How AI Is Rewriting the Rules of SaaS Monetization

So today, Scott and Martine will be

discussing how AI is rewriting the rules

of SAS monetization. Um, so we'll kick

off on how that shift is happening. Uh,

we'll go into the investor lens on get

Martine's in perspective on what great

monetization looks like and the patterns

he's seeing across the market. From

there, we'll talk about how AI companies

can better define and capture value,

turn pricing into a true operating

model, and then we'll wrap up with some

practical advice on how startups can

scale it the right way. And with that,

I'll hand it over to Scott to kick

things off.

>> Oh, hello. Uh, nice to meet you or nice

to meet everyone here. Um, super excited

to welcome Martin on to the webinar. Um,

Martin has been on our board since, you

know, for at least four or five years at

this point and is one of the kind of

key, I would say, like voices on how

infra and AI are intersecting in this

current moment. and I think he has a ton

of very interesting thoughts around

pricing and and thinking like an

investor in this moment. I think there's

like a lot of very interesting kind of

rabbit holes we can go down and I look

forward to going down them with him. But

with that, maybe Martin do a quick intro

for folks and then um and then I'll jump

into a little bit of a vignette on on

metronome and then we can jump right

into the conversation.

>> Well, first off, that beautiful image is

definitely pre-COVID. That was probably

10 years ago. So

>> pre pre- startup maybe

>> pre- startup. So, so I appreciate you

putting it up there, but uh you know

things have evolved then. So, hey,

listen, I appreciate the time from

everybody. Um, so I'm Martine, I run the

infrastructure fund at Andre and

Horowits. Uh, I did my PhD at Stanford

in computer science in infra in

networking. Um, and then I I started a

company and uh I was a portfolio founder

for Andre Horowitz. So, Ben Horitz was

actually on my board. Um, we sold that

to VMware. We were kind of the first

unicorn exit like coming out of the

recession in infrastructure. So that was

a billion dollar exit and by the time I

left VMware that that year that I left

it was a billion dollar business billion

dollars a year and then I joined 10

years ago in Jason Horitz to run the

infrastructure fund um and so I've been

here for 10 years I sit on a bunch of

boards you know names that you probably

know five cursor um a lot of the big

labs and so I'm very very excited to be

here.

>> Very cool. Um awesome. Well let me um

I'll do a quick intro about metronome

and then we'll jump straight into the

conversation. So um for those of you who

don't know uh Metronome we power

essentially monetization infrastructure

for some of the fastest growing AI

native companies and also some very

large uh enterprise SAS businesses. Um

our core value roughly is we make it

really easy to price and package however

you want and make sure that you can get

your AI products into market as quickly

as possible while making sure that

everything is exactly accurate behind

the scenes. Um, and I think the other

key thing I will say before I turn it

over to Martine for some questions is

that one of the key things that we

understand here at Metronome and we

believe very deeply is that like right

now there's a lot of talk about usage

based billing, outcome based billing,

all these like kind of billing models.

But the reality is that the problem that

billing represents is actually much

deeper. It's actually much more of a

workflow problem. It's much more of a

customer experience problem. It's much

more of a like how do you architect a

business around a monetization model

rather than just like okay how do I

actually make sure that like my invoice

looks correct at the end of the month

and so we built metronome to kind of

service all of those needs. Um and

actually that kind of takes me to the

the first question that I have for our

team and one that I think is really um

interesting is I think you you know you

built NERA in the late 2000s early 2010s

and then you worked at VMware. Um I

think everyone would say that like AI is

a wave and I am curious how how uh how

you view the monetization like where we

are in the monetization journey of this

AI wave like you know I think I think

one of the things that I would observe

about social media is like Facebook for

a very long time didn't make any money

it just ran at a loss and so actually

where are we in the like maturity curve

of AI as an industry in terms of

monetization and how does it differ from

prior generations that you've seen

>> okay so um it feels Like every platform

epoch, we have to think about kind of

pricing and billing and how we value

software, right? And so when I started

my company in infrastructure, we're

still in the perpetual days, right? So

you'd sell someone like software or a

box and they pay you some money and then

you'd leave. You'd have to come back in

a couple years and try and sell them

another box, you know, and then

hopefully like you had a new better box

or whatever. And uh yeah so you know

from when we started in this era to to

um you know after I left the industry

had gone through kind of the SAS

revolution and gone to recurring and the

view was like you were going to deliver

software as a service and then you know

we're going to build from recurring and

you get all of the upgrades that you

want from that and so you know I I I had

a number of learnings from those one of

those is like these tend to be customer

aligned like you do this because it's

actually better for like the user better

for the customer but the second thing is

you have to like redo your entire entire

business this way. Everything from go to

market to sales comp to how you value

features to I mean like it cuts through

the entire business and we actually had

to come up with an entire new playbook

um in in order to to to deal with

recurring and I know that you know Scott

you were in the industry as well. I'd

love to kind of hear your thoughts on

that. Um but you know going forward so I

think we're having a a similar shift now

as we go from kind of traditional

seekbased recurring to usage based there

there was kind of smatterings of that in

in infrastructure before and I'm an

infrastructure guy like you think kind

of like cloud cloud credits but now it's

everything um and this one is much

bigger much faster and much more furious

than the one before and it just tends to

be a lot more strategic because it's so

closely bound to the actual usage and

value that it gets. And so again, we're

having to rethink everything. I think

we're in quite early days to your

question. I actually think I mean,

listen, in some ways, I feel like I'm

talking to Fermy about physics. Like you

actually are one of the world's experts.

