There are people that are spitting up

agent workforces with hundreds of agents

and sub agents and they're getting

incredible amounts of work [music] done.

But how do you do it? And how could you

think about it? And what are the

strategies to actually create an AI

agent workforce

>> [music]

>> that under promises and over delivers?

Well, today I brought on Ali K. Miller

and she's one of one of the most

well-known AI voices ever. She's worked

with IBM, she's worked with AWS and

she's managed multi-billion dollar P&Ls

in the AI space.

I asked her a simple question, how do

you manage your fleet of agents? In this

episode, we cover a lot of ground, but

by the end of it, you're going to

understand how should you should

strategically think about spinning up AI

agent workforces, where there's

opportunities to create startups in the

B2B space with AI agents and a lot

[music] more. Enjoy the episode and I'll

see you at the end. Today's episode is

brought to you by Brex. My company's

been on Brex for a year and a half and I

started because I kept hearing companies

like Vercel, OpenAI, and Anthropic were

using Brex and I figured if they're

using it, why shouldn't I? It's been a

game-changer. The thing that got me is

how smooth it is. It's got high-limit

cards, it's got banking, it's got AI

that handles the back office busywork

like expense reports, which I don't want

to do, on its own. It's really just

built for this agentic world. If you're

building something new, it's time to get

Brex. Check it out at

brex.com/solutions/startups.

Link in the description.

>> [music]

>> I can't tell you how excited I am to

finally have Ali Miller on the podcast.

I've been begging her to come on. She's

one of my favorite people in AI and I

don't say that lightly. Um welcome to

the show, Ali.

>> Thank you, Greg. And you are also one of

my favorite people, so like I'm actually

very excited to to talk about all the AI

things that we're working on.

>> By the end of the episode, what are

people going to learn?

>> I hope like one of the biggest mindset

shifts that I'm going through right now

is I feel like the term managing agents

is wrong. And my hope is that people

will understand what that mindset shift

is, see a few examples, and figure out

how to start how to make that mindset

shift, what the first step should be.

>> Okay, perfect. So, where do you want to

start?

>> So, this is and and I'm happy to to

debate you on this cuz we haven't

chatted about this. But I feel like

managing agents feels like I'm their

direct manager and I'm like, "Suzy, go

over there and Betty, go over there and

Jeremy, go over here." And I feel like I

am three rungs above at like an SVP

overseeing level

where I feel like I am setting up the

infrastructure and then

they are figuring out the best way to

execute within that. Um and so I feel

like I'm moving from managing to like

waiting for escalations.

Um or I feel like I'm moving away from

delegating and more just deciding what

should or shouldn't happen. And so it's

a little bit more of the like a like a

liability role where I just get to be

the the final say of what happens

um and come in for like critical

thinking stops. But does it like am I

the only one that feels like that is

happening? I just it feels like that

word is wrong. Like I see managing

agents everywhere and it just feels like

anyone that is still talking about, "You

should manage agents." feels like early

2026 talk.

>> Also, like do do we want to manage

agents? Is also the question. Like

managing people is hard, you know what I

mean? Like

a big reason I think a lot of people

like AI to do stuff for us is so we

don't have to manage things, you know?

So, that's something else I've been

thinking about.

>> Like, I I ran an org of about 100 people

at AWS. The parts of people management

that I loved, it was

the the making them better and

empowering the out of them and

seeing them completely blow past their

ceiling, watching them get promotions.

Like, that was the fun part and also

seeing what we could do together. Things

like, "Oh, we have to fill out this

thing with the paper and the button."

And like, get me out of there. So, I

think the admin side of people

management and the admin side of agent

management, I want that fully gone.

The things that I that I love about

people, I'm bringing that over into

agents, which is just like, "How do I

act as as ambitiously as possible and

get you to break through your ceiling?"

And one of the best prompts that I have

done with my AI workforce is three

words.

>> [laughter]

>> And with like a little bit of

explanation, but like, at its core, it

is three words. That is the best prompt

ever. So, I have

uh my AI chief of staff is Simon. Simon

runs like this whole org. And so, I have

34 AI agents that work in this

workforce.

And it dawned on me

that I was already functioning at the

limit of my own imagination in my

business. And that I could be doing way

more ambitious things if only someone

could manage me, right? Like, could

break help me break through my ceiling.

And obviously, I have a lot of mentors

and you're amazing at at, you know,

shaking people up and and making me

second guess how I'm doing things. It's

really helpful. But, I it dawned on me.

I was like, "Why am I not leaning on the

AI agents to help me with this? Like,

why is everything that they're working

on initially prompted by me? Even if

it's

um a scheduled task, I still had to come

up with that task and tell it to do it.

So, the best prompt, three words, and

it's just do smart things.

Like, my AI workforce has access to

every single context doc I've got.

