Rise of the AI Workforce
How did we go from generating LinkedIn posts to onboarding AI Workers? Aaron Levie, CEO of Box, joins us to discuss the AI Workforce and what's next in the workplace.
I had welcome everyone.
I'm pleased to kick off the AI workforce summit
with a very special guest.
Aaron Levy is here, the co-founder and CEO of Box.
Aaron, welcome to the program.
Thanks for having me, Greg.
So, you know, we've both been in the industry
for quite a while.
We were part of ushering in the cloud in the mid-2000s
when you were starting Box.
It was a new bottle for enterprise software.
It was a huge platform shift.
But this is perhaps the biggest thing we've seen yet.
You know, Jen AI burst onto the scene 18 months ago.
And for us as software professionals,
we were discussing like,
what's the impact on enterprise software?
2023 kind of became the year of the AI co-pilot.
Everybody kind of announcing these features
right along the software writing shotgun
with the human workforce,
making them more productive.
But 2024 is really the year that AI worker,
software that can accomplish the jobs of, you know,
of humans or tasks.
And it's really caused us to think, you know,
what is the future?
So, first question for you,
when did you see this trend emerging?
And what's making you so excited about its impact
on the industry?
>> Yeah, so I mean, realistically,
probably started seeing it at the, you know,
second half of last year.
And I think you captured the timeline really well.
If I sort of even zoom out a little bit,
I think for about a decade,
we were talking about AI models.
And AI models being able to do very discreet,
you know, kind of single tasks,
label this image, auto-complete a sentence,
you know, the GPT-2, GPT-E, early GPT-3 era.
Then obviously with chat GPT,
we got the initial inklings of kind of AI assistants
where I'd communicate back and forth
in an ongoing dialogue,
or you'd have a kind of a co-pilot-esque dynamic
where it's sort of, you know,
sitting alongside me, helping again,
still auto-complete, auto-completion type tasks.
But what I think, you know,
the ecosystem eventually figured out,
you guys include it is, is wait a second,
if we don't just use the AI model as,
as really a kind of a database
where you're asking the question, getting an answer back,
but instead really a reasoning engine,
or almost a brain,
and we're giving it and feeding it possible,
not only tasks, but information and tools that it can use,
well, all of a sudden,
you can actually turn these things into agents.
And so we've kind of gone from AI models
to AI assistants to now AI agents.
And we are, you know, quite literally in, you know,
day one of that.
I think the actual agents at scale in the world right now
is, you know, is probably measured by a very,
very low number.
But I think in three or five or 10 years from now,
we'll look back and realize this was the start of a,
of a new wave of software where actually, you know,
software is kind of providing work alongside it
as opposed to, you know,
just helping answer questions
or helping enable you to do work.
It's actually gonna deliver labor itself
inside the software.
So you're, you're right.
You're spot on.
We are in day one.
I think companies are first,
they're starting to see what AI can do for tech support.
They're starting to see what AI can do for frontline sales reps
or paralegals or recruiters or content producers.
And it's kind of bending all of our minds,
but kind of going back to the cloud
and this last major tectonic shift
that you were part of,
given your experience kind of ushering companies
through that mega change.
How do you think this AI worker,
AI agent movement is gonna play out?
Like what similarities or differences do you see to the cloud?
>> Yeah, I think, and this is probably a liability,
but I do tend to think about patterns
through just a kind of clay,
Christensen-esque dynamic of blowing disruption,
which is there's a new technology.
It's bad at some things.
It's good at other things.
You know, what is the axis in which that,
that technology is really good.
And performance better.
And then which axis is it performing worse at?
And will that axis improve over time?
And so you kind of look at, let's say,
let's say cloud as an analogy,
cloud, you know, when it first emerged,
let's not even say SaaS,
let's just say cloud infrastructure for kind of,
let's say AWS, the axis it was super good at
was very flexible compute infrastructure.
So I could deploy it, I could scale it up,
I could scale it down.
What it was bad at was having all of the capabilities,
all the security, all of the infrastructure,
kind of robustness of a traditional data center
at the point when it was first introduced in kind of '05, '06.