And so I would really love for you to

kind of dig in and I I'll love to like

kind of sprint with the anecdotes that I

see from the board for you to dig in and

how you're seeing it,

>> you know, the the key considerations

each company should do. But I will say

from my perspective as a board member,

as an investor, you know, if we're in

the AI space, a company is not

considering usage based billing. um you

know we

it's a conversation I will force very

quickly because it's so it is so fun it

is so fundamental to the business um and

it is so strategic so

>> yeah totally actually so you know I I

think you mentioned it but like before

this I worked at at Dropbox and in a way

I think Dropbox was one of the vanguard

companies for this subscription business

model um so I'll give a little bit of an

anecdote and then I turn the question on

you so um so one of the things that I

observed very quickly is that our

monetization model at Dropbox for the

team's product scaled with the number of

users. And so like the number of people

in your company like you know was your

essentially your value metric or how

your how your monetization scaled. What

this led to was like wall-to-wall

selling where you would basically try to

get as fast as possible to distribute

across as many different people kind of

logging into the network of your Dropbox

um instance. and and um and so the one

of the interesting observations of that

business model was that we spent a lot

of time reducing the friction for

bringing an incremental user onto the

platform both in the web UI but also

support for every operating system and

making sure that every operating system

kind of like continued through end of

life and so you know I um there's a lot

of like really great Dropboxers who

spent like endless years making sure

that like every version of Windows like

was supported to like the ninth degree

because that was how they monetized So

that was like uh you know uh the

monetization model dictated the product

strategy and those two things were

deeply tied. Um one of the interesting

things though is how much a user used

your product had nothing to do with

monetization. Like it's like it's like

if a user stored a trillion gigabytes or

one gigabyte or zero, you actually made

the exact same amount of money. In fact,

your incentives were kind of weirdly

perverse where your cost scaled as the

amount of data that is stored, but your

you make you recoup nothing. And so that

what that led to is this kind of uh you

know it's like there's a reason why

Dropbox is famous for its like virality

and network like essentially expansion

but also you could say the value no

offense Drew if you're listening hasn't

really changed in the past 15 years for

any one user in that product and that's

actually I think a a good to me that was

like an object lesson in business model

is destiny and so I would love to hear

your take on some of these more AI

native companies and how that how that

intersection between products strategy

and business model is playing out in

these companies that are based on top of

these LLMs. Like I would be super

curious to hear you just like reason

through it or maybe take an example or

an abstracted example, but I would love

to hear your thoughts on if you have an

equivalent story for like some of these

AI native companies.

>> Yeah. So, um I think the thing that's so

remarkable about AI is how much it

actually changes the cost structure for

the user. There's actually two ways. One

that's not that's kind of not answering

your question. The other one that is

answering your question. The one that's

not answering your question is is I

don't remember in the history of the

industry where you could take a a

company and actually pour money into it

and it could make use of that money

early on, right? Like in the past, like

you'd pour money into a company, they

hire a whole bunch of engineers and then

they would like build something really

complex and it wouldn't work, right? Um

but with AI, you could pour money into

these companies and you train a model

and that model is incredibly useful,

right? And so in some ways, companies

have become these receptacles of putting

large amounts of money and that's

actually turned into revenue on the

other side with with small teams. So

it's a very very different construction.

But then you look at the other side of

that. So what do these models mean to to

to users? And again I don't remember the

last time that we had a product that was

so close to actual value creation. So my

favorite example is uh is image. So I'm

not going to talk to LM. I'm going to

talk something else because it's so much

easier to reason about. Right? So uh

let's say that I wanted to create a

Pixar image of me, right? Uh in in the

preai world like I don't know like I

can't draw. So what do I do? I go hire

somebody and like listen what it said it

takes them a week because of their

backlog and then uh what a h 100red

bucks let's say right so if you compare

that to the inference cost what's the

inference cost 1/100th of a penny right

so you've got four orders of magnitude

difference right so you have got this

very very clear dislocation

between um what you could before and

what you can do now with a model right

and so so because you're bringing the

marginal cost down so much there's very

direct value to the user and this kind

of necessitates different pricing

because you know you could use it for an

hour and you can get four orders of

magnitude you know more value or you

could use it for a week and and I think

necessarily the trade-off there is

actually the cogs end up increasing for

these companies right I mean you have

you've created kind of like a nuclear

reactor that costs some amount of to run

and so on one hand there's tremendous

amount of value on the other hand like

you know like it actually costs to run

these things. And so it just kind of

necessitates a different model both, you

know, so you can kind of um uh you know,

align yourself with kind of the customer

value, but also so you can make

decisions on how you run your business

on like how much do you want to like

optimize for margin for distribution or

things like that. And so it's just the

symmetry between the value created

versus the cost of service is so much

different now than it was during the SAS

era. It's just really require these

different business models. So that's why

I think so many of these are going

towards you know usage based um billing.

>> Yeah. Well so actually that kind of

dovetales with and because you mentioned

cost I have to ask the question but

sometimes people ask and they say okay

like you know they look at you know I I

don't even know where they get this

information but they're basically like

this company is running at a loss like

it's like a margin negative product. I

guess you sit on the boards of some of

the fastest growing companies in the

history of the world. uh talk to us a

little bit about our behind the curtain

like what is actually like you know is

that true? Is this stuff like running

cost negative and then or margin

negative and and um and and talk a

little bit about like that that kind of

meme and like why you know it either

makes sense or

>> Yeah. Yeah. Yeah. Yeah. Okay. So so

before I was saying like listen like you

have this kind of nuclear fusion that

you've created and that adds tremendous

value and so that it's just like a it's

like a different set of primitives to

work with as you build out your

business, right? So you can you can make

all sorts of choices that you want,

right? Um I will say um and so so now

like you know people make different

choices based on that. Sometimes they'll

go for distribution, sometimes they'll

go for for margin. Unfortunately, we as

an industry every time that there is you

know a technical epoch. We have this

weird shud where we are like oh like

these companies are margin negative or

whatever and we have that conversation

without realizing that these are board

level decisions where often they're

quite intentional to do that. And the

great thing about userbased building and

this is a knob that you can decide to

turn up or down, right? So, so let me

just go ahead and talk through this. Um

um

many of the so I'm not going to say any

specific companies but we we're in all

the largest AI companies. So many of

these companies um if you segment the

business and you're like okay I'm going

to look at just the enterprise revenue

or like the enterprise and teams revenue

they're actually beautiful margins right

um they're gross margin positive

normally not software margins but let's

say 30 to 50%. They're actually great

examples of AI companies at scale that

have been profitable from day one that

are at hundreds of millions of revenue.

So there's clearly proof points of like

very profitable. But there are other

ones where if you segment it on like

let's say you know enterprise and teams

is cross margin positive but if you look

at like free plans and you know maybe

pro plans or the TPM tier um it's

negative or the entire thing blended

pencils out to zero. And so okay so what

conclusion can we draw from that? Well,

if you have a situation like that, I

guarantee because I'm in these

conversations, the board was like,

"Listen, what's more important to us is

distribution, not margins." And what's

amazing about usage based billing, by

the way, what's amazing about this epoch

is like you actually have that level of

granularity and control. Like, you can

view pricing very strategically, much

more than during the SAS days, by the

way, because this stuff is just so

useful. It kind of solves retention

problems. So, you you can at the board

and say, "Listen, we're going to kind of

give away this many tokens or this much

stuff over here." Um, we're going to

pencil out blended to to zero. Over

here, we've got great margins. At any

point in time, if we want, we can kind

of go ahead and like limit the free tier

and go to a margin positive. And so,

again, you know, this is that this kind

of silly stupid meme that happens every

epoch. I mean, I mean, listen, it

happened during the com days when they

actually weren't making any money.