Context docs about my business, my

friends, family, my 2026 personal goals,

business goals. It has access to my

meeting transcripts, email, calendar,

Notion, Stripe, Supabase, GitHub,

whatever.

And I just several times a day want it

to look across all these things and just

do smart things.

And seeing how Fable 5 and GPT 5.6 and

that level model is reacting to that

vague

um flavor of prompt. Like, you could you

couldn't do this a year ago. Now, you

absolutely can.

>> So, when you hire a human being, I think

there's like three types of employees

that you can have. One is uh someone who

doesn't complete tasks, not a good

employee if they're not completing

tasks. Um the second is they're

completing the tasks um like

satisfactory or exceeding, but like

they're not really like thinking about

new tasks. Um so, they're not You can't

just like if you step away from the

business, you're probably not going to

see insane growth. Um

and then the best employee that you can

possibly hire is doing the task,

exceeding expectations on it, but also

thinking about new tasks that they

should be doing, and actually going and

doing those, and

exceeding expectations or or you know,

or being very satisfactory on that. So,

what you're saying is

basically, you're just giving more

responsibility to your team of agents.

Um you're giving in a way cuz you're

giving these three words to it and

you're saying like, "Hey, I'm shifting

the responsibility of like

you know, do smart things to you." Like,

you have to you you have to like there's

a bunch of fog that you have to figure

out.

>> Yes. I would say I'm giving them more

breath, more scope, more flexibility.

I'm not allowing them to now send 100

emails and before I used to have to

check all the emails. I still check all

the emails. So, the the tier of risk has

stayed the same, but the width has

expanded.

>> It's it's almost like unbelievable that

those three words actually make a

difference.

>> Yes. This is like I And by the way, so

so I I agree with your assessment on

this like tiers of employees and Alex

Lieberman shared this like pyramid of

proactivity that I turned into I'll I'll

send this to you so that you can pull it

up right now as I'm talking about it.

But, it is five levels of proactivity.

And at level four, it's like I've

already solved this thing. Here are the,

you know, tradeoffs or whatever. And at

level five, it's like I've already

solved this thing. Here's how I'm going

to deal with it if it goes wrong. Here's

the next steps, all the things that you

just laid out. I would say that the

difference between someone who's at

level three and two in in your um

analogy is someone that understands

goals and someone who's been given the

power

to rethink how things get done and the

power to actually execute. And I give my

AI workforce goals. Like, those are

written out and every single quarter

also um share with you this tweet that

has like the prompt that I think

everyone can use. But, every single

quarter I'm going through a goals review

with my AI agent workforce so that the

goals documents that are living on my

desktop and are duplicated in the drive

so that all this cloud like workflows

can actually work. Um all of that is so

that AI, when it is in that expanded

scope world and it's taking on that new

tasks,

it's doing it in a goal-oriented way.

It's like giving it a product mindset.

Like I think it would be extremely

limiting if you only treated this thing

as an engineer when it could be the

greatest product lead you've ever had.

>> I think you tweeted about like your your

like men you're really focused on

proactive agents, right?

>> Yes.

>> about when when you talk about proactive

agents? Is this what you're talking

about?

>> So I when you talk to the AI labs and I

know you do and I know I do and a bunch

of others probably do. But the the word

of the year feels like it's proactive.

So I don't want to be the first domino

anymore. I don't want to be the

bottleneck in my own work and any single

moment that I realize that I am the

limiting factor of helping a billion

people transform their lives, work, and

business in the AI age, I have to remove

myself from the process and go, "Bad

alley, like what are you doing?"

[laughter]

And so a lot of that um especially in

the in the kind of tail end of 2025,

first half of 2026 was switching into

proactive agents. So we we had proactive

automations that were trigger-based. Um

I'll give you a really easy example.

Every single time I drop a video

recording into our video folder, so

basically anytime I do a screen

recording, goes into this one folder and

automatically it gets generated um

automatically generated is a transcript

of that video um that gets, you know,

then saved into our little transcripty

thing. Social posts get generated that

are in my voice, so nine different

social posts get generated for X and

LinkedIn and Instagram real scripts and

all this stuff. So that presumably the

thing that I was filming was for a

social video. Um so that was easy

automation land, but that is just one

example of like a proactive um

very

um

well-defined workflow.

What I think is more interesting for the

back half of 2026 is proactive of

undefined workflows. So, like AI is

probabilistic all the time and not

deterministic, but I want to take that

probabilistic nature of reasoning, like

the step zero of reasoning, and apply

that to the actual tasks that it takes

on. So, in order to do that, whether

you're talking to a human or an agent,

they have to know what's the goal,

what's the star, what's that vision.

They have to have access to tools,

permission to use these tools in the way

that actually gets work off your plate,

and a sense of what would normally

trigger that sort of action.

So, and I can I'm going to share

one thing [clears throat]

on on screen here, which is every single

day,

um let me just give me 1 second.

So, essentially, like

I want my whole company to be queryable.