SaaS, similarly, I think had that same dynamic.
You know, when Salesforce first launched,
it probably looked like a worse or simpler seabull.
But the variable that it was really good at
was you could deploy it in a three or five or 10 person team
or a small business.
And it was like the best CRM ever for that size company
because now for the first time ever,
that size company could have an actual CRM system
as opposed to hiring Accenture and having a full
kind of deployment inside your data center of seabull.
So I look at AI through that same kind of,
and then, you know, naturally what happens is,
is, you know, Salesforce invests more engineers,
builds more features, eventually, actually,
you know, the axis of that simplicity
gets more and more advanced, it still starts simple,
but it can be more robust.
And all of a sudden it kind of overtakes
the kind of incumbent.
And, you know, cloud computing did the exact same thing.
Starts out simple, dev tests workloads,
people using it to build micro apps
or small startups would use it.
But then ultimately had more and more features,
it became more scalable,
it could support the largest workloads in the world,
and then it became the dominant computing platform.
So if you kind of think about SAS
how that trend and trajectory,
cloud how that trend and trajectory,
and then do we see this in AI and AI agents?
And I would argue yes,
because of the same, you know,
kind of similar axes are at play,
which is right now, you know, what we think of as,
as, you know, best in class AI,
let's say GPT-4 or Gemini,
it can do something super well,
it can answer questions about contracts,
it can answer a single customer support inquiry,
can generate an email.
But, you know, the more complex the task,
the kind of worse it'll be,
it'll get confused, it'll do the wrong thing
after kind of three or four iterations of that task.
So we know that the models have to improve,
we know that it has to get cheaper,
because sometimes you need to be able to have
the AI model do multiple things,
and so it's not just like a single question, single answer,
so we gotta run it through the AI model multiple times,
so we need the cost to come down,
and we need to be able to continue to scale it up.
So it's sort of worse at very complex tasks
that you wanna do, you know, at scale,
but it's really, really good at simpler tasks
that you're just starting out with.
And so I think AI will follow that same trend,
which is you're starting to see the very early phases
of what it's gonna be good at,
but you know for a fact that it's gonna get cheaper,
higher quality, better and more scalable
because of the, you just extrapolate the curve that we're on,
and I think that will follow
then the same exact kind of disruption curve
that cloud or SaaS or PCs even kind of went through.
And so the initial tasks are, you know,
again, like limited amounts of complex workflows,
but still where you have to generate some degree of text
or an answer or a customer communication
or reviewing a document or processing an invoice,
it's really good at that.
Completing the entire business process,
the entire workflow with 99.999% accuracy,
still probably not, you know, we're not there.
But again, that's exactly why it starts as a low-end disruptor.
It means it works better at some things,
you're not gonna scale it for everything yet,
but you know that you can ride this technology curve
and it's gonna get better and better and better.
It's gonna get cheaper and cheaper and cheaper.
It's gonna get higher quality over time.
You guys sell to the Fortune 500, right?
You're out talking to CIOs and the C-suite
of the Fortune 500.
And for some reason, this megatrend feels a little bit different.
Like enterprises seem to be leaning in and wondering like,
how do I need to change my business?
'Cause it's gonna take a while.
What are you seeing happen in large enterprises
with respect to AI workers and agents?
- Yeah, I think, well, well, you have this benefit
of you have an existing crop of, let's say,
technology leaders or even people in business
that now within one generation have already seen multiple point,
like, eras of disruption in their enterprise.
You know, that wasn't necessarily true within SaaS.
You know, the internet did happen and that was pretty disruptive,
but that was less of an internal operational shift
and much more of a kind of customer facing dynamic
of distribution and marketing.
And so because you have IT leaders and business leaders
that experience kind of pre-SaaS, post-SaaS
or pre-cloud, post-cloud,
all of a sudden, you know, it's the same people in charge now
that are seeing AI happen and they see how much shifted
because of the cloud transformation or the SaaS transformation.
And I think that means that they're more prepared,
strategically, mentally, operationally for, okay,
a new technology disruption is emerging.
