at the moment. So we have pivoted from just AI
to everybody needs to put a genetic in their in their
frameworks and names. I'd be curious to hear from your
(20:33):
perspective, where are we on this journey?
Are you feeling we're like it's good enough already?
We're like seeing actual real productive use cases.
It's very early. How do you see this?
And then related to that, as we're talking about this like
weird in between space, the Twilight Zone is agentic, this
kind of piece which unlocks thisTwilight Zone, which gets us
(20:53):
from like the old way of doing things to the new way.
I think maybe robotics because you worked for the singularity,
so you can go way deeper on thisthan I can.
I'm just trying to consume wherewe are on the curve and you have
the ability to look further out than I do.
So in my little view, let's lookat a 18, two years, 18 months, 2
years. I do think Agentics going to
unlock a lot of people that a lot of you know, the Aunt Jenny,
(21:16):
normal people that aren't as deep in this.
The chat thing is just not, it doesn't do anything right.
It's a way to get information and whatnot.
But at the end of the day, it's not like accomplishing anything
other than that. And what Agentic does is it
gives you have a little LM kind of a little brain and it has,
it's consumed all the Internet. So it's very wide and deep.
So it's a square of knowledge. And in some ways it's almost
(21:38):
like too wide and too deep. And then what most of them do is
you take, you want to narrow that to solve a specific
problem. So you use a retrieval augmented
generative database. And, and my layman's way of
explaining that is you take thishuge brain that can see
everything and you focus it through a lens.
And you say, I really only want you to think about this.
And by the way, here's some moreinformation about this that you
may not have because it's not out on the web.
(21:59):
And then you give it some tools to go do stuff.
And that's where it gets interesting in, in my little
sliver. It could go shopping for you.
That's going to give people a lot more time back.
Another framework I love is there's a value oriented
consumer and a convenience oriented consumer.
I started a whole company aroundthe convenience oriented
consumer. And what I'll tell you is people
will pay a lot to save time. It's split evenly amongst the
(22:20):
population, but the more affluent tend to be in the
convenience oriented bucket. So it's 80% of the wallet out
there. They value their time more than
things and money. And so these agents are going to
have a really big impact on our lives because you can spend more
time with your kids, you can go golf or play pickle, whatever it
is you want to do. You will have more control of
your life because these agents will go book your travel shop
(22:42):
for you, manage your task list. I think there's, I'm fascinated
by the service economy. It's super analog, but now these
digital agents are going to be pushing into that and they're
going to be managing projects for you.
Like right now we had a water leak in my house and we got 10
people in there spackling and drywalling and all that kind of
jazz. Eventually an agent will project
manage that kind of stuff for you.
So I think that's what it's going to be like.
(23:03):
I think then when we get some robotics going, I do think AI,
all indications are it's going to, and I think this is more
like 3 or 4 years out, it's going to really juice self
driving cars. So things will change more in
the physical world in that way. But I'm more dialed in on this
kind of, I do think the digital stuff we do is going to
accelerate very dramatically. That'll be the fastest to go.
(23:25):
So one of the leaders we interviewed for our work with
the AICPA, the American Institute for Charter
Professional Accountants is mentioned to me or brought up
the question of so they're in the software space building
agents. And he said one of the
challenges I believe people haven't really thought about is
for the CEO or executive in a company.
(23:47):
How do you think and manage a company where you will have a
million agents working for you? How does the world of
organizational design and management change once we have
access to an abundance of these agents which will do things for
us on our behalf? It's really weird because we are
pretty deep into this in my company and we're only like, I'm
(24:10):
not a, we're learning it as we go.
So this is just hot off the presses.
When's your when this is going to be released?
In about a week or so. Okay, all right, two weeks.
I'll give you breaking news, butby the time this comes out,
we'll announce it. So we built our infrastructure
all based on agents, and then wewanted to see how it scaled up.
So we threw 1000 skews at it andit just adjusted it instantly.
And then we're like, all right, 10,000.
(24:31):
So then we just got to the pointwhere we just gave it a million
skews and it just processed it very relatively quickly.
So we built this pipeline of agents and the way we did it is
we're in some ways we're replacing a function.
Most of these agents are doing things a human could do.
OK, so then what you need to do is just give it the instructions
to go do that and the tools to achieve it.