I want AI to have context on everything

that's happening, and it dawned on me

that yes, it had access to all my

meeting transcripts, and it had access

to my Gmail and all this stuff, but

there was a lot that was not yet

codified, and it was things like

everything is becoming proactive, I want

to be more proactive, or

this client, they think that what they

need help with is workflows, you know,

under the CMO, but actually what they

have problems with is reskilling and

finding new roles for this one

department. So, anything that is not

codified inside of, again, meetings,

emails, whatever, or Slack, I have asked

AI now to prompt me every single day

with this, and you know, I got to put it

in my brand colors, and I didn't want to

have to think with, you know, maybe 10%

of my brain still working at the end of

the day, so I give it like a little

prompt. It reminds me to dictate because

that's four times faster than writing.

And so, I will bank these entries to be

like

you know, I talked to Greg. I feel like

the entire focus is on proactive agents,

proactivity,

um and flexibility. And I want to look

more into his three levels of employees.

And so, like I might do this for 5

minutes or 40 minutes at the end of at

the end of the day. I might do it

throughout the day. And then I just save

it out and then it's like it this goes

into my personal wiki. And all I want to

do

is make sure that the agents that are

working at that really flexible layer

where again, I am not managing them.

I am enabling them and they're coming

back to me with those escalations and

decisions.

I want to make sure that they have the

right context or else all their stuff is

going to be wrong. And and we saw this

in the beginning of our AI workforce

stuff. It was like, oh, I saw that, you

know, Greg confirmed that interview. And

it's like, no, Greg confirmed it, but

we're still figuring out dates and I'm

doing it over text and you know,

IMessage MCP broke since you can't see

that. So, there was a lot of stuff that

we had to continually fix and it took

probably months to get to where we are

now. But, we have Claude in every single

one of our chat channels. I had a very

weird I I have to send I have to show

you this.

Um

Let me just share my whole screen.

>> By the way, this So, the Brain Meets

Diary thing, so when you

>> Yeah.

>> when you you know, you add today Well,

you had like 86 entries, right? So, your

AI agents do all of your Does your Does

your entire AI workforce workforce have

access to that or just some? How do you

think about that?

>> So, great question.

Um essentially, my AI workforce right

now is one AI chief of staff with six

directors. Those directors are largely

over like business functions. So, one is

education, one is all the client work,

um one is kind of operations, one's

marketing, one product, and then Phoebe,

all these are named after Friends

characters. Phoebe is like the chief

dreaming officer who's just like being

wacky and weird in a corner. Um and so,

she's this is let me take another just

like moment here.

Um the reason that it took us months to

get to where we are now with our AI work

forces is that you have to take

stock of what assumptions you have made

about your work and how you are living

day to day and you have to be willing to

be like, "Oh, that thing that I've been

doing for almost 40 years, I feel like

we should change it."

And that's a really jarring

uh change to work, especially when

you've like been an overachiever, right?

I'm sure you feel this, too. And so, um

one thing that I am constantly having to

remind myself is we have all these

agents that do all these tasks and we

have skills and we have this and that.

And I have to remind myself that like

that is operating in 2015 world if I

give all of them job titles that existed

in 2015. So, if I name them CMO or chief

product officer and the person

underneath it is a front-end engineer

and a back-end engineer and all this

stuff, then it feels like I am operating

in 2015 org structure.

And one of the uh

most wonderful uses of free will

uh and just delightful things is going,

"Oh my god,

all of these employees basically cost

$0. And so, at the margin, I can hire

any flipping person I want to."

And so, I just wanted this weirdo. So, I

hired Phoebe as like a weirdo in the

corner

who's just looking at all these things

that we're working on and Phoebe acts as

this like almost end layer for things

that are getting generated to go like,

"How do we 10x it?"

Like I um

I joke, there's this guy David that I

worked with at Amazon who was one of the

reasons that I joined there and he is

like one of the most ambitious thinkers

I've ever met. And I joked that I would

pay him and I still I it's a joke but I

I

>> [laughter]

>> would pay him to do this.

Like um I want I wanted him to put me in

a room like Spanish Inquisition

Inquisition style with like a bright

light on my face and to ask me a

question. Like I was running a

multi-billion dollar business at Amazon

with 400,000 global startups running AI

strategy and if he asked a question of

like how would you do this and I

answered, I wanted him to just slap me

across the face and be like how would

you 10x that?

And that

>> [laughter]

>> I want a David um for for how I'm

structuring my AI workforce but I'm now

able to do that uh on my own. I'm sure

David would be disappointed to hear that

but it it's rethinking roles. It's

rethinking how you're spending um again

how how are you thinking about that

margin

um and so Phoebe is one of them that I

would have never hired in human world.