>> Here we are making any money. Um, you

know, it's being applied now, but like

the reality is is it just really shows

the flexibility that you have with these

business models. Yeah. And I I think

what you said is really important again.

I think it's like um

>> I'll bring in a Dropbox analogy because

I just know it so viscerally. It like we

offered a free tier, okay? Like that

obviously cost us money. Like literally

definitionally, right? And and yet it

was like it's like if you just looked at

the free users like this product doesn't

make any money and yet the company was

free cash flow positive for like you

know 98% of its existence. and um and

and and and the reality is is like the

free tier in that case is serveing as

like lead genen for companies that are

actually getting super deep value from

it and then scaling it

>> and and and and oh by the way like what

a great lead genen mechanism it is like

100% how many companies have just like

literally pissed away money on like

marketing that doesn't do anything right

like like having having come from the

SAS era let me tell you like like the

most common discussion on the board is

like we have all these MQLs that's great

you're beating your revenue none of them

are converting to SQL so none of them

are are like actually converting to to

to to wins right so there's this massive

massive disconnect between marketing

general demand gen and then lead genen

and then closed and you would spend

millions every year on this because

because the marketing was not aligned

with actually product and product usage

it was a totally different motion and

here you actually have like this magic

thing that you can give to people that

actually draws them on the product

journey from day zero. And so it's

almost like the most efficient marketing

budget that any company has ever had. Um

and then and literally you just make a

decision to what extent like you're like

okay I'm going to like grow distribution

versus I'm going to do uh conversions

into sales. And so I just think like

this kind of mimetic view on margins is

just doesn't understand what's actually

>> happen. Exa Exactly. And it's even more

it's like you know the other thing

that's true is that when you capture

these users like attention is zero sum

and so if you can get them on your free

product then basically you are training

them on your model of the world and then

getting them in your ecosystem. And so I

think it's like again it's like it's

like um you know turns out big

businesses doing hundreds of millions of

dollars you can find a cohort of users

that are margin negative. Okay cool

great thank you for that useful

information. I think like the reality is

that like I yeah it's like um these

these businesses are like printing cash

on day one. So actually that kind of

leads me to an interesting question

because you know I did a startup in the

early 2010s and it was like you know the

company

>> getting to 100 million was like this

insane milestone that just no one ever

got to and you were just like oh my god

you're a god if you can do it. Okay. Uh

now like I mean I still think it's an

very impressive achievement but it's

being achieved super fast. What is

different about the founders in this

moment such that they can handle this

like or you know like there's like

obviously the internet has matured

channels have matured but like yeah

>> let me let me let me work towards so

this is the right question but let me

work towards it with something that's

very adjacent to what we were just

talking about which is

>> so our view um uh as investors is that

AI actually solves the customer

acquisition problem beautifully

>> but it doesn't necessarily solve the

retention

actually it doesn't necessarily solve

the defensibility problem, right? And

so, you know, you've kind of got this

hack that you can actually throw kind of

pricing at to solve to get a bunch of

users, but then there's this this great

question is like, okay, so how do you

retain them? How do you stay defensible

and things like that? And that is very

much an evolving

um playbook. Like I you know like right

now it seems that as far as I can tell

these markets are so big and growing so

fast that brand modes are actually

taking effect meaning like it doesn't

like like I don't know OpenAI may or may

not be better than the others at like

language but it's the leader because

people know it and my mom uses it which

is crazy right and you know Midjourney

which never took on any VC funding is

still kind of the leader in image and

everybody knows cursor and it's the

leader in code and so you've you've got

things like brand effects that are

taking over um to get these companies to

scale. But then also these companies

have to move to very much traditional

modes, right? Like what are the

traditional modes? Two-sided

marketplaces,

long tail integration modes, things like

that. And so I think the dynamics we're

we're seeing are the following, which is

why this stuff gets so complicated. The

first one, like we said, is you can

throw money at the the the the the user

acquisition problem. So you have these

massive like really quick growths and

then you have a separate motion which is

like so how do you kind of retain those

users once the brand goes away right and

so you see like a lot of kind of product

work in order to do that and then you

have this third one is because now

you've got this hack for customer

acquisition like h how do these teams

actually you know keep um keep the

companies running while this is going on

and and this is kind of all to something

I said previously which is like there is

a gift in all another gift in all this

AI is you could actually make very

functional products with small teams.

It's more of like almost like a science

project than an engineering project.

Engineering is incredibly incredibly

hard to complicate. And so I think uh to

scale it's very hard to scale because it

gets so complicated. So I think we're at

a point where the industry where we

understand the acquisition, we

understand some of the defensibility.

Part of that defensibility now is you

have to go horizontal which becomes a

bit more of a product engineering and

problem. And I think that's kind of

where we are. Like if you even look at

like OpenI released Sora 2.0 right

recently and if you actually look at it

like was it the state-of-the-art video

model? Not really. But there's like a

lot of product features on it. And so I

think now exactly where we are is like

we're we're in the product phase for

things like retention for things like

verticalization. Um and I don't think we

actually have a good answer but I do

know that right now these teams are

forced to be more product teams than

they had been in the past for exactly

this reason.

>> Yeah. So, okay, this is super

interesting because like okay, like

>> and by the way, I t I touched a whole

bunch there, but I just wanted to make a

point that like just because you've

solved the customer acquisition problem

does not mean you've solved the

defensibility retention problem.

>> The companies are focused on that right

now and that's kind of I think this is

where we are right now is right there.

