Latest / Founder Mode / AI Agents Are the New Employees

Transcript

AI Agents Are the New Employees · Founder Mode

0:00
The through line here is simple. The founder's job has changed.
0:02
Yeah, there's just less doing and more deciding, right?
0:05
Yeah, it used to be build the product, hire the team, scale
0:07
the OPS. Now it's design the system,
0:09
orchestrate the agents and make the calls.
0:11
That can't be automated. You're listening to Founder
0:28
mode. Where we breakdown what it
0:30
really takes to build and scale companies.
0:32
Today we're talking about what might be the biggest operational
0:35
shift since cloud computing. AI agents that work like
0:38
employees. Not Co pilots, not chat bots,
0:40
actual workers that get hired, assigned work, evaluated and
0:43
maybe replaced. All right, Jason, I feel like
0:46
every week we say this changes everything, but this one
0:49
actually might. Yeah, I, I, I think that today
0:52
we're not just talking about another tool, we're really
0:55
talking about like a staffing model.
0:57
Yeah, and once you see it that way, you can't Unsee it.
1:00
Yeah, these are like, I think there's this shift that's
1:03
happening, right? It's not, it's not the Co
1:04
pilots, it's not chat bots. These are workers that are being
1:09
hired by, you know, people like us and and other founders out
1:13
there like you listening that are being assigned work.
1:16
And then if they don't do a good job, they're getting feedback
1:19
evaluated, replaced, swapped out for a better agent, right?
1:22
Yeah. I mean, for sure.
1:24
I think what's interesting is when you start to think about it
1:27
that way, you start to treat the tools that you use differently.
1:30
And I think that's kind of, you know, so like typically, like, I
1:33
mean, you'd buy a SAS app and you'd be like, OK, use this
1:36
thing. I set it up, it kind of works,
1:38
but now you get something you're like, oh, can I coach it?
1:40
Can I teach it something? Like you start to have this
1:43
sense where it's like, oh, it's going to have memory and context
1:44
and it's going to remember things and it's going to learn
1:47
so that if I teach it something today, I won't have to teach it
1:49
tomorrow. And I think that's the part
1:50
that's it's kind of crazy, right?
1:52
And so I mean, what it was just a few days ago, what Feb fib
1:55
open AI launched frontier And you know, the reason it matters
1:59
isn't the tech. It's like this mental model.
2:00
And it's funny, like I think Frontier was out there and
2:04
honestly kind of late to the party, right?
2:06
And you would throw it a task or throw it to do something and it
2:08
wouldn't just be like answer the question one shot.
2:11
It would be like, oh let me just keep trying.
2:13
But the the Frontier announcement is interesting and
2:16
the idea that agents can be staff and so you give them their
2:20
credentials and they log into Salesforce, they run work day,
2:23
they handle workflows and and you give them a performance
2:27
review. Like all that stuff feels kind
2:30
of wild, but also very, very 2026 I guess.
2:33
Yeah. And so the real question for
2:35
founders becomes like, if 80% of the execution is automated,
2:39
what's the job left for like the humans to do?
2:42
Yeah, there's an article that's going viral right now on X about
2:45
this, which is like basically this guy is predicting that all
2:51
that basically most of if not all knowledge work, white collar
2:55
work is going to just evaporate. And that, you know, the 1st
2:58
place that is being affected tech, because that's where AI
3:02
has been embraced the most. But it's going to come for all
3:04
of us and all roles and all jobs everywhere.
3:07
And also the humanoid robot stuff that's going to then, you
3:10
know, the kind of blue collar is next, right?
3:13
And that's going to take longer. But you know, I think that
3:16
that's the like there's a little bit of a doomsday kind of
3:19
dystopian element there. But then, and this is what
3:22
you're, you're working on all the time is like, but if we can
3:25
take the automate that, if we can automate the tasks that are
3:28
really repetitive and manual labor that humans are doing all
3:31
the time, what couldn't the humans then do?
3:33
And how can we actually take advantage of of human capability
3:37
and, and, and and brain? And I think that's kind of what
3:40
you're building with the you you were in the hat again.
3:42
Well, you're building a pretty good AI.
3:43
So I don't know if you want to talk a little bit about that.