How do we respond to it?
And they have the capacity, they have the talent
to try and respond.
So I still would bet on SMBs adopting agents
at scale first, to be clear,
but there is less reluctance, less sort of conservatism
or let's say skepticism of AI than what we saw
with the initial phases of cloud for, you know,
a large enterprise simply because they saw
that cloud became real, SaaS became real,
it became the dominant standard.
And I think that you have a large enterprise
that's not wanting to miss out
when that also happens with AI.
So which departments or roles within an organization,
do you think are the best suited for AI workers
or AI agents today?
And why do you think, like, why are those roles really well
suited for this type of disruption today?
>> Yeah, so I think that there's probably a couple
of different axes to think about this on.
It's sort of like what is the, for lack of a better term,
interface in which people communicate
with that particular type of work or labor?
So if the interface is, you're always talking
to somebody via chat or email,
then you have a very good kind of interface
in which, you know, AI can respond to things
and to the end consumer, you know, it doesn't,
it's not obvious that something has changed
or maybe even the convenience has increased
as a result of that.
So that's sort of one dynamic.
You have another dynamic, which is,
which is, is there type of work where you're kind
of largely sitting at a computer screen,
reviewing things, typing things, you know,
generating text of some sort?
And is there a way for AI agents not to replace
what you're doing, but to augment what you're doing
to make it so you can do it even faster?
And so, so I'd say I just described
probably two totally different categories of work,
but both of which offer great opportunity for AI
to become a huge point of leverage.
So, so on one hand, obviously, we, I think we can,
I've come to expect the idea that, you know,
if a customer comes to your website
and just wants to figure out what's the pricing for,
you know, my size company and what product
would I be most interested in?
You know, that, that, it's a far better convenient experience
if the customer can get an instant answer back to that.
And you'll probably generate a faster lead, you know,
when you can reduce the barrier to having that conversation
or asking that question.
Similarly, if you have a customer, an existing customer
that, that, you know, just is like,
I need my password reset or the software's not working
or something about the system is down.
Again, that, that sort of time to resolution,
that time to answer matters more than kind of
almost any other variable to that customer.
And so, another great example where AI can really augment
the work that, that people are doing.
So that's kind of a, you know, let's say a, a, a, a,
a B2C kind of component.
And then you have internal operations
which lend themselves well to AI,
which is, you know, we've seen the most classical one,
which is let's say GitHub co-pilot.
GitHub co-pilot is, is brilliant because you have a workforce
that, that other than when they're collaborating
with other people, you know, verbally
or in any other kind of situation,
they're, they're typing lines of text
in a computer terminal and, and, and, and you can, you know,
you can basically bring the world's intelligence
to that process by typing that text faster
in the form of, of AI intelligence helping you write code.
And so it's kind of like the perfect,
it's kind of like the perfect dynamic,
which is like the interface is like a linear interface.
The, the, the sort of knowledge of coding is, is very, you know,
prevalent on the internet in the form of open source
and, and lots of code libraries and sack overflow information.
And so that's kind of like why I didn't get how co-pilot
has been the killer app of AI.
But things get more complicated
the less your work looks like that.
So, you know, lawyers, I would say, is a great example
of, of a lot of potential.
But there's a lot of back and forth
with an external client or counterparty.
There's a lot of like information that they have to, you know,
pull in from another human.
There's a lot of collaboration with, with other parties.
So, so I think there's areas where AI offers a lot of leverage.
Like let's review a contract very quickly.
Give me the clauses that I need to improve from this.
But at the end of the day, you're still going to go back
to the client, you're going to have to go explain it to them.
And so you can't fully automate that work, you know,
anytime soon, at least.
So I think you're going to see a continuum of, of kind of like,
how often is that task kind of executed in a business?
Is it executed, you know, one time a day, ten times a day,
a hundred times a day, a thousand times a day?
How much information is sort of involved in that task?
How many people need to kind of be involved in it?
How sort of autonomous is it as a task?
And I think all of those variables will, will then, you know,
drive how much AI can automate or augment that work right now.