They like they like Yes. And and so

actually now I'm super curious because

in historic modes you kind of are

solving these problems kind of in par

like at sorry at the same time right and

right now it's like but but you just

like somehow got this like hundreds of

millions of ARR overhang which needs to

renew next year and didn't exist last

year. And so, uh, what is like

>> and I know you're working with some of

the like like some of the founders. I

I've met a number of them and I and I've

worked with some of the people who

they're now hiring as their lieutenants

and these are like the smartest of the

smart people like just like the true

truly amazing humans. What are these

companies doing differently in this

phase? Because they're essentially

compressing an entire company formation

into like a freaking year. They're

taking a decade and compressing it. And

and so like how are these companies even

doing this? Like how are how are people

how are founders running these

businesses? Like what like what and

obviously it's like an emergent thing.

>> Let's be clear. It's total mayhem out

there.

>> Yes.

>> It's like it's so let me give a bit of

an analogy which is before you kind of

build this complex product and nobody

would use it and so then kind then

you're kind of like lost in the forest

and you're kind of trying to find people

to use it. And so like by the time

people start using it, you have a fairly

like mature R&D or and product or and

this and that. Like that was kind of

like the standard path before now. And

so it's like you know like this adage of

like you're trying to like you know fly

the airplane while you're building it

type thing. I mean like you know this is

such a cliche right now. It feels to me

is like what you did is like you took a

you know like a like an airplane kit and

then you launched it into the air.

Now you kind of build it before it kind

of hits the ground. I mean the saving

grace is that AI is so magic that you do

have uses to work with and you can AB

test your way into how you do it right

if you keep things together right. So

the so the good news is like you know

we're talking about usage based billing

like you can act you're so close to like

the customer value that you can really

follow

user behavior patterns to know what to

build access like you're not lost in the

wood. It's more of kind of a strict

execution problem. I found the teams

that do the best here are very very user

focused, very very usage focused, very

customer focused. They build the obvious

stuff because the customer acquisition

problem has been solved. I mean I think

that's kind of what works the best. Um

and then you know and then listen I I

don't think we've seen the end of this

story now because um

you know we've got a bunch of leaders um

that are now branching out to a number

of products and every time it seems to

me that there's going to be competition

the market just grows in fragments again

and so I I think the story to date has

just been fragmentation so it's mostly

whites space and until this market slows

down that's what we're going to see and

then the consolidation phase will look

very different right I Let me just give

you like just a like a quick a quick um

historical anecdote on this. Remember

openai was the first to image right with

Dolly.

>> Yeah. Yeah.

>> But they lost that right and then they

were the first to code with co-pilot

really and then you know now they're not

the lead there and then they were the

first to video with Sora and they're not

the lead there but like they're they're

the largest company by far with

language, right? And so the the

conclusion that I take away from this is

is these markets are so big, you can't

focus on a whole bunch of them. You kind

of focus on one and you win that one.

And I I think while that's happening,

you're just going to continue to see the

companies that execute the fastest that

kind of win the white space win. And

then when things slow down and

consolidate, we'll have a very very new

phase where it's going to be much much

more about kind of defensibility

sticking out the ground like competitive

dynamics etc.

>> Cool. Okay. So yeah, I I mean I think

that that general thesis of it's like

you know these spaces were all much

bigger than any of us thought or they're

much more like they're so much more

multiaceted. Um and and obviously one

way of winning a space is with product

and the other is through business model

and monetization. So um how uh how do

you think like when you're when you're

talking with a portfolio company um how

are you like who are the who are the key

people to help a like a a founder or CEO

reason through the like right way to

turn that dial between like where should

be distribution focused we should be

margin focused like generally like how

are you like what is the ideal setup in

this mode knowing that like if you get

it right it's like catching lightning in

a bottle and the flywheel just spins

incredibly

Yeah. How are companies set up against

this?

>> Yeah. Yeah. This is a super interesting

question. So, um so I've always viewed

that company goes through three stages.

There's the product stage, the sales

stage, and the operation stage, right?

And like you can't kind of skip phases

and that the leadership is going to kind

of align to the phase. And historically,

um the product phase is just about the

product CEO or the product founders and

kind of R&D and execution and whatever.

And then you get to the sales phase. And

the sales phase is when like basically

the business side kind of kicked in. And

so like when I was in the sales phase

for my company when we were going from

10 million to a billion um you know my

my my closest business partner was the

CFO and the reason or was finance and

the reason that was is because literally

like sales comp and marketing spend and

going international and had to handle

partners like that was like the heart of

the business. Of course when you go to

operations that's the case. So the

biggest shift that I'm seeing to answer

your question is finance is being pulled

into the product phase. It's so

interesting right? Like we've never seen

this be like what is someone in finance

going to do with like you know early on

with like while you're still building

the product and finding product market

fit and the reality is because you do

have this knob towards ad value so much

more directly and it so directly impacts

the running of the company that things

like the margins like you know even

things like you know like like do you if

you have like a new feature like do you

incent this by like investing money in

it by like giving it away or not and so

>> I mean I am seeing finance being pulled

much more strategically much more

has always been incredibly strategic for

the business, but it was it was always

in these last two phases. It was in the

sales phase and in the um the operations

phase and now I'm seeing it in the

product phase. And so, listen, you have

you now you have product teams that are

incredibly finance savvy. You know,

you've got R&D teams that are incredibly

finance savvy. I mean, I work with

companies, no joke, where the engineers

themselves have dashboards on how much

it costs for what they're working on.

Does that make sense? So even as part of

R&D like kind of the billing is going

through it and so I think just finance

has become much much more strategic as a

result of kind of this new this new

epoch.

>> Do you think that like kind of so it

sounds like one way that companies are

reacting is every

I would I actually love to hear your the

question that I

>> Yeah. Well, so actually it's so actually

it's interesting. I I I actually

strongly agree with like so one

observation I've had is that like our

engineering team for instance is like a

lot more financially literate than any

team ever was at Dropbox like ever even

at the like terminal phase of the

company you know except for like maybe

the senior info leader or something and

um and it's like it's like it's like

they don't own a P&L but they understand

the P&L like in depth and they're like

okay this feature you know I actually

literally just yesterday saw a feature

request for one of our largest customers

and like an engineer was bringing like

here's the cog's analysis of how it was

going to take to serve this feature and

I talked to the PM and we think we can

charge X and I'm like oh my god like

what this is coming from engineering

like cool like working as intended I

guess like that's it and um and I think

it's exactly goes back to this idea that

like

>> when you're in a consumption business

like you're it you're just talking about

value and then you're talking about

value splitting between you and the and

the and the customer but like everyone

should be obsessed with value like it's

like it's everyone's job to like be

building value for your Otherwise, what

are you doing at this company? And and

so engineering, it's like it's like

actually very empowering, especially

because like actually math is like very

easily like it's like engineering brain

works really well with thinking about

like P&Ls. Um so that's one observation.