3:45
Yeah, when you say doomsday, like I almost kind of giggled
3:49
and cut you off because like I it's like to me it's everything.
3:52
Not that like it's. No, I agree.
3:54
But I think so a lot of mainstream anybody that's not
3:56
us, that's not working in tech and in and everyday, I think
3:58
they're scared. I think people are scared.
4:00
But it's like, man, there's just such like, and again, it's hard
4:03
to even describe like how exciting and cool it is.
4:06
And I think like I think back to.
4:07
Be clear, I'm with you. I think it's exciting, it's
4:09
cool, but I think a lot of people.
4:10
Can be on the cool kids side today because I do think like
4:14
man, if you look back there was this notion when like Uber and
4:17
even Instacartoon extent came out.
4:19
It was like, well, are you going to be building the robot or are
4:23
you going to be getting told by the robot what to do right?
4:25
If you think about it like, you know, and the robot being like a
4:28
phone and sort of a lot of these network businesses, you would
4:30
say like the phone tells you where to go get groceries, where
4:33
to pick up the person, where to drive them to.
4:35
Or are you the person inside of Uber or inside of Insta card or
4:38
one of these companies building the robot that then you know,
4:41
sort of like builds that model? Before I go there, I do want to,
4:44
there's something about where you're just said that just, it's
4:46
very interesting to me because we're also trying to train the
4:49
AI to handle like edge cases like a human would, right?
4:53
So like, what do you do when you can't log in?
4:55
Like how do you figure that out? But I think there's the other
4:58
side of it that I know you've thought a lot about, which is
5:01
it's just agent talking to agent.
5:03
Like the web as we know it today changes a lot because you don't
5:06
need a login screen and an interface when all of it is
5:10
agentic and everything is just, Oh yeah, I sent my agent off to
5:13
go book me that reservation or go do that task for me.
5:17
And it's now talking to another AI.
5:19
Remember the, I think we talked about this in a prior episode
5:21
with a video of like the two AI's realizing their AI's and
5:23
then they just went to some like fax machine style sound talking
5:26
to each other. But that's going to happen in a
5:28
massive scale. Like what happens that?
5:30
Yeah, I think a bunch of cool stuff happens.
5:32
I think you get some really interesting outcomes where
5:34
you're like, oh, that's amazing. It kind of was super delightful
5:37
and that's kind of the happy path.
5:38
I think today the happy path is very narrow.
5:40
And So what more than likely happens is you get things
5:43
spinning off and, you know, in the case where the agent has the
5:46
ability to keep trying, may just burn a bunch of tokens, waste a
5:49
bunch of time, create a big mess, right?
5:51
Quite frankly, just doing the wrong thing over and over and
5:53
over, expecting, you know, this is the definition of insanity is
5:56
right, Doing the same thing over and expecting different results.
5:58
The agents definitely get into that sort of insanity more often
6:01
than they should. And then I think in the failure
6:03
case, they just kind of die and shut down.
6:05
So they fail closed where they're off.
6:07
And then the question is, do they notify before they fail,
6:10
where a lot of times they don't. It's like, oh, I just shut and
6:12
you're like, wait, this thing hasn't run in a day or two days
6:15
and nobody checked on it. Somebody check on the agent.
6:18
So, yeah, going back to your question about my, the companies
6:21
we work with, yeah, it's interesting.
6:23
I think, you know, you're, you're the engineers that you're
6:26
working with every day are much more plugged into like the
6:29
latest and greatest tech and all the frontier models and the
6:32
developments that happen literally every day from from
6:35
the kind of the all the AI companies and players, right.
6:37
I think in growth and marketing, which is where I spend most of
6:40
my time, you know, it's, it's less like 2005 now.
6:46
It's definitely more like 2015. And so the, the kind of TLDR and
6:49
everything I just said is if you're a founder and your growth
6:52
stack involves like the same tools that people were using
6:54
five years ago and no AI, you're like, you're going to fall
6:58
behind. And so it's like, how can you
7:00
start to integrate this stuff into your workflows tomorrow?
7:02
Yeah. And I'd even go a step further
7:04
and say like in many, I mean like I we use HubSpot, which
7:07
obviously the tool that's been around 5 years ago and
7:09
Fireflies, which always been five years ago or more.