Yeah, I think some people are calling,
when, when AI just performs a specific task, like, you know,
find me all the contracts in my contract repository with customers
that have this unique clause and, and, and bring them to my CRO, right?
Now somebody used to have to go do that.
You used to take four hours.
Now, like that task can be given to you like an agent.
And in some cases, as you mentioned, like, like an entire job role
might get replaced with the AI, like, I, I think I'm kind of calling
those AI workers.
So you, you mentioned the example of the, of the tier one tech support person,
which is like, I forgot my password, you know, reset my username.
It looked like, like all the basic stuff that you just look in the knowledge
base
and get an answer back to a customer, like, that's probably going to be the
work
that gets kind of displaced by AI sort of first, because the AI can do like
that
job perfectly or the frontline sales rep as you've been injured or things like
that.
But what do you think companies are going to do with all these efficiency
gains and productivity gains?
They get by hiring AI workers and bringing AI agents onto their teams.
What are they going to do?
Well, I think, you know, this is, um, uh, I think there's a way that you could,
you could look at everything I've said thus far and be sort of worried for the
workforce.
Um, I, I, I, I let the human workforce, which is, I think the more important
workforce.
Um, I actually take the opposite lens and, um, and I think that the, um,
I think history has proven that this is the case, you know, time and time again
with technology, productivity gains, um, with an asterisk that this will not be
true
of every individual firm.
It's more of a kind of a macro view, but because every firm might make a
different
decision on what to do with that productivity gain.
But I think, you know, by default, my view is in areas of your business where
you get productivity improvements and let's take like the most obvious, you
know,
easiest productivity improvement of all time to understand at least mentally is
like, let's say sales rep productivity, which is like, I spend a dollar on
sales
rep salary.
How many dollars do I get out on, on the ultimate, you know, ROI or revenue
generated and like that's like, like, you know, that's at least classically one
of the
more, uh, uh, you know, isolated ways you can look at productivity.
It's very hard in engineering.
It's very hard in marketing, but, but sales rep productivity is like, you know,
one of the more classic areas.
If I'm just speaking for box right now, but I think this extends to most, most
companies, if I can, you know, pay a dollar and get back $2, I might be willing
to spend a, you know, a certain amount of dollars to get the $2.
I can pay a dollar and get $2.50.
I'm probably going to spend more money on, on the dollar side on the, on the,
uh, on the investment side because the return has just gone up dramatically for
that dollar invested.
And I think what people think of initially your reaction is, well, if you get
it,
if you spend a dollar and you get $2.50 back, then you'll just spend over time
that same dollar and you'll just be, you'll just give the 250.
And I think most, most people that participate in capitalism kind of say,
well, wait a second, actually like that 250 I want as many of those as possible
Because that might clear a new threshold where the ROI is so positive that I
should
keep investing as many dollars as I can until, until that basically
productivity
number goes back down.
And so my instinct is in most areas where you see productivity gains,
what you'll actually see is reinvestment back into the business.
Now, sometimes that will be reinvestment back into technology and AI, because
you'll just, because it's just working so well, you just want more of that.
But oftentimes in the business as you're growing, you still eventually will
then
need, you know, humans to go do all the things that AI can't do.
And, um, and that's where I'm actually pretty optimistic, which is if I can get
if I could generate X more leads per dollar, um, you know, using an AI platform
I'm probably going to hire more sales reps on the other end of that workflow,
where I was previously constrained because the dollars invested in my lead gen
system were taking up a lot of my dollars and it was only X level, you know, X
percent productive.
Um, but now if it's more productive, I'm going to actually need more sales reps
to handle all those leads.
And I think the same thing is true for engineers.
If I can, if I can, you know, generate 20% more code per dollar of engineering
work, um, I'm going to want, I'm going to eventually have more features that I
'm
building, which eventually if I'm doing my job should mean more revenue for the
company.
And if I have more revenue for the company, I'm going to want more features
again.
And those dollars will again go back into engineering.