(24:52):
And then what we found is initially it was very flat and
then we had to give it some structure and we basically did
it as if we ended up using just typical organizational
management things. So we have a, we have a super
boss. I should use manager.
So if that's a southeastern, we call him the boss, but we have a
super manager, the CEO agent andit's overseeing this whole
(25:13):
pipeline we have. And then the pipeline is divided
into let's say 10 different things.
The pipeline is data excuse comein and get processed and come
out the end and there's 10 specialized different agent
teams in there, sometimes with different specialization, but
they're individual contributors.And then each part of the
pipeline has a manager. And then the front of the
(25:34):
pipeline actually has another intermediate layer.
And then so we basically built an org chart and then we talked
to the boss and he tells the team what to do.
And they actually have pretty defined roles and an org
structure. And we almost you, when I talk
to the developers, they talk as if they're humans, not in a
weird way, but they're like, this guy's going to go do that,
(25:55):
this guy, but he doesn't do thisbecause that guy's going to and
they have names for him. They're like pipeline Agent 1 is
going to do this. And so I can't sure.
The answer to your question, to the best of my knowledge, is you
manage them. Like when you say it, what do
you do with a million agents? Everyone's Oh my God, they're
going to overwhelm me. But what you're going to do is
you're going to the manager agent, though a human manager
agent, a human can manage 10 people is a common thing.
(26:17):
But the CEO of NVIDIA, everyone reports to him or he has like an
insane number of reports. So it's going to look more like
that. And it's going to be the manager
can almost manage infinitely individual contributors, but you
want them to have a very specific role.
Have you ever been in a company where you're spending all this
(26:37):
time on job description? You're like, this is ridiculous.
And it's in the human world, it's usually because an employee
doesn't want to do something right.
So it's not the tail is wagging the dogs.
They're like wanting, is this inmy job description?
Because I want to be able to tell Sally I don't want to do
it. So it ends up being like a total
waste of exercise. But in this world, the job
description is important becauseif you can nail it and
(26:59):
individual contributors, there could be 100 of them working for
one boss. And then the boss layer is
pretty thin. So you've got this very thin
middle management layer of agents that are very
specifically doing something. And then you have a lot of ICS.
So it's in a million, it's probably something like 990,000
ICS and then some layer of management in there.
(27:22):
And then you'll have a as CEO, you'll have a go to market agent
that is the CMO and you talk to them just like you would ACML
and then they have a team that does stuff.
That is so interesting because it, you know, what came up for
me is for, I don't know, like 1020 years now, we keep telling
people that you have to move away from the org chart lines
(27:44):
and boxes and the idea that you can like linearly manage an
organization towards something which is much more fluid, much
more dynamic and agile. And this here sounds to me that
at least on the agent side, you get to like fairly rigid lines
and boxes because they are machines you need to like and
the machine intelligence. Yeah.
(28:05):
And it ties into a part we don'thave in the human world where
there's usually a little bit of DevOps in there too, right.
So the boss agent is saying, allright, I've got this.
You have tasked me with this task.
I'm going to spin up this many ICS and I'm going to know to
stop because you've given me some constraint.
Maybe it's dollars or a compute or something where the IC
(28:26):
doesn't need to know that. So this is where the job
description becomes important. The way we the way we've done it
is we just built the ICS and we managed them.
And then we're like, we did it enough where we're like, OK, we
could just, we know the and the job description somewhat becomes
almost a prompt. It ends up being code, but you
could almost write it in a prompt where you're like, you're
the boss agent and you give it all this context and I want you
(28:48):
to spin up a new, but what you're doing is what we're doing
is we're backing out of it, if that makes sense.
So we started with ICS and the engineering team was managing
them. And then they're like, we could
write an agent to do what we're doing because I'm babysitting
engineer one is babysitting the front of this pipeline.
And then it gets to the point where you get and you write a
prompt and then you're like, nowyou're managing that thing and
you're like, I could actually, man, I could write an agent for
(29:10):
that. So that's like how we've backed
into it, if that makes sense. And it ends up being a very,
very clean job description. It's so fascinating, truly a
different. It's funny because it's in some
ways it goes back to the beginnings of the industrial
revolution and like. It is a fortification of yes.
(29:31):
Exactly. We're going back to the conveyor
belt. Yeah, yeah.
So to use your CPA thing, there's at the end of the day, I
don't know as much about CPA world.