7:11
And so I think that, but the question is, are you logging
7:13
into them every day the same way you were a year ago?
7:16
Or are your agents logging into them and using the data from
7:18
them? Or do you have workflows where
7:20
that all that stuff is automated and instead you're getting your
7:22
daily meeting prep sent to you, which you you send me sometimes.
7:25
And they're so good now it's like, and no EA or virtual
7:28
assistant did it. It's a complete download of all
7:30
the data of like what's gonna happen, right?
7:32
And it's hard because there'll be meetings after someone left
7:35
the office or their meetings that I wasn't in, that someone
7:37
else joined, that I was gonna join, or my agent joined and I
7:40
couldn't go to that meeting. And now you can combine the data
7:44
from that meeting or the transcript kind of from that
7:46
meeting combined with like the status of like and the activity
7:50
in the deal and HubSpot combined with, you know, some document or
7:54
plan that you have and then that can go then generate you an
7:57
e-mail. And again, like, I think that's
7:59
sort of the the sort of state-of-the-art today of like,
8:02
you know, things where they're like taking these kind of
8:04
desperate disparate systems and bringing them together in sort
8:07
of interesting ways where a human normally would have had to
8:10
reason over that. I think the next level and, and
8:12
sort of where we had started this conversation was like, can
8:15
you then build an agent to actually act upon that e-mail
8:17
and say, oh, because of all this, now here's Kevin's 12
8:20
meetings for the day. Just go to all the meetings for
8:22
Kevin and do the and do what Kevin would have done, right?
8:25
And I think that's the thing that we haven't jumped that.
8:28
People want to meet with your agent or the do they want to
8:30
meet with Kevin or are agents going to meet with each other?
8:32
Or you know, is it still going to be 1 sided?
8:34
Yeah, No, I mean people still need people.
8:36
And I think there's that piece, I think, but the, the types of
8:39
meetings that you need to attend in the way that you can show up
8:42
in a much more prepared and sort of educated way is dramatically
8:45
easier. I had a meeting with CEO that
8:48
I'd reached out to that AI had finally recommended because I
8:51
was like, hey, I'm looking for this kind of peak companies and
8:53
this kind of thing. And, and I literally sent I
8:56
think 9 requests and you know, four of them got back to me and
9:00
I met with this gentleman and, you know, I was walking through
9:02
this and I did my research and obviously used a ton of AI to
9:04
research to understand it, but then was able to synthesize that
9:08
and have a very real time discussion.
9:09
It was like, wow, you're super prepared.
9:11
You understand the where the business is.
9:13
And there was some things I definitely got wrong, but there
9:15
was a ton of things were like my assumption of like what I saw
9:18
from the outside plus with some reasoning, plus AI plus what I
9:21
know about the space was sort of able to open a door in a way
9:24
that was unique. And I think so like, could you
9:26
have set a robot to that room? And there's no way that that
9:29
would the robot would have like, won that conversation.
9:31
I mean, he would have hung up. Of course, yeah, they or
9:35
wouldn't have the meeting wouldn't have started like it's
9:36
all these things, right. But I think the you know, again,
9:39
it's it's like the old days of like, well, you could do a bunch
9:41
of math by hand or you could run a calculator and get a bunch of
9:43
it, right. You still got to could have
9:44
explain and know how it works to get there.
9:46
But I think a lot of the synthesis is gone.
9:48
And I think, you know, that sort of spins into this like, you
9:51
know, is Frontier an AI product or is it really more of an HR
9:55
system for agent, right? You know, I you know, we'll see
9:58
right. I think like it's you can
10:00
describe how it's been presented.
10:02
But I do think the piece that the idea of being able to sort
10:06
of manage agents and guide them. And my co-owner JJ always keeps
10:09
saying, because I've hired, I mean, thousands of kind of
10:12
freelancers over the last 30 years through like, you know,
10:15
odesk, Upwork, yeah, fiber kind of thing, right?