Um and Toby is another. I'll send you a

screenshot of my workforce but basically

Phoebe is that chief during officer and

Toby is Simon's assistant whose only job

is watching the AI workforce work take

down notes, what still has friction um

and who needs access to what. So going

back to your point of hey I have this AI

diary that I'm maintaining.

If we found that one agent did not have

access to this and Toby was like every

single time you keep correcting this one

agent's output have you thought about

giving your agent access to this? Now

this is just context that lives on my

desktop so any of these agents can

really see it. Um but if it was a

specific tool um if it was a specific

folder that is outside of normal cloud

land um that I try and have hard rules

on then I would absolutely use AI as a

means of figuring out those friction

points to then expand.

Um yeah.

>> Question on designing your actual

workforce. So, I agree by the way. I

think like

um you have to think about like

how do you create an

an AI native workforce like without job

titles from pre-AI native land. So, I

agree with that. But, like

tactically, if I'm a founder, like how

do I It's so much It's so much easier to

be like, "I need a CMO. I need a CPO. I

need this." So, how do

>> I think everyone should start there.

>> Yeah.

>> I think like the the the starting point

is What does it feel like to work with

one agent? After that, I would say,

"What does it work What does it feel

like to work with one agent who is doing

things on my behalf proactively?"

Then I would say, "What does it feel

like for two agents to work together on

a task or for one to direct the other?"

Um like one to route to the other.

Um and then then I would say, "Okay,

what does a workforce look like and how

do all those things interact?" And I

have, you know, like a mission control

where I'm seeing how all this stuff is

moving around.

And then you go, "Oh, now I understand

how they're trading notes. Now I

understand how context is passed. Now I

understand that things have to run in

parallel. Now I have to understand

um that that this agent actually didn't

need access to these tools. Now I

understand that that agent can run off

of a smaller model. Like not everything

needs Opus. All of my, you know,

sub-agents are like Haiku and Sonnet."

So, all of that is in the discovery

phase of building out the AI workforce.

I think start with traditional job

titles.

>> No, I just I was thinking to myself like

I wish it wasn't that hard, right? Cuz

like it it

it does feel like there's like a ramp up

time to actually get to a point where

you have an AI workforce that's working

for you that is efficient. And I think a

lot of people the what happens is like

they try, they fail. And they're like,

"This isn't for me." or "The models

aren't good enough yet." or and and you

know what I mean?

>> Yeah. So so here's here's my

take on that.

Um

I think that you can spin up a workforce

with one prompt. Right? Like I've shared

this prompt publicly.

Um you can just prompt and say, "I am a

founder. I am building an AI personal

shopper. My team is three humans. Here's

what we do. Here's where we're based.

Here's our goal." whatever. You can say

that and just say, "Interview me. Um

we're going to build out an AI workforce

together. Something that runs more

efficiently and achieves my goals of

saving at least 5 hours a week. Um

uh capping my my meetings to to 15 hours

per week and make sure that I get into

my capital raise by October." Right?

Like you can you can do that in one

prompt and have it interview you, and

then you have a workforce.

To go from uh yes, all these agents

exist and they all have markdown files

and they're doing some stuff to ooh, now

it's at the 90% plus level and ooh, I

needed this extra little context with

this diary and mhm that role isn't

working. I'm going to switch it.

That is all going to come through

iteration cuz it's so specific to each

person. The advice that I would give is

stop relying on only yourself to find

these blockers. Like AI as a watchdog is

one of the best use cases that exist

right now and almost no one is doing

this. So like having an AI watchdog in

Slack to catch for duplicative work or

having an AI watchdog on your calendar

to see when there are conflicts or an AI

watchdog over your meetings just to see

where disagreement is happening.

Like 10 years ago, I remember working um

this was at a at a large-scale

enterprise.

Um but we were working on like comparing

contracts. Right? It was like before the

edit, after the edit. And it was like

compare and contrast with AI.

And 10 years ago, that was like the

greatest use case ever.

And yet, no one today is using AI for

this like weird cross-functional gap

analysis

at a more advanced level than we would

have done 10 years ago, and it's still

just like such a meaty use case. I think

Claude Tag is a big help here. I think

it's a mess right now in this exact

moment that we're recording this. I

think it's a mess to set Claude Tag up,

but I'm sure it'll be fixed by the time

this comes out. I've also set up my own

Claude code to come in. I have a Slack

channel

that is called Loop Alley. I'll send you

a screenshot of non-private information,

but it is called Loop Alley. My freaking

human team can talk to my AI workforce

in that Slack channel.

So,

there is no ceiling to this stuff. Like

I'll have a teammate who like if I'm in

private emails with someone, that the

teammate will write into the Slack and

go, "Hey, did

Did that large financial services client

like did they respond to Ali's email?"

And my workforce will respond back to

that person. And that person will not

have to wait for me for 5 hours to get

back to them.