I think I'm the other thing that I see

is it's exactly what you're saying like

the finance the person who runs finance

is is they're extremely young. First of

all, like we're seeing like CFOs who I'm

like, whoa. Like you are like a CFO and

you're legit, but you are way younger

than like, you know, it's like the meme

of a CEO CFO is like it's like you're

like this like 80year-old dude and you

have white hair and you're like you're

the risk mitigator. And now

Exa. Exactly. And now it's like the

exact opposite where it's like they're

actually

>> they're like they're like pushing the

risk pedal and they're they're they're

smart about it and they have full

control but because their business model

is like so tied to value it's exactly

they're they're able to like see through

actually oh because I control both sides

I don't just manage costs I actually

manage upside as well. I can identify

here is this and and so what you're

seeing is this like compression of like

in at at Dropbox we'd have like the

Stratin or who's all about upside and

then the the the accounting org that's

all about minimizing downside and now

it's like compressed into one human and

that person's very generalist they're

generally very like mathematically

inclined but they look they you know

they don't they don't they're not the

accountants anymore they're not like you

know they they work they outsource a lot

of that stuff and so in a way it's

actually a much more

>> interesting

super strategic role and it looks like

and and so one of these things that I've

observed just is that these people are

>> like I think at at public companies the

CFO is always the right hand and there's

like a reason why the CFO frequently

becomes the CEO at

they're in the operation stage exactly

about like

>> a few and now it's exa exact like to me

it's actually the opposite now it's like

the CFO is like a mini CEO And it's just

like it's like in the in the in the

maybe the best possible sense which is

like the possibility space and they're

like yes they're managing risk but

they're really there to like see upside.

But to put the finest point on this I'm

going to repeat myself. It is so

important. Like listen I've been in this

industry for a long time 25 years man.

I've done two companies of my own. I've

done over 200 investments. I have seen

it all. I have never in my life been in

like a fivep person company where you're

talking like about like like product

design like you haven't even started the

product like you're literally talking

about like okay what are we going to

build who are we going to build it to

and a key pillar of that conversation is

is is pricing cogs in the finance I mean

it's this is just it's the closest we've

ever been to delivering direct value and

paying direct cogs and so if it's not

pulled way way forward you're probably

not giving it to do that it that that it

warrants.

>> I agree. And I think the the the kind of

the the company that taught me this

lesson super viscerally was when I like

we embedded in Snowflake for a while and

just like got to watch that team

operate.

>> And

>> they told this story that like has never

left my head. This is like six years ago

at this point. And they're basically

like, "Look, the truth is at this

company engineering, it's like when

they're sh thinking about shipping a

feature, they're like, you know, at at

Dropbox, if you could ship a cost-saving

feature and just like drop cost by 50%

or whatever, you'd have shipped it

yesterday." You're like, "Of course, I'm

just going to go book the win." But in

these usage based businesses, the cost

and the usage profiles are so deeply

related. And so the engineering team,

it's like actually they have to deeply

think about okay, how does this like

cost savings affect our top and our

bottom line at the exact same time and

and it's like when you have these like

tied to value metrics like yes, reducing

query time against some uh database is

like in theory good, but if it if your

monetization metric is too tightly tied

to that thing, then it actually could be

negative. And it's like it's like it

introduces these really interesting set

of incentives. And so you realize that

like if engineering can be tweaked in

that way like think like think about

sales think about every other department

now it's like like snowflake you know

they have this saying or they at least

they did which was like basically the

contract ne sorry the uh the the like

sales starts when the contract is

signed. They sign a $10 million

contract. Now sales starts because now

you need to burn down that contract and

actually make sure that like the the the

product is actually getting used. And

it's this like NRRbased like mindset

around like it's like bookings, who

cares? It's like much more about like uh

actually like uh uh making sure that it

gets consumed. And that is such an alien

concept for people who kind of came up

during the SAS era where like vaporware

sales is like just as good as as as as

software that actually gets used. And I

think this new world is different

>> and and oh by the way this is totally

changing orgs like it changes like you

know what is pre-sales versus what is

post sales. It changes what is a

customer success or it changes you know

how do you incent um the teams on these

things? Who do they report to etc. So I

mean the the the change is kind of as

dramatic as as you can get. It's

everything from product to sales to

salescom to organizational design and

and right now by the way like I would

say you know outside of like the

researchers that you know Mark

Zuckerberg is poaching for billions of

dollars like some of the most in demand

people in the industry are post sales um

just because like actually expansion is

driving most of the economics um of of

these companies and so like really

>> well it also fits your

it fits your thesis around this idea of

like actually Um it's like this like

being multi-product thinking about like

okay acquisition is not maybe not a

solved problem but it's like a much

easier problem.

>> It's a much easier problem now than it

has been. Yes. Exactly. So it's it's

almost like we've shift like it used to

be everything was like just get people

to like get the product like it was so

hard and you like spend especially in

spend years starving for oxygen like

waiting and now you basically can get

users by subsidizing it and and almost

everything in the company strategically.

I mean, listen, I'm trying not to be

categorical. It's not it's not

categorically true, but so much focus

has shifted to. All right, you have

users now. What?

>> Um, and that's really where a lot of

strategic bulk is. Yeah.

>> Okay, one last question and then I'll

jump to Q&A. But, um, as as an investor,

I guess like um what what is the biggest

like kind of mindset change you've had

to make over maybe the past year now

that this stuff is like slightly

maturing? It's like it's obviously we're

like injuring the maybe the like some

amount of stabilization phase. Maybe

it's just going to get crazier from

here. Who knows? But like what's your

biggest update as an investor as you're

like looking at these companies on day

one of because you invest primarily at

seed, right? So like when you're seeing

companies for the first time, like what

is the thing that has updated the most

for you?