10:18
And to work that correctly, you have to be able to size and
10:23
scope and delegate really well. Like, what kind of person
10:25
handle? How do you sort of like give
10:27
them something, give feedback? Oh, that was too big of a task
10:30
or too smaller. It was too discreet or it was
10:32
too, you know, complex. And then a lot of working with
10:35
AI is very similar, right? You try something and you're
10:38
like, oh, it didn't really understand what I wanted.
10:39
Let me give more context. Or it knew what I wanted, but it
10:42
was like it didn't actually get me, saved me much time because I
10:45
didn't give it a hard enough problem.
10:47
So then it's like, oh, I need to make the problem harder.
10:49
And I think that sort of idea of this kind of like how do you
10:52
decide how high to fly when you're kind of coaching these
10:54
agents has to me been a big thing.
10:56
But I'm curious when you think about this from the growth
10:59
stack, marketing stack, like if anything, people always joke
11:03
like the technology stuff's actually easier because it's
11:05
very discreet. It's ones in the end, it's like
11:07
you write code, code fails, code passes.
11:10
Is code good or or bad is like somewhat subjective.
11:13
And at the end of the day, you can argue that codes good or
11:16
bad. But if like if it does the job
11:18
and it scales to meet the job, you're like it's good, right?
11:20
Like from a business impact where I think marketing and
11:23
growth and branding super hard. Like we can all chat GPTA logo
11:27
which I've done before and maybe you can tell like how what
11:29
you've worked out. Do we evaluate the people and
11:32
the agent employees differently? Like how does this stuff?
11:34
Where is this going right? Part of it's like, well, you're
11:36
only going to find out and tell you no, you have to try it,
11:39
right? And like, there's probably a
11:41
point where they were moved to the fax machine.
11:42
It was like, do you think it'll actually make a copy of what's
11:45
going in that'll actually come out on the wire on the other
11:47
side? We've been so used to like Pony
11:49
expressing this thing across the town that now the fax machines
11:52
like more, more default than e-mail they trust because of
11:55
Nipa and things like that. And so to me, it's like, you got
11:57
to use this stuff and you got to try it.
11:59
And if you're not trying it, you're going to fall behind.
12:01
Yes, and. The stakes are, the stakes are
12:04
high when you're in the car and it's driving you.
12:07
The stakes are pretty low if it's like just this agent that's
12:10
going to go optimize your sales pipeline.
12:12
Yeah, unless it deletes it all right.
12:14
But it may, right? It might like it.
12:15
Might have a backup copy before you let the agent loose.
12:18
Maybe. Yeah.
12:19
So, so then I think it's like, OK, is the new founder skill
12:22
like just bring me back to founder mode?
12:24
Like what am I'm a founder. What do I need to be able to do?
12:27
Is it orchestration? It, you know, coding maybe in
12:30
the traditional sense becomes more optional and it's more
12:34
about systems thinking, right? Like I'm not and don't take that
12:37
as a slight. I can see you getting worked up.
12:39
But like learning to code in the traditional sense might be,
12:42
might be different now. And so, you know, is it less of
12:46
can we build this and more should a human touch this at
12:50
all? Yeah.
12:51
And I think the, yeah, the thing that I was kind of winding up
12:54
on, as you were saying that is thinking back to what Toby from
12:57
Shopify's notes, that was like, you should flip to default AI.
13:01
Like I'm about to do something. Is there an AI tool that I have
13:05
in my quiver or something that I want to go investigate that
13:08
could do this? And I think it's moving from
13:10
that sort of mode of like, well, you know, I use AI when I, when
13:13
it falls into something that's very clear.
13:15
I have a question and I'm like, I'm no longer Googling.
13:17
I'm chat chatting or ChatGPT ING or Claude or whatever your
13:20
flavor is. But it's moving to this motion
13:22
of like, no, no, no, I'm about to like sit down at my desk or
13:24
pick up my phone or pick up my keyboard or type something like,
13:28
is there a way that AI can do that?
13:30
And I think it's that sort of, you know, the, how do you get
13:33
strong in the gym? You got to lift the weights.
13:34
And if you're not lifting the heavy weights, you're not going
13:36
to get stronger. And so part of it's with, if
13:37
you're not using AI and trying, you know, you're, you're not
13:40
going to get stronger. And so part of it's like forcing
13:42
yourself to like, think about, is there an agentic or AI
13:46
solution for everything you do? And it does create like this
13:48
kind of crazy spiral effect in your mind at times.