So,

that sort of thing, the ratcheting up of

how advanced your AI workforce can be,

how multiplayer it is, that's going to

take time because people are still

figuring out best practices now. Things

are not easy to set up right now, but

that baseline of hey, interview me, I

want a workforce, I want something just

doing stuff for me at a high enough

level, you can set that up and connect

into tools in under 3 hours.

>> The other thing is because a lot of

people are not doing it, that's the

arbitrage opportunity, you know?

>> Yes.

>> So, it's kind of like

it's kind of like it's stick through it,

optimize it. I'm curious actually from

your perspective like um you know, what

are opportunities

that are you seeing that people could be

you know, building, you know, making

money, type that sort of thing. I'm just

curious, you know, what comes to top of

mind.

>> I think so certainly

I think AI workforce first of all, like

of all AI users, if you look at the

percentage of people who are paid AI

users and if you look at the percentage

of those who are using things like Codex

or Claude code, it is minuscule. So

already, if you're just trying to be in

the top like 1% of AI users and you're

using the stuff and you've built out

even a basic workforce

you're already top 1%.

Probably top 0.5%.

Um getting it to that advanced level I

think is absolutely arbitrage because it

feels like I'm operating a company of a

thousand people and not my small, you

know, scrappy Gremlin group. Um that is

still absolutely one. I think the second

that um

that I would do is that AI is a watchdog

over any single thing that I am normally

tracking. So maybe it's and and I don't

just mean visibility. I think dashboards

are dumb, but I want visibility with

anomaly detection or insights or

something. So don't just tell me what my

social media following is or views or

whatever. Tell me what are people

talking about? What are people best

reacting to? What is not performing

well? What should I do tomorrow? Write

me a script that helps me for that. So

kind of this AI is a watchdog but with

insights into action I think is the

second. And the third that very few

people are talking about but is probably

one of the biggest arbitrage

opportunities because of how good the

models are now

is to instead of building out the thing,

build the factory for the thing.

>> What do you mean by that?

>> So let's say that you want to build um

a product and we just released um

there's something called the AI first

index that I run with all of my Fortune

500 clients where I interview their

executives and I evaluate how AI first

they are across like 16 different

dimensions and all this stuff. And we

decided through a combination of humans

and AI to create a product um, for the

public to be able to benchmark

themselves on how AI first they are as

individuals and as a company.

In that process, I could have done one

of two things.

I could have gone to CloudCode or CodeX

or anti-gravity or whatever. I could

have gone to any of these and said,

"Hey, I want to build out this thing,

interview me, you know, look at my my

um, AI first index reports that I've

used with previous clients,

um, find every single workshop I've ever

done with clients where I mentioned the

AI first index, whatever. Do that and

build out the product and then we

iterate for several hours, days,

whatever until something is perfect and

we release it." That is option one.

Option two is realizing that that's

probably not going to be the only

product you build or will not be the

only iteration of that specific product

that you build.

And so it's it's like going one level up

in abstraction. It's like what dev tool

companies did for engineering, but

you're creating

dev tools that level for yourself.

You're going to like the kernel level

for yourself.

Um, and so you're moving down the stack

for yourself.

And instead of just building that

product, we instead built out a mini and

very beginner software factory.

Where we're building out primitives

obviously we have to deal with login,

obviously we have to deal with payments,

obviously we have to deal with social

sharing.

Um, we have to deal with writing

newsletters to promote these things. And

so you end up instead of just building

that one product, you go,

"There is going to be a flywheel that

comes out of this. There's going to be

explosive opportunities that comes out

of this. Why not take advantage of that

now?"

And so it's like a measure twice, cut

once kind of thing, but the measurement

is building out that foundational layer.

So that the next product that you build,

the next iteration of the AI first index

or whatever you're building out, is so

much faster, so much better, so much

stronger.

Um and so

we're we're we're building these like

loops, these optimizing loops again that

aren't super autonomous and are very

heavy-handed with humans.

But that is the arbitrage opportunity on

products that are revenue jet like

that's already that product's already

profitable.

And now I have the ability to build

endless products that are profitable at

faster speeds than I built the first

one.

>> That's crazy. That's absolutely crazy.

And like no one is talking about this.

>> No, it's the it's the the dark headless

factory. Headless like AI headless,

[clears throat] not you know.

Um but that is that's what I want. I

want that I I want to learn through the

mess. Like we had a webhook issue,

whatever. Like I want to learn through

that mess and then I want to never make

that mistake again.

And so you're you you have to think

about how this factory works, not just

for product building, but you know,

maybe it's for how you want to run your

content engine, maybe it's how you want

to deal with net new leads. Like think

of the factory behind the one singular

task instead of the one singular task

itself. That is one of the biggest ways

to rethink work in the AI age.

>> What's uh

what's Ali Miller's current POV on, you

know, software

you know, the SaaS apocalypse and

software, the value going down, down,

down? Like in a world where everyone

could create a software factory.