>> I mean, I just think that we kind of

know what's working and there's areas

that's speculative and I think we just

tend to conflate these things. So So

what do we know that is working? um

bridging bringing down the marginal cost

of creation to zero is working. And what

is that creating? You're creating all

sorts of stuff, right? You're creating

images, you're creating videos, um

you're creating stories. So like that

like those companies are really really

working. Um and you know, we hear a lot

of these like whatever that stupid MIT

report of like 95% of enterprise things

fail. Like like these just don't make

any sense. It's like this is a proumer

revolution of like creatives and there's

tons of like you know profitable

companies at scale in that space. So

like that's definitely working right. Um

coding is definitely working right. Um

you know it just turns out that like

let's say there's 30 million developers

let's say and there's there's probably

more. Let's just say 30 million. Let's

say they make on average of 100k a year

right? you're talking about a three

trillion dollar market and you've got a

tool that touches all of them and you've

got multiple products at over a billion

dollars. I mean, this is a really really

meaningful shift. So, we know that's

working. You know, there's a there's a

less talked about thing that's working

which is like um this kind of emotional

connection stuff, right? Like some

people were kind of grumpy about GPD 4.5

because it wasn't as obsequious as like

uh you know four. And so like we do have

an emotional connection. We have a bunch

of very interesting companies that are

kind of doing something computers have

not been able to do is to do that. And

so we now have a lot of clarity on

things that are working. Language

reasoning is working very well. Like if

there's a language reasoning test, you

can do research like that's working. But

there's things that are still not

working, right? And and we just conflate

these two cuz like so again like

>> you know like the MIT, you know, 95% you

know fail. Probably what's happening

there is like okay so like there's some

big company and the board was like we

need AI and then the CEO's like we need

AI. So they go to the platform team we

need AI. They're like, "Let's do AI."

And then they hired Deote and the Deote

doesn't know what the hell they're

doing, right? So like everybody comes

in. I mean, in this new thing, I mean,

Deo's great. I'm just saying like in

this thing and then they do AI, which of

course is going to be a total failure

because

>> like, you know, the enterprise is just

still trying to figure out how to do

this. It's very much a secular

phenomenon. It's very much coming from

users. There's a shifted user behavior.

That means new product. This is an

entirely like new super cycle. So you

can't look at kind of like enterprise

buying. And so I would say my biggest

takeaway is like we know what's working.

It's this very secular change in user

behavior. We know what's still on the

come which is deep enterprise you know

stuff and that's kind of what where it's

at right now. Um uh but we know that

there's enough momentum that it's going

to follow the same cycle as the internet

or the browser where it always starts in

the secular way and it goes towards

there. And so listen I think we can say

like this stuff is magic. It's working.

It's a big huge wave. Uh but that does

not necessarily extend to things where

it hasn't matured yet and uh we just

need to be kind of very sober about kind

of where we are in the adoption cycle.

>> Yeah. Actually this I said that was the

last question but there's there's one

question I've been meaning to ask you

for a long time

to let's

>> No, there's a million more. Uh but this

one is um I think it's like I think one

of the things that's striking about you

as an investor is you're very

consistently in my opinion kind of

operating at the edge. you're like kind

of in I know you're in all the discords

and all these weird things. How uh

>> how do you source these new things? Like

I I I can't imagine like you're like

like taking emails from your inbox like

what is your what is your algorithm for

exploring the edge and kind of

consistently finding it?

>> Oh, I mean listen I've got the I mean

I've got the best team on the planet. I

mean I've got a team of you know 8 to 10

investors who I mean we we kind of

segment up the world. I mean the way

that we view about the world is you know

I only focus on infrastructure and you

know if there's a set of like two or

three good founders in a space we just

view it as a you know it's a viable

space like if a good founder is going to

like risk their you know opportunity

cost their you know family's wealth on

it it's a good space

>> and then we literally we use that as

kind of an indication of a of an

interesting space and then we just do

all the research we talk to all the

customers we talk to all the competitors

everything that we can to learn about

that space so you we kind of follow the

founder network market. It requires a

team to do that. And then again, we're

fortunate to have like a top portfolio

um that you know tends to have platforms

that a lot of you know early uh adopters

will use. And so we kind of have like

that visibility into this. But it's a

very multifarious

um approach that requires a big team and

and then you know honestly it's like my

favorite thing is innovation and

startups. So it's where I spend all my

free time too which helps.

>> Yeah. Cool. Okay. Um, I'm gonna jump to

some user questions because there's a

bunch of them and uh and actually maybe

maybe Maggie's gonna jump on them.

>> Yes. Yeah, we have a ton of questions.

So, I'm hopping back on. Um, I think

there were a couple questions around I

think Martine what you said about AI

solving the customer acquisition

problem. So, can you go a little bit

deeper there on what you meant by that?

>> Yeah. So, um,

so for these models,

uh, this the closest we've ever seen to

just exposing a model to, um, you know,

the market and then having users that

you can actually monetize, right? We

just haven't seen this before. There's

so many examples of companies going from

zero to 100 million or 50 million pretty

quickly. And so if you look at, you

know, if you look at the meme company

that's a bottle company in this space,

the problems they have tend not to be

like, can you get someone to use it?

Like even let's I hate the term TPT

rapper because it's actually not a real

thing. It's this weird majority of

people come up with. But let's say

you're doing a startup and your startup

is like I'm going to um I'm just going

to take GPT and I'm just going to change

the website. Like you're doing nothing

but you're like I've got a website

that's doing TPD5. Like people will use

that because it's so magical, right? and

you've done nothing. All you've done,

right? And so you're basically spending

money to acquire users. And so you've

got this kind of magical thing that you

can expose to acquire users. And now all

of the questions come down to how do you

add value on top of the models? How are

you defensible? How do you get away from

kind of this kind of seesaw problem of

of model releases and then not being a

kind of a frontier model, you know, how

do you do defensibility, etc. So the

problem has really moved from the

acquisition problem which these models

tend to solve many times to kind of the

rest of the the rest of the business.

>> Cool. Great. And then there there are a

couple questions too around what's

shifting to usage based pricing. How do

you think about revenue predictability

and forecasting? Question for both of

you.

>> That's a Scott. It's a great question.

>> Yeah. Yeah. I I think it's a complicated

question.