13:51
And that's where I'm again, I think of it more as curiosity.
13:54
Let me push on the let me push on that metaphor, though,
13:56
because I think like the go, go lift the weights and get
13:59
stronger is agree. That's what you what you want
14:02
people to do is like use these tools to make yourself better.
14:05
I think there's a lot of a parallels right now to GLP ones
14:08
or steroids where it's like you can use AI and you can, you
14:11
know, it's a cheat code or a shortcut.
14:13
And then you get we've talked a lot about this, like the slop
14:16
bot of you know, and you I even watch you call people out all
14:18
the time on this. It's like this is clearly AI.
14:20
You have done nothing here. You've just copy pasted from
14:23
ChatGPT like go do it again, right?
14:25
Like go actually do some work. Think about it.
14:28
If you're going to do slop, then like do the next step which is
14:30
take the slop and publish the slot slot.
14:32
Versus like your interpretation like don't just send me a sloppy
14:35
e-mail. Go do the work and don't send me
14:37
the e-mail, send me the answer. No, I think that's good.
14:39
And then so maybe it's taking it too far, but you know, is there
14:42
a world where we go back? I don't I don't think we do with
14:45
AI either. So it's like, you know, again,
14:47
it's how are you using the tools to more effectively utilize your
14:52
human brain? And the part that I just can't
14:54
see the answer to you and maybe you have a stronger perspective
14:57
on this is like what is going to be left for us to do?
15:00
And it's not just the like, are you the Instacart person or the
15:03
Uber person that's like building the machine versus the person
15:06
using the phone to be told what to do by the machine?
15:08
I get that. But there's a very limited
15:10
number of those jobs. So like, how else can this is
15:13
more for regular people? But like coming back to the
15:15
founder audience is like, you want to be a founder that's
15:17
crushing it and you want to use these tools effectively.
15:20
What is the next, what is the next thing that you go do?
15:22
Right now you're I think we're keeping it a little bit too
15:25
broad. Like where, where do we want to
15:26
send people, right? So.
15:27
Yeah, So I, I, I think you can't put it back.
15:29
I think what's the interesting thing for the humans to do is a
15:32
version of like the the humans again.
15:35
What what would you do if you had a bigger team?
15:37
You know, there's it's to me, it's like that's, that's.
15:39
How can you get? How can you put your big AI
15:41
agent workforce to work effectively?
15:43
Could you do more things? Could you have more fun?
15:44
Like could you have plan more cooler adventures?
15:46
Could you plan cooler vacations? Can you get more time back to
15:50
enjoy humanity in life? That's that's the choice where I
15:54
fail like where I'm like, no, I'm going.
15:56
To spend more time orchestrating.
15:57
And do, yeah, I do more in the same amount of time or, or think
16:01
I'm doing more. But I think like, I do think
16:03
that that your answer is more appropriate and probably more
16:06
sane, which is like, hey, like, because I can spend less time on
16:10
drudgery, I can get the same impact with less work and, and
16:15
have more free time with family, kids, entertainment and things
16:17
of that nature. I mean, there is a view that
16:19
like it goes to like you spend, you know, 10% of your time at
16:23
work and 90% of your time entertaining and relaxing.
16:26
I think there's that view. Now you get to do these random
16:29
political things about like, you know, universal.
16:31
Basic income, yeah. All this kind of stuff I don't
16:34
know. Man, you're you're not going to
16:35
be a man of leisure. I know that so So what about
16:37
like hiring people that can see this for what it is, right.
16:40
So like the idea of bringing an operator who understands
16:43
orchestration, what, what would we, what would we ask them?
16:46
How do we how do we assess that ability?
16:48
I mean, honestly, like when we're hiring now, we just have
16:50
assessments and we're like, hey, go do this thing.
16:53
And we used to. Go async.
16:54
Yeah, used to be like, oh, answer these five questions.