>> Also, I like I wish I had an agent that

was yelling at me about my posture. So

like maybe I'll I'll create a new one

[snorts] for that. As I as I realized.

Um SaaS apocalypse, I think mediocre

software is dead in several years. And

the reason that I think it's actually a

longer timeline than most people are

predicting is because of what I shared

about like how often people are actually

using this stuff.

So, you could go into one of the most

AI-first, you know, banks or AI-first

software companies. And if you ask them,

"Have you rebuilt DocuSign? Have you

rebuilt parts of Salesforce? Have you

rebuilt all these things knowing that

you can?" They would say something like,

"No, because we're already so bandwidth

constrained." Or, "No, because we've

prioritized this other thing."

Um as long as we are still bandwidth

constrained, and as long as there are

still

billions of people who have not used

these sorts of tools, you're not going

to have

mass

adoption inside of the enterprise of of

the replacement to SaaS.

Does that make sense? Like Like if it

continues to take

I don't know, 100 hours or something to

rebuild something at the scale of a CRM,

companies that only have people who are

sitting there and can work for 100 hours

and who know how to do this are going to

be able to take advantage of it. And

it's only going to be when that drops

down to like under 3 hours and is a fun

click and drag interface, which I would

even argue and say Replit lovable or not

at that level yet, right, for that

complexity of software,

you're not going to see uh a

high-complexity

enterprise-grade

highly secure SaaS

do that.

>> Also, people don't want to maintain that

software, too, right?

>> Oh my god.

>> People don't People are willing to pay

someone else to maintain software.

>> Absolutely. I I built an app, this was a

a year and a half ago or something. I

built an app that only lives on my

desktop that allows me to like better

manage photo stuff. And someone

yesterday uh brought this up in a call,

and I was like, "Oh my god, I have

enough just for this." And then I opened

it and it was aired out. And I'm like,

"I don't want to deal with this right

now. Like, this is

>> [laughter]

>> not at all what I want to do." So,

you're totally right. The the

maintenance is rough. I think like Boris

kind of describes one of the like future

employee types is just like the

maintainer. Um but I I have a really

hard time seeing mass SaaS-pocalypse

until the ease of making prototyping,

making, customizing, and maintaining,

and securing

um is

is at like 95% plus.

>> I mean,

even even in a world where there's the

maintainer,

if something breaks and you're an

enterprise, you want someone to call.

You want to go into someone's office,

right? Like

>> Yes. You also want someone to blame.

>> You want someone to blame. [laughter]

Right on.

>> That's an important piece. I think a lot

of people are forgetting that like

the question of is AI going to replace

this, this, this, whether it's a task, a

job, a company, a product, something, um

often I am asked the first question I'm

asking myself is who's liable now, who

would be liable in that other world, and

do I think that that trade-off is worth

it right now? Like, I work with Fortune

500 CEOs every single day.

They No way.

>> [laughter]

>> No way. They want to be able to call

because they want someone to unblock,

they want someone to secure. The other

thing is that um let's say that um let's

just say it's a Salesforce example, and

that you could build a shitty CRM or a

simple CRM or something that's just

running on your own, um but Salesforce

has relationships with all the AI labs.

They are, you know, getting into early

testing. And so, by the time a new model

comes out, you are facing it as a day

one person. They're facing it as a day

30, maybe. And so, you're also going to

be on a very big lag.

And so as you're thinking about that

cost trade-off, I think in addition to

all the things that we just talked about

with enterprise grade security and

maintaining, whatever, you just also

don't want to experience that lag.

Like we're moving to a world where

being fast to the punch and getting a

30-day, 60-day, 100-day leg up on

someone is going to be massive for

business.

>> What about for consumers? So like I I

get that like an enterprise, you want

someone you can speak to and and you

want security, but for consumer it's

like like for example, your app idea

around,

you know, let me know when my posture is

bad.

>> Yeah, which

I'm just going to keep

>> [laughter]

>> Here, I'll move I'll even move the

camera up. Okay.

>> By the way, I also have horrible

posture, so

>> Okay, well then let's build a product

using my phone.

>> Yeah, exactly. And and it's like, okay,

let's say you build a product and I

build a product. It's like

uh

you know, ultimately may the best

product win.

Um but like

>> Yeah, hopefully.

>> Hopefully.

>> I I don't think that's ever been the

case though.

>> That's right. I mean, the best the best

songs aren't on the Billboard 100, you

know, like in the sense of like the

marketing Yeah.

the the promotion of a of, you know,

piece of IP is really what drives a lot

of awareness and

and

>> also an arbitrage opportunity. Like you

it's almost kind of exciting that it's

not only based on code

or design for who wins. It's like kind

of nice to know that if you're someone

who's really personable, that you can

get a leg up if you're able to like open

doors that other people can't.

>> Exactly.

>> Like on the one hand you could say it's

not fair because it's so subjective, and

on the other hand you could be like, oh

yeah, but if I lack that one skill or if

I'm not the best in class at that skill

and I'm just kind of passing muster on

that skill, I still have a chance.