I've had big finance teams. At the end of the day, there's
data entry people, right, Because you've got to take this
world of data receipts and digital things and Docusigns and
all this noise and you've got tofigure out how to entry.
(29:51):
So that's clearly like an IC level data entry agent, but you
don't want that agent to be thinking about Rev rec or
something. Then there's going to be like
something, maybe there's like a bookkeeper agent and the
bookkeeper thinks about a lot layer up like charts of accounts
and things and starts to see, Ohmy gosh, there's so many
expenses in going that the data entry people are putting into
(30:12):
miscellaneous. I'm going to look in there.
Oh, I need to actually have a new entry because I don't know,
our company is consuming a lot of AWS and I need to put that
into a new so on. So you could actually whiteboard
out this org chart, I think if you have to take the human part
out of it, which is very hard atfirst for me, the engineers,
they're very good at it because they deal with computers and
(30:33):
they did it by backing out of the role, if that makes sense.
So that's like a weird learning that we've had as humans do it
first and figure out how to put themselves out of that job and
then go layer up. And so far we're still the Super
boss. But at some point, like when do
we get replaced is kind of like the because some everyone
crosses this moment in this kindof singularity ish where you're
like, I'm sitting here consumingall this information.
(30:56):
We talked about this and then I'm building frameworks and but
in a way, am I pattern recognition rising?
Could I train an LLM to do that?So is your vision.
Could I back out on my job? Yeah, absolutely.
So that's like where your head starts to explode.
Right. But you don't start there when
you're talking to people outside.
That's a Pascal, Scott, kind of.Yeah, that's fair.
(31:17):
But I'd be curious, in the midterm, would you expect
companies to become leaner as asin physical staff people?
Yeah, absolutely. I'm a big believer in this.
You're starting to see it. It's a little unfair.
So you start to see these chartswhere I think Carta has one they
talked about on the All In podcast.
This is why you got to listen tothese really like big signal
(31:39):
kind of things out there. And if you dig into that and
they just showed AI companies, they get to 5 million ARR twice
as fast and with half as many people or something like that.
If you dig into that data, it's because they're looking at
cursor and windsurf. It's like there's two data
points in there. So cursor is 300 million an AR
and has 20 people or something ridiculous.
I don't think it's going to looklike that.
I think that's a very specific use case.
(32:02):
I do think in our world, let's use Channel Advisor.
When you start a SAS business, you have maybe you're losing
money because of the way SAS works.
You get paid in the future. So each employee cost you like a
hundred, 150 and they're making fifty.
I think revenue per employee is a good metric.
So you're in a SAS business, you're 0 to 5 million.
You got 50K per employee. If you're lucky, then you get to
(32:23):
a kind of 100K and then you could get to really good ones,
get to 202 fifty I think. I think that we'll be all that's
going to shift down. So I think AAI native company
will get to to 150, zero to fiveand then two 53105 to 10.
And then I think you got a shot to get to 305 hundred.
(32:44):
Fascinating part. Of me is an entrepreneur.
The interesting question is should you take that extra and
invest more? Because if you see enough
opportunity, you should probablynot harvest that hard.
Like I would argue cursor may bedoing the wrong thing.
So I don't know. Time will tell.
If they had another 50 people, could they, could they have a
(33:05):
dev OPS product? Could they have a, should they
do some cloud hosting or something?
I think, yeah. Each entrepreneur has to
navigate that complex matrix. Yeah, reminds me to my early
days at eBay, way back in whatever in 1999 where we were
crushing it in Germany and made,I don't know, like $0.70 on
every or made a dollar on every,whatever $0.60 we spent or
(33:28):
something. And we took all of the money and
poured it back into the marketplace just because that's
how you get to dominance, right?That's how.
And it worked. Today's eBay would not do them.
They would cash out every penny and then they'd be shocked when
it stopped growing. Totally.
Hey, I have a last question for you as a bit of an oddball, but
I would love to hear your opinion on it because you're an
(33:50):
expert and I see you and our dear friend Jason talk about
this on LinkedIn. I follow the two of you quite a
bit in this kind of liminal space, this in between moving
from an old world order to a newworld order, the old world order
being Google search being the dominant way for us to refi and
buy. Is this changing?
(34:11):
Is it the end of Google? Part of the fascinating part of
the liminal space is the innovator's dilemma.