16:56
Now it's like, go write this project and go, you know, one of
17:00
our assessments for the engineering is like go write AAI
17:04
calling bot and make 10 phone calls to our test system and
17:08
find bugs and then and generate that and send it to us with your
17:11
code. And what's in you're like, OK,
17:13
that's and saying you're having somebody write the entire QA
17:15
tool for your company. I would say so far, about 30% or
17:18
40% of the people can pass it and, and send you something that
17:21
works, which shows it's like a reasonably fair to like it's
17:24
like not 10% and it's not 90, but it's possible.
17:27
And it's like for a reasonably competent for person, they can
17:31
finish it in like a few hours, like a definitely a day or two
17:34
Max, like, you know, depending on how you want to Polish it.
17:36
And so I think that's one way to hire people just say, say, like
17:38
go use some of these tools and, and pick.
17:41
And what's crazy there is you really see how they think like
17:43
they they use off the shelf tools and SDKS and open source
17:47
or do they try to like Claude code in like raw code the whole
17:50
thing and vibe code the answer because I mean, honestly, like,
17:54
what would I do it? I'd put the whole brief into a
17:57
chat and ask it like, how would you build this?
17:59
How can I build this in two hours?
18:00
What tools they and just run that loop and you'll get the
18:02
answer. And that's one way.
18:06
And I think so to me, hiring people that know that and think
18:08
that way, they're just going to have much more plasticity, I
18:12
guess is the word for the brain. I don't know if that's the right
18:13
word, but like some version just start.
18:15
They're used to that sort of model of thinking and there's
18:18
almost a penalty if you've done it the old way because you're
18:20
too proud of the craft, right? And so you're like, no, no, the
18:23
craft. Is coming back two weeks later.
18:24
They're like I built the perfect app that does this.
18:26
I hand coded every line. So I think there's a time in
18:30
place like you're not going to probably launch the space
18:32
shuttle. But I think I don't think
18:33
they're vibe coding. A lot of SpaceX like launch
18:35
code. I bet somebody's.
18:36
And if they are somebody's, there's a lot of checks and
18:38
reviews. But again, like those checks and
18:40
reviews can be deeply automated in QA and stuff, probably at a
18:43
level that was never possible before.
18:44
Well, that's a big, that's a good segue to to big team versus
18:47
small team, right. So there we're now.
18:49
Are we in the solo founder renaissance?
18:51
For sure that it exists the, the number I so I follow a lot of
18:54
these kind of solopreneur, sort of like, you know, man in a box
18:58
and a computer kind of code thing.
19:00
And they're, it's incredible, like what some of these folks
19:03
and now a lot of them fall into this like screaming on the mic
19:06
course kids where they're like $72 a month.
19:11
Code will change your life, you know.
19:13
Watch this video, pay me $100 a month and I'll teach you
19:16
everything. Yeah, and I and I think there's
19:17
a business for that and I think that's on and there's a lot more
19:19
people that want. It's just like the, you know,
19:21
what sells better gym memberships or you know, the,
19:24
the wid you know how to like how to get strong and how to get 6
19:28
pack abs. Like first, sure, they have sold
19:30
more 6 pack ABS courses then people that have went to the gym
19:33
to do the six pack ABS exercises.
19:36
And so I think the course is always the cheat code and you're
19:38
not putting in the work. But I do think the solo
19:41
business, and there's a ton of them out there that are these
19:42
little SAS businesses or little custom tools that it can
19:45
generate great income very quickly with one person.
19:48
And then it's just a matter of what's the impact and what can
19:50
you do? And I think part of it, where
19:51
are you in your career and your life and what do you want to
19:54
focus on? I think I like the idea of the
19:56
ten person sort of team where you get enough diversity, you
19:58
get enough different different skill sets together and
20:00
experience where you're like, you know, the prompting is more
20:03
unique. Even if you're all using AI,
20:06
There's there's something about like a single person, a single
20:09
point of view that diversity of thoughts missing.
20:11
So I do think it like my sense is the 10 person teams are going
20:14
to be again, picking that as a random number are going to build
20:17
bigger and more impactful things than the one person teams on
20:20
average, but also better than the thousand person teams, the
20:23
hundred person team somewhere in the middle.
20:25
I don't know. Yeah, those sound very big.
20:27
Now I just think about the like the this the myth of the solo
20:31
founder Unicorn, you know, billion dollar company, right?