>> Yeah.

Yeah, so

I agree. So like when people say, just

to like sum this up, when people say

like

>> Yeah.

>> software is going to zero, on the

enterprise side, we both agree like

yeah, some software might go to zero,

but it you know, you want someone that

you can speak to, you want

security, you want something to maintain

it. On the consumer side,

um what it feels like it's sort of

shifting from science to art. And now

the people that are going to win are

going to be the more creative, maybe the

video first people, the people that can

like understand how to create Instagram

reels that a posture app can go viral,

and the code is actually going to matter

a lot less, but the amount of

opportunity that exists both in

enterprise and consumer,

to me couldn't be higher.

>> Like I So, I think a lot of people will

say the phrase like look for the

bottlenecks and solve the bottlenecks,

and I always kind of disagreed with or I

don't think it's fully complete. The

phrase that I say is like look for the

bottlenecks, then evaluate the value of

fixing those bottlenecks, and then pick

the bottleneck that is high value to

fix.

>> Mhm.

>> And so, if right now the bottleneck is

not on writing code, and the bottleneck

is not on coming up with good design,

but the bottleneck is getting something

from a local HTML file into like an

actual iOS app, then that might be where

you spend your time.

Or if the bottleneck is that no one's

really figured out how to get

um you know, stronger word of mouth and

referral codes, and like that's still

kind of messy. Um and I and I know this

as a product maker and advisor,

whatever, like that is still a messy

spot. So, like

maybe if you fix that, your your uh

whatever they call it, like the the

covariant, the word of mouth covariant

thing, um

um could be above one. Like that is what

I would be

spending my time on. Finding the

bottlenecks and finding what is still

high value. I think video creation, no

matter how much AI is helping me edit

or, you know, edit the script or

whatever, it is still a slog to be able

to make video. So, that is still a

bottleneck and it's very high value. Um

but, you know, people

in the B2C space, I'm sure can think of

a lot more. I don't know, I just think

of like certain B2C products that I use

and I'm like, why did I pick it? Um

I use WhisperFlow every single day. I

don't like their mobile experience at

all, but I still use it. Um because the

value is so high. Have I seen a single

video about Whisper Did I see a single

video before I started using it? No, I

now see them, you know, everywhere, but

>> Could it be subconsciously though? You

like see their brand places, like you

might be watching I don't know, you

know, Chris Williamson and then they

sponsor Chris Williamson and you kind of

you kind of just see it, you know?

>> Yeah. I think like influencers still

have a ton of sway here. The rise of the

B2B influencer, which like I feel like I

was one of the first [laughter] and it

is

it's so amazing to see more people

creating business content, but that is

still a bottleneck um in in building

like B2B trust.

>> Right.

>> That is a massive bottleneck and so

finding creators that can help you

there.

Um

but I think B2C has a ton of

opportunity. I worry um if you look at

the YC splits right now,

um when I was working with YC when I was

at AWS compared to now, the ratio of B2B

versus B2C has skyrocketed.

Like there's just not as many B2C

companies in these incubators getting

built.

Um you could either say when they're

zigging, I'm zagging and double down and

do a B2C thing. Like there was this

woman who created an app. She's never

coded a day in her life. She created an

app that takes a few photos of your face

and she takes that and creates an a

model of your face and gives you like

aesthetic photos that are like you in a

grainy rainy day riding a bicycle or

whatever.

She had 300,000

users out the gate.

Like there's still a lot of opportunity

in B2C even if the big incubators are

seeing that activity less.

Um

and so maybe that's another opportunity

for people to explore.

>> Well, yeah, and I think like, you know,

we we've we've been talking a lot about

agents and I think there's just an

opportunity to create agent-first

version of some of our favorite apps.

You just like look at, you know, a bunch

of different B2C apps. Just to go look

at centurytower.com.

Um not affiliated, but you can just see

like what's charting and what are people

downloading and it's like, okay, in a

world where superintelligence is now on

tap, how can I make an AI-native version

of this?

Um

or undercut, you know, from a price

perspective or just drive more value.

Like there's ways there's now like

opportunity to

to to

to enter some of these markets.

>> I I completely agree with you and I

think agent-first software is absolutely

one. Um two things that I actually think

are really interest or maybe three by

the time I get to it, but interesting

research avenues to learn more

opportunities like the one you just

mentioned. So, one, YC posts

uh videos on Instagram for what type of

applications they're looking for and

agent-first software is one of them. So,

listening to what YC is asking for,

assume that they are already thinking 18

months out.

Um so, that's definitely one arbitrage

research opportunity.

The second is Matt Van Horn's last 30

days research skill, which is just

amazing. I've like inner

um I've integrated that with my like

Claude wiki. Love it. Um and the third

is

>> Wait, can you tell people I've had Matt

on I've had Matt on the pod, but just

quickly like what is it and why why do

you think it's chef's kiss?