>> I think it's Yeah, exactly. complicated

as I think the the short answer. I think

um I think the the way that I like to

think about this question is like every

company wants some amount of

predictability, but you also want your

you also want unbounded amount of upside

value, right? And so like the good thing

about usage is that as it grows like

your value is growing, the value you're

providing to your end customer. And so,

um, getting to predictability and also

like having a a value model that scales

like as as your customers find value,

like it's like it's almost like, um,

it's almost like, uh, why if you had

like if you're building a product like

Facebook and for virality, like you

wouldn't really like if you found a way

to like drive virality through the roof,

you wouldn't really question it too

much. You'd be like, "Okay, cool,

great." Like I this is a really good

problem to have. And so what I find is

that frequently when companies are

asking the predictability question

really they're kind of either worrying

about like a fake thing that they

haven't yet run into or they actually

have a real version of this where

they're actually like trying to manage

things really really tightly and they

have a lot of data. And so the reality

of like managing essentially the uh the

predictability problem is it is a

problem that isn't really fundamentally

solvable in the early phase and you

really shouldn't stress about it too

much. investors will always ask about it

because like that's what investors do.

They're trying to like minimize risk.

But the reality is like I don't stress

about it. And then at the point where

you do have like a a super highly used

product, yes, then it becomes a problem

and now you have the data to be able to

go solve it. And it just it requires

like eight honestly just a lot of work.

And it like it like you know the the

things you need to get like basically if

you just study like growth as a concept

and you think a lot about like

cohort-based growth and you get really

good at like dialing in kind of very you

know it's like fairly standard

essentially growth man like uh growth

tactics you can usually get to a model

that's like good enough to kind of be

like I can manage my business on this

side. Now sometimes the the

predictability question comes in the

form of customers where customers are

like I don't really want to have to uh

like I'm worried about unbounded uh

spend on your platform which is like a

fair concern for a CFO or budget holder.

Um there I actually look at this like a

product problem. So if you study OpenAI

for instance, when they sign their like

super large enterprise customers, they

like basically build in product tools to

basically limit the amount of money that

can be spent at at the CFO's discretion.

So they can basically say like, look,

this user is only allowed to spend

$10,000 this month. This user allowed to

spend $100,000. This user is only

allowed to spend a thousand. And so what

they're doing is they're transmuting

this like concern around unbounded spend

into a product that actually helps them

control spend. And so actually their

budget tools or cost control tools are a

core part of their product experience.

And so frequently when customer and and

then and then that plus like EDPs which

everyone's super familiar with buying

from like Amazon from the hyperscalers

those two things combined like product

experience plus like a commercial kind

of structure kind of give customers the

predictability like essentially the

downside protection that they need in

order to feel good about buying. Um

that's that's generally how I've

experienced it is like almost always

like there's a theoretical problem but

in practice it's kind of solved with

like fairly straightforward commercial

structures and then investment in

product resources and then also just

kind of realizing that the flip side of

being unable to protect the like your

rapidly growing revenue is actually like

a business that's growing at an

unbounded rate which is like generally a

good like falls into the champagne

problems part of the curve. So I don't

know Martine if you have a different

view but no

>> I think that's great. I totally agree.

This is kind of interesting. I think we

talked a lot about the SAS monetization

shift. There are a couple questions like

if you're a more established SAS company

and you can't move at the fa the pace of

SAS monetization and how it's changing.

Is there any advice the two of you have

there for companies struggling there?

>> Yes. I I actually

the basic answer is like I have like an

hourong talk for boards about this exact

conversation that I've given to many

public company like sea level seuitees.

The answer is like I will compress it

down into like three simple

recommendations and then if you like

really have this problem come talk to

me. I'm happy to like go bend your ear.

So the first thing is um to Martin's

earlier point the move from seatbased or

you know let's say you're on a plat like

a a flat fee subscription to usage is

bigger than you probably think it is and

it affects every department. And so in

order for that to work, it's like a

major transformational moment. So the

right way to handle this is to deputize

someone to basically push this through

the organization because the entire

organization is inertially set up to

resist this change. Like just straight

up your sales comp won't work, your

product development process won't work,

your product doesn't work for it. So you

actually have to take it super seriously

and if the CEO doesn't care or the board

doesn't care, it's not happening and

don't even try. Like that's my my first

advice. Um the second advice that I

think is really really critical is start

small but with conviction. So sometimes

you see these companies they try to do

this like boil the ocean like I'm going

to move five billion in ARR from a

seatbased plan to a usage based plan in

a year or two. That is never going to

work like never. And so what you

actually have to do the smart companies

what they're doing is they're launching

new AI native products. They're using as

a brand like kind of product building

moment. And what they're doing is

they're starting that product as a Z-

dollar product and they're treating it

like a Z- dollar product where they

don't know anything about the viral

loops or how to sell it or anything like

that. And what they're doing is they're

planning over time to shift that five

billion over to like 10 billion on this

new usage based product, but they're

doing it over like half a decade. And

they're like they're and so they're both

like acting with urgency and they are

also like having a like a stomach that

can survive like for the next three to

five years. And uh if you try to do it

too fast, your business will reject it.

If you try to do it if you don't treat

it like a new product, you try to hold

it to the old like you know your your

like super safe product management

philosophy where every change review

takes two months, you're also going to

lose because your competition is AI

native and you're going to move way too

slow. And so the best companies what

they're doing is essentially they're

they're using it like a little bit of a

refounding moment and they're kind of

taking this like AI native product and

they're putting like their most

enterprising startup aligned like PM or

product leader or GM and just saying

here's a budget just go run at as at

this thing as hard and as fast as

possible. And what you will see by the

way is if you do this right is that

product line will grow as fast as a

cursor or a lovable. Like I have many

different examples of public companies

that have launched AI native products

that started at zero in January and are

now at 150 million like a year into or

less than a year into the product.

>> Why is that? It's because they you like

M Martin called I I actually like like

this frame that Martin uses which is AI

is magic. When you're selling magic,

turns out people love magic.

Yeah. And you have a distribution

channel that is huge. And so it is like

you have this backbook that you can like

if you actually do have magic you can

find these viral loops and you can get

growth that it like looks like you know

looks like a an AI native startup within

this giant context. It just requires

though that you strip away all the like

bureaucratic crap that comes with being

a large company. And if you do that you

actually can capture this lightning in a

bottle. And there's tons of examples in

the wild. like you know I won't name

them but like all the public companies

that have this you can bet they're

talking about it in their freaking 10ks

like it's like it's like front and

center in these things. I I think

there's a very important aspect of this

um to call out which is the only real

sin in this epoch is zero sum thinking

and I think that a lot of companies are

in paralysis because they're like oh SAS

is dead like this is going to change and

that's just patently not true this is a

new behavior it solves a new class of

problems like user expectation is going

to evolve but that doesn't mean that

these companies are irrelevant right and

I just feel like I feel like there's

this big explosion And then a lot of

companies like wait I'm still here like

I didn't get this. So now they're kind

of like coming up with like they kind of

survived like the first wave of these

things. And the reality is is is listen

AI is not great at system of record.