20:34
And then, you know, can AI, could AI replicate that
20:37
diversity of thought and, and point of view, especially if
20:39
you're using multiple frontier models and LLMS and kind of
20:43
agents that are coming from different places.
20:46
You know, we we had it all up. So we talked about kind of like
20:48
the board of directors model, like I wonder if that's if
20:51
that's possible. And is it, is it a founder with
20:53
their swarm of agents or is it, or is it, is it a still a team
20:56
of 10 people? I, I know, I know where you sit
20:58
on it. But I, I think, I think also, if
21:01
you were to start tomorrow, you might not hire a bunch of
21:04
people, would you? No, I think there's a chance.
21:06
I think it really depends on what you're trying to do,
21:07
honestly. Like I think there's certain
21:09
problems that. Baby sass is different than like
21:11
if you're doing the. Place good care business yeah or
21:13
a media company or something. I do think there's a chance
21:16
where you could go like single thing only agent.
21:19
But I do like for some of these, you know, again, if you're
21:22
building an HVAC business or a plumbing business, like we just
21:25
don't have the tools in the context yet to like unplug a
21:28
toilet with AI yet, right? But if it's like, and so it's,
21:32
I, I would describe it as this, it's like, where are you on that
21:34
digital sort of sense of like, you know, it's the, it's the
21:37
same, it's no different than the Nomad with the backpack and a
21:39
laptop. It's like there's a set of
21:41
businesses that work really well for moving to Europe or
21:44
Southeast Asia from America or vice versa.
21:46
If you're in one of those countries and then just working
21:49
out of a backpack to the hours of your users, only working on
21:52
every 3rd Saturday, surfing the rest of the time, and you can
21:55
build a very sweet digital business.
21:57
A lot of it's been done. Now you're going to build a
22:00
business selling to the, you know, Fortune 500 and having to
22:03
walk into offices or sit on like synchronous Zoom calls.
22:06
And I think you've seen a couple businesses like this for
22:08
yourself, like you've, you know, started with businesses that are
22:11
heavily human based and now move to businesses that are maybe
22:13
more digitally focused. Yeah.
22:15
Yeah, I think that's right. So do you think that this
22:16
creates more founders or fewer? Definitely more, 100% more.
22:20
There's no doubt in my mind that that, and I would even go a step
22:23
back and say it's not just founders, builders like, yeah,
22:25
just, I mean, how many things have you built?
22:28
Yeah, I could never build and now lovable is like my best
22:30
friend. You would never have considered
22:32
yourself a builder. I'm gonna build something for
22:34
you in Lovable later today. I'm ready.
22:35
I can't wait to see it. But that's the magic, right?
22:38
I think. I don't think that existed a
22:40
year ago. Yeah.
22:41
You would not have called yourself a builder in that.
22:43
Sense no, I'm, I'm, I'm starting to dabble in cloud code.
22:46
I'm, I'm not, you know, I'm getting there.
22:48
So so then, you know, there's this like point of how and when
22:51
to keep humans in the loop and we've talked about this in prior
22:54
episodes too. It's like the, the failure mode
22:57
here is if you automate the judgment too early, right If you
23:00
if you lose the human judgment today, I you know, potentially
23:05
you can delegate away things that actually do require some
23:08
oversight and like you want the agent to come back and check in
23:11
with you or you want to give it more specific instructions and
23:14
tune it and refine it right. So like, is it, you know, you
23:17
talked about enterprise, like enterprise sales, sales
23:19
relationships, key customer relationships, you know, the
23:22
various trade-offs that we have with around strategy decisions
23:26
that are going to involve real resources.
23:28
And also like I talked about a minute ago, creative direction
23:30
and brand voice. Like there's all these companies
23:32
that have popped up and I even got one earlier today.
23:34
It's like create 600 videos a month for $10,000 and syndicate
23:38
them across every social platform.
23:39
I don't know what the quality of those videos are going to be,
23:42
but you know, there's whole companies that are built around
23:44
this now. And so I think the question is
23:45
like, when do you automate and delegate to these agents?
23:49
And is it because you can do it or because you should do it?
23:51
Again, it's more fun. Like, if that drives you crazy
23:54
and you can't do it, like, then don't.