>> So, there are a lot of public skills

that I think are done by geniuses in

their space. One that was kind of first

out the gate or one of the first out the

gate that is made by a lovely man named

Matt Van Horn is {slash} last 30 days

and it's on GitHub. You can just grab

it. But, it is the ability for AI to

figure out today's date, scan the news

of the last 30 days, but scan it in

interesting ways, synthesize it in

interesting ways, and just fan out crazy

amounts of agents in parallel to be able

to bring it back to you. So, as I'm

thinking about, you know, if I'm going

into a company and I'm running a

workshop for their 200 executives,

I don't know about the insurance space

as well as I should. And so, like if I

need to quickly get spun up on an

industry, I'll use it.

Um or quickly get spun up on a specific

company, I'll use it. So, I use it

there. But, for this in particular, you

could just do {slash} last 30 days and

then say like startup ideas that could

be built by someone with the following

background or the following skills or

um had the last three jobs of this this

this. Like, use it in interesting ways

to see how you can carve out a new path

that people are not doing.

Um the third, which I have access to and

I think there are public avenues to get

it,

um is that I might Let's say I I am at

like a CMO summit. And so, every single

person in the audience is a CMO.

I can hear the types of questions that

they're asking, right? I can hear the

the fear zones that they have. I can

hear questions that they used to ask 3

years ago and are no longer asking

today. And so, finding companies,

people, influencers, creators, Gregs of

the world to like follow to hear the

inside scoop of what these people are

thinking of. Like I can tell you that

CMOs, all of them are asking about like

how do I get discovered by agents? How

What is the agent for shopping

experience look like? What is brand

consideration in the AI age look like?

You know, all all of that is being

considered right now by CMOs, but it is

often coming from a place of fear that

they are worried that their business is

going to be

depleted, that their pipeline is going

to be crushed in 2 years if they don't

figure it out now.

So, figuring out paths to find those

fear points would probably be the third.

>> I love it.

Allie, anything else you wanted to

cover?

>> I just want to screen share the insane

Claude reaction because

this

um and this is me also cursing at

Claude, but whatever.

So, I wrote I wrote um a a not super I

wrote a not super nice thing about

Claude in one of our Slack channels.

[laughter]

And this was like late at night and I

was just like getting it out there so I

could talk with my team about it later.

And all of a sudden there was an emoji

reaction of a salute.

And I was like, I don't think a single

person on my team has ever used a

salute. And I hovered over it and it was

Claude. I was like, [laughter]

"What are you doing?" And so, I wrote

back to it, "Did you just

you know, emoji react like is that you?"

And Claude was like, "Yep.

That was me. I'm here."

And I just if there's one thing that I

want people to to think about, it is

the leaning into the weirdness of what

it looks like to have not just an AI

workforce, but to have a multiplayer AI

workforce that other humans can chime in

on

and have it be proactive.

Right? That is absolutely second thing.

And giving it that flexibility to more

roam free. Um and the third is what it

actually looks like for a teammate or a

system to up level, whether that's in

dark factory type space or just

answering better questions inside of

Slack. Those are the things that I would

be considering and don't be

scared like me if Claude emoji reacts to

one [laughter] of your messages.

>> Yeah, I mean it's

You know what that is like? It's kind of

like um

you know, it's a winter day in New York

City and for some reason it's like

middle of February and all of a sudden

it it it feels like summer. Like you

know, there's like random hot days and

you're like, this is amazing and you're

like 90% excited but like 10% frightened

cuz you're like, it's not supposed to be

>> Yes.

>> It's not supposed to be so hot now. That

was kind of like

>> are always so You're like a genius with

analogies. Yes.

>> That's what it's like. It's like

You

and that's 90% cool but 10% frightening.

>> Yes. Yes. I'm like I'm like still going

to continue to try and lean into that

weirdness and find ways that I can like

take that weirdness and use it to my

advantage. Um

but

I'm going to keep that fear next

[laughter] to me so that I don't lose my

mind.

>> 100%.

>> Yeah.

>> Uh

I hope people enjoyed this episode as

much as I did. Ali, I absolutely love

chatting with you. You're one of my

favorite people to talk to. Please

comment on YouTube to let just to just

to hype Ali up honestly and have her

hopefully come back on the podcast

again. Uh Ali is a a follow. I'll

include where you can follow her on her

socials in the show notes and the

description.

>> Yeah, Greg, thank you so much for having

me. I My hope is that every single

person got the tactical things that they

need to just like immediately

immediately take action on this. If

anything was not clear, let me know. I

am going to like jump on and help

people.

And Greg, I will absolutely come back.

You're one of my favorite favorite

creators. You can always call on me.

>> I appreciate it, Ali. I'll see you next

time.

>> Sounds good. Bye.