So, so I think in, in, at a macro level, I think that most of this
ultimately results in the productivity gains or efficiency gains going back
into
the business to drive more growth, more ingenuity, more creativity, um, as
opposed
to just being harvested.
I think what, what an economist might look at is like, well, that was, get
harvested as more profit.
The problem with that flaw is the flaw with that is that nobody's in a static
environment.
So if you're harvesting those profits, you're going to have a competitor that
says, no, I don't need to harvest the profits.
I'm going to grow faster.
So, so you won't be able to, you won't be able to have, you know, sort of a
ridiculous level of profit as a result of AI, because everybody's in a
competitive
space and we will compete out those profits over time.
So again, this is kind of why I think ultimately the human side of this
equation is, is, you know, ultimately, you know, probably actually thrives in
this
model because it gets us from doing maybe the, you know, more, you know,
road work and we get to now do a little bit more incrementally creative work as
result of this.
Yeah, we saw speaking of, of the competition we saw with the cloud movement,
we saw a little bit of FOMO, right?
It's like your competitor was kind of moving into this like new architecture.
But with AI, I think we're seeing some serious FOMO.
And so we have, we have, we have companies taking action, which is kind of why
we're
here at this, at this summit, talking about the future of, of the workforce and
what it looks like.
Um, do you, do you see companies taking action?
You know, 2023, I think was a year of paralyzation where everybody was
wondering,
what do I do?
Is this thing safe?
What are the rules?
What's my competitor doing?
But what's happening in 2024?
What are you saying?
Yeah, I think that we're definitely seeing absolutely more action than 2023.
Um, uh, I would say that I think this, this still pales in comparison to, to,
you
know, a couple of years from now.
And the reason why, um, uh, ironically, actually the, the rate of change in the
industry, um, is, um, it is, is, is because it's actually so fast.
It actually causes customers or companies to, to really have to be very
thoughtful about architecture decisions.
Um, it actually pays that there's a, even higher premium right now to making
the
right architecture bets on the right platforms.
Because if you just dive in to the first thing that, that you hear about, you
know, it's, it's, you know, it's, you get any odds that, that, that that thing
is
the wrong architecture or the wrong platform that will not actually accrue to
get all of the, the accruing benefits of, uh, of, of, of AI.
Um, and so, so I do think that's what we saw last year, um, is a lot of people
step back and say, you know what?
It's probably a little bit too soon to bet on one AI platform because I'm
hearing
about this open AI thing.
Google's all, you know, getting their act together.
This anthropic thing's very cool.
Um, and so what do I bet on?
And now, you know, a year later, you're seeing that, okay, I, I kind of see the
contours of the industry sort of shaping up.
And so I'm willing to now make an investment, um, and start to get, you know,
some of these, these gains, um, but, but I, but I, I still think we're earlier
than, um, then, then, you know, we're like almost last year's kind of a wash.
I think we're, I think it's almost like this is the first year where you can
actually dive in, which means that probably next year, things will be five
times more mature in terms of where things have settled out.
Um, but I think this is a moment in, and it's, and this is always true of, of
technology transitions.
This is a moment where your architecture matters a ton and who you bet on it
matters a ton.
Um, you know, obviously like if you made a CRM decision in like 2003, um, you
know, it may not have been a hundred percent obvious, which, which, you know,
direction you should go in.
Um, but the people that let's say chose sales force got all of the
compounding benefits of the entire super cycle of cloud and SaaS, but you
know, you make one wrong chain, one wrong decision there and you might be
stuck for five years on the wrong architecture.
And I think that that companies are, are sort of realizing they actually have
to make the right decisions in this point in time, which does, you know,
lead to a little bit more evaluation, a little bit more trying to figure out
what, uh, what, what, what the future looks like.
But, um, but this definitely the year where we're first, you know,
seeing these types of deployments.
Yeah.
It's, it's really happening.
It's pretty crazy.
Let's talk about box for a second.
Um, last I heard you guys had 110 plus thousand customers, something
like that, my directionally sort of right.
Yep.
Um, who are, who are your workers?