23:55
Like, I mean, part of this is like it, dude.
23:57
It should be fun. Like, if it's not fun, you're
23:58
not going to go to, you know, you're not going to keep it up.
24:00
Yeah. I don't know what the real
24:01
answer is. I think for me, I just love
24:03
thinking AI first. It forces me to be more curious.
24:06
What's the Moat now? Yeah, we debate this one a lot
24:10
because I think you can go prompt like some of the most
24:13
interesting things on it. Go go build me Yelp, go build me
24:16
Uber, right? And it can build a good version.
24:18
And so then it's like, well, do you have the community?
24:20
Do you have the brand, those kind of things?
24:22
I think so those like the branding and the community, the
24:24
marketing thing, like do you have interest?
24:25
You have a megaphone, can you get people to come to it?
24:28
And then it's like, are you integrated deeply enough into a
24:32
workflow or system where it makes it very hard?
24:34
I was talking to a company that does like these courses.
24:38
And for them, the mode is that they've got a decade of like,
24:41
not just the courses and what people have built and learned,
24:44
but they have like what people are actively searching for
24:48
today. Like what are people trying to
24:49
learn today? Because that's actually more
24:51
interesting than the course. Like they have all the course
24:53
data that sort of evolved slowly over time, but it's like when
24:55
people wake up and they go, oh, they all saw open claw.
24:58
They're all like searching for the class on open Claw.
25:01
You want to learn it? Yeah.
25:02
And a lot of times doesn't exist.
25:03
So then they end up on YouTube or X or somewhere where they're
25:05
like a more ephemeral sort of quick RIP sort of thing.
25:08
So I, I I think that ends up being the game, but I'm curious
25:11
what you said in marketing and bringing like what is the Moat
25:13
for you guys? Yeah, I mean it's distribution.
25:15
The people that already have, you know, I was talking to
25:16
founder the other day, their website, you know, their big
25:20
subscription business and game online gaming and and very cool
25:24
business. And they have like hundreds of
25:28
thousands of organic views on their site every day just
25:30
because they've built this huge community and audience.
25:32
People know them. And so, you know, when you're
25:35
talking about testing and running experiments to get
25:37
additional traffic, everything is a drop in the bucket compared
25:40
to that. And so you have to really
25:42
calibrate like, what is the thing we're going to do that's
25:44
going to be high leverage? It's going to have high impact
25:46
and you know, agents and AI can generate a bunch of ideas, but
25:50
it's the the little bit of the experience and knowledge of
25:53
knowing what has worked in the past for similar business,
25:55
similar users and and kind of stack ranking based on not just
25:59
a kind of AI slop list of ideas, but better.
26:01
Like what are the backs against the backdrop of this kind of
26:04
unique problem space and reach of this brand?
26:06
Where can we? Where can we grow?
26:08
A through line here. Simple.
26:09
The Founders job has changed. Yeah, there's just less doing
26:12
and more deciding, right? Yeah, it used to be build the
26:14
product, hire the team, scale the OPS.
26:16
Now it's design the system, orchestrate the agents and make
26:19
the calls that can't be automated.
26:21
Yeah, Kevin, what what were your big, you know, 5 takeaways from
26:23
our chat here. Yeah, I mean agents are
26:25
employees now. I mean Frontier is an example
26:28
that's you hire task and evaluate AI like staff, human
26:31
plus agent systems is really that core founder skill.
26:33
Like how do you map together Global founders can scale.
26:37
I mean, one person can do 10X the work or maybe even 100 X
26:40
depending on what it is of, you know, a team of couple years
26:43
ago. Some decisions just have to stay
26:45
human. Trust, taste, judgement.
26:47
At least today. Don't delegate well to the
26:50
agentic world. And then finally, like, the Moat
26:52
isn't the agent, it's your data, your workflows, and quite
26:55
frankly, sometimes restraint. If this episode made you rethink
26:58
your org chart, send it to your Co founder.
27:00
And subscribe to Founder Mode. We'll keep breaking down what
27:02
actually it takes to build in 2026.
27:05
That's it for today. Because the founders who win
27:07
next won't build the most software.
27:10
They'll design the best system.