I'm sure you actually know that book better than I do.
And I never, I read the book andit was like academic.
And then I went to meet the CEO of Borders at Channelvisor and
he was telling me how the Internet was a fad and people
love coming into the coffee and they would gladly pay more for
books because they could have a cup of coffee.
(34:32):
And I was like, wow, this is innovator's dilemma right here.
This guy is the living embodiment.
So I think it's going to, I don't know the answer, but I
think it's going to be fascinating.
You have two companies, Google, maybe 3, Google, Meta and
Amazon, who have basically explicitly said we're going to
beat the innovator's dilemma andit's going to be interesting to
see if they do or not. Amazon did one time with Kindle,
(34:55):
coming out with Kindle and self eating.
The book market was genius at the time.
It seems crazy, but Netflix is another one that has had a
history of doing this. So maybe we're in this new
generation clearly defining the innovator's dilemma.
People can get out of it. So Google's doing everything
they can. I don't think they're going to
because their innovator's dilemma is they've got the best
business model in the world where money just rains out of
(35:16):
the sky. It's going to be very hard to
get out of that individual enough.
But yeah, the they're doing someof the right things.
But is that enough? I don't know.
What I can tell you is personally, I've switched
perplexity and ChatGPT because the user experience is so much
better. You get to the answer faster,
but now you start to use some ofthese deep research models, you
get to not only the answer, but you get to the next 12 answers.
(35:38):
It's pretty crazy. And this is what makes it an
exciting time to be in where we are.
I found this interesting. I saw you posting on LinkedIn
that you switched essentially most of nearly all of your
searches to perplexity, which was roughly the same time.
I did the same thing. And I don't know how you're
dealing with this at the moment,but I found myself using ChatGPT
(35:59):
with the search feature significantly more in perplexity
these days because the results in ChatGPT I find actually are
better presented, typically tendto be better.
Yeah. So it's weird and interesting to
see how fast we switched from like we go from Google search to
perplexity and now we're just already like on to the next one.
Yeah, that's the liminal space. Yeah.
(36:21):
I also have a softness in my heart for the underdog and I
really have been the CEO of the perplexities interviews and
things. So I haven't switched yet, but
it's maybe just a little bit of loyalty there.
I agree with you. The ChatGPT has surpassed it at
this point. Scott this was an absolutely
phenomenal conversation. I loved literally every single
bit of it. I learned so much.
(36:41):
We will put links into the show note for people to follow the
exciting annulments which will come out of your company I'm
100% sure knowing your track record.
Also talking about good podcaststo follow, we'll put a link to
the Jason and Scott show which is an absolute must listen to,
particularly if you're in the e-commerce space.
But also even if you're not in e-commerce and just want to hear
(37:04):
2 very smart people talk about the future and banter quite a
bit, which I love. Highly recommend that.
Jason's the grumpy, the guy in the liminal space.
I hate change. So it ends up he's the guys on
The Muppets, the grumpy guys on the balcony.
There you go. He's the cremation of the two of
us. Scott, thank you so much.
This was amazing. Thanks Pascal.
Thanks for introducing me to theliminal space.
I'm going to go take a deep diveand swim around in it for the
(37:27):
rest of the day. Perfect.
Thanks. Disrupt disruption.
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Las Culturistas with Matt Rogers and Bowen Yang
Ding dong! Join your culture consultants, Matt Rogers and Bowen Yang, on an unforgettable journey into the beating heart of CULTURE. Alongside sizzling special guests, they GET INTO the hottest pop-culture moments of the day and the formative cultural experiences that turned them into Culturistas. Produced by the Big Money Players Network and iHeartRadio.
Crime Junkie
Does hearing about a true crime case always leave you scouring the internet for the truth behind the story? Dive into your next mystery with Crime Junkie. Every Monday, join your host Ashley Flowers as she unravels all the details of infamous and underreported true crime cases with her best friend Brit Prawat. From cold cases to missing persons and heroes in our community who seek justice, Crime Junkie is your destination for theories and stories you won’t hear anywhere else. Whether you're a seasoned true crime enthusiast or new to the genre, you'll find yourself on the edge of your seat awaiting a new episode every Monday. If you can never get enough true crime... Congratulations, you’ve found your people. Follow to join a community of Crime Junkies! Crime Junkie is presented by audiochuck Media Company.