It's not great at transactional

guarantees. It's not great at things

that traditional computers is really

good at. So if you are one of those

companies and that's what you do. you

have to realize that the the the the

user expectation is going to evolve and

you're going to have to evolve with

that. You know, there are probably areas

of your product that AI is very good at

and you're gonna have to evolve to

those. Um, but this isn't the

existential crisis that I think most

people are are are viewing it as. And so

I I would just recommend like listen, if

you're in the stage, do not like succumb

to the paralysis of like, oh, all is

lost. You know, realize that you can

incrementally do this. I think Scott

articulated very well. Um, but like the

zero sum thinking is the worst. Like

it's like, oh, like, you know, this is

going to come at the cost of everything

we do. It's just it's just the wrong

mentality.

>> I agree. And I think that the key the

key thing I've observed is that I you

can actually because we sell to a lot of

these companies, you can kind of sort

them on the first call where you're

like, are you are you both uh

enterprising and willing to question the

status quo, but you're also not like

fatalistic. You're kind of like you're

just like you're you're in this like

mindset of yeah, we know that there's an

huge opportunity for us here. We know

that these markets are way bigger than

we ever thought they were. And uh and we

have the air cover to go do the things

that we need to go do. And I think that

like balance is pretty rare, but when

you see it, you can you can tell like

it's like extremely obvious. And I think

um the the smartest CEOs, they're kind

of they're understanding it now and

they're seeing the positive examples.

Like I think Figma's like been pretty

successful on this front and I think you

can kind of see it and I think the smart

CEOs are like okay cool let's just do

that let's just do that like let's let's

copy the playbooks and the playbooks are

being written right now I don't think

anyone has a perfect one yet but I you

could start to see that you can start to

see it happening and I think um the good

thing is a lot of these companies are

public so they're kind of forced to talk

about it and in fact they love talking

about it because it's like pure

marketing upside for them.

I think we have time for one last rapid

question, but I know both of you talked

about like acquisition kind of being

unbounded now and the shift towards

defensibility. Is there any advice you'd

have folks take away on thinking about

defensibility where monetization plays a

role in that?

>> Martine, I'll let you go. I think you

thought about this a lot.

>> Um,

where monetization impacts

defensibility. That's

>> or just defensibility in general and if

my

>> So here listen here here here here's the

best that that we have which is um

you know there's probably as far as we

can tell no endemic modes in AI that are

new that are like endemic to the

technology. Um in fact you could argue

there's perverse economies of scale

because these models uh get distilled

pretty easily. So it's it's pretty easy

for incumbents to catch up and they're

actually very resistant to being layered

by software and so like it's kind of you

expose it directly to the user and so

like users can just kind of move between

them without getting locked into a

complex workflow. And so like there's

they're almost like um like def like

defensibility resistant in a way. And so

um a lot the companies that do do

defensibility they actually kind of get

very good at more traditional modes.

Like I mentioned two-sided marketplaces.

one brand one is another one. You'll see

a lot of these companies actually really

lean into brand because like and marks

are explaining this um this quickly.

It's very important. I say distribution

is very much a mount which goes to the

monetization like I for me if a company

comes in I'm like we're gross margin

negative and I look and I look at the

company and I'm like oh 60% of it's

positive. They made the decision to go

for distribution. I'm like, "Yeah, I

would definitely want to invest and be

in that because like distribution is

very much a strategic mode in in an

early market like this, which is

directly tied to it." Um, and then

there's just a lot of the enterprise

modes like a lot of these solutions

require understanding the enterprise

needs um and kind of the longtails and

set of integrations etc. And so I would

if you listen if you're an AI company or

launching an AI product, I would I would

lean heavily on on traditional modes. I

he lean heavily on distribution which

again is I mean I think to have the

effect of not to do that correctly

especially as you go multi-product you

want to do it um usage based um and

listen this is one of these areas we're

still learning so maybe Scott you and I

have this webinar in a year

>> yeah things have changed

>> we'll update everyone on on on the data

points that we'll collect in in a year

>> yeah I think my my my belief on it as

someone who's like a founder in this

space is that you know I like kind of

fundamentally distrust the concept of

modes in the sense that I think it can

be like you know it's like u it's it's

hard to it's hard to it's hard to say

that anything in the fullness of time

won't get evaporated but what I do know

is

>> uh doing one hard thing is hard doing

two hard things is really like uniquely

hard things is impossible and so if you

think you can do it that is like a

really uh that is a really attractive

thing for us and so the way I think

about it is when you're building a

product in market pick a market where

your product is disruptive and if you

can find the synthesis of also where the

business model is disruptive then your

incumbent is both forced to follow on

the product and they have to update

their entire go to market motion and

they're like commercial motion and so

it's why I'm like you know obviously I'm

a vendor in the space but like why I

think usage is so disruptive because it

is you're getting a double disruption at

the incumbent level where you're getting

disrupted on the products and on the

business model and so you actually have

to make two massive changes at the exact

same time and that's just really hard in

the inertial context. And so if you're a

if you're a founder in this space,

finding those like soft spots in like

the market substructure is just so it

what you will find is that your if if

your competitor is a tech company,

they'll probably be able to follow you

on product eventually. They'll clone

what you're doing, but they'll be so

slow on the go to market side and on the

distribution side to Martin's point. And

I think that's to me the magic is like

you just do both. And the good thing is

you're young. You're early. You can

build your company to be usage or

consumption native or outcome native.

Like you can build all the incentive

structures from da day one to kind of

fit in this like you know price to value

model. And uh and especially if you're

competing against uh companies that

aren't that way like you can you know

basically every excess seat that

Salesforce sells is your opportunity.

Like their their margin is your

opportunity very literally in this

space. And I think that like combining

the two is actually like the easiest way

to build a really massive business. and

grow really fast. So,

>> well, I think that's a great place to

end it. Um, thank you so much Scott and

Martin for the conversation. We still

have a lot of Q&A, so we'll follow up

with folks we didn't get answers to

today, but we have one more webinar in

the series with Lovable, so stay tuned

for that one. Awesome.

>> Awesome. Thanks, everyone. My name