And when you think about bringing, you know, AI into the box platform
in the form of an AI agents or an AI worker, like, what does that
mean for your software?
How are you thinking about it?
Yeah.
So, so, um, uh, we're big believers in this kind of based on the premise
of your, your initial questions, which is if you think about, you know,
what are all the things that you could actually do with content?
So what we are, we're a platform that helps enterprises manage their,
their content.
So think contracts, marketing assets, invoices, financial, you know,
documents, HR records, all that, you know, the, you know, the most important
content in your enterprise, you're going to want to secure, you're going to
want to manage, uh, uh, effectively integrated across your software.
And all that content has a tremendous amount of value in it, but it's largely
untapped by most organizations.
And so our, our sort of view is, well, what if you had an intelligent platform
that let you actually get more value from that information?
Uh, your, your question earlier about showing all the contracts of a
particular, with a particular criteria, um, you know, that right now in most
organizations could be like a five hour process at best.
It might be like a multi month process at worst.
Um, if your data is just fragmented across, you know, everywhere.
Well, what if it could become like a one second process?
Um, you know, what would, what would you need to do or have done in your,
in a kind of intelligent content management system that could cause it.
So you could ask any question and get an answer back from your data or, um,
where you take, you know, all of your marketing assets or all of your sales
information or all of your product, uh, data, and you let employees just ask a
question of that information and get an answer back as opposed to get a link
to a file that then they have to go read.
And you know, maybe spend an hour of their day just trying to find one, one
specific answer.
So this is the potential of AI plus enterprise content.
Um, and so our, our view is that, is that you need to go on a platform
and makes that incredibly easy to do.
Um, and so we're, we're building that platform.
That's box AI.
And, uh, it lets you ask questions of your data.
It'll let you not yet, uh, but both our, our sort of direction is very similar
to this conversation was let you automate workflows where AI agents are,
you know, often reviewing content, extracting data, making, you know, to
some extent decisions about, about, you know, that, that, you know, the process
that that, you know, content should go through, uh, in invoices of us, of,
above
a certain amount that should route to one person.
A contract has this pretty good language in it.
That's route to a different person.
Um, so we believe AI agents will, will do a lot of that type of work.
Um, and the amazing thing is, I think for our business and, and for a lot
of other, uh, SaaS platforms out there, I'd say 95% of the use cases
that customers come to us for with AI are not use cases that humans today
are doing in that business.
So this is not that, that there's some, you know, you know, group of people,
there's a hundred people out there doing all this work and AI is going to go
replace that.
It's actually all of the work that we never got around to because it was
just, it just fell below the line.
It was just like the thing that took too long or the ROI just was not the ROI
was just not optimized for, I can throw, you know, a six figure, uh, employee
at this problem.
It's just not, it's not valuable and not, but could I throw up, you know,
$50 of AI compute at that problem?
Absolutely.
And so, and so this is what AI begins to unlock is like all of the stuff
that we never got around to all of the reviewing of contracts or,
or getting our data cleaned up or being able to ask questions of information
that right now people are just going into a slack channel and asking a question
on AI, beginning to be able to do all of that, which means that we just get
more again, productivity for the actual human employees in the organization.
So that's what we're excited about.
Yeah.
2024.
We're here in, in July and we're seeing,
in one of the AI workforce upon us, AI customer support reps on the front
lines, AI inside sales reps, the examples you mentioned with box around AI
operations or AI analysts.
Uh, we're talking about AI paralegals.
We haven't even talked about recruiting, right?
And sourcing candidates there.
There are so many examples and I think it's captivating people's minds,
which is why we're all here at this summit.
But I know you don't have a crystal ball, Aaron, but you're pretty good at kind
of seeing what the future looks like in the next couple of years.
So as you look out now on two years from now or three years from now,
what's the world going to look like and how fast do you think we're going to
see all this progress?
Yeah.
The problem with any crystal ball stuff these days is like, if you would ask
that question in October of 2022, uh, my answer would have aged, uh, you
know, within a month as like the dumbest answer of all time.
So, so, you know, I'd be talking about like VR AR as like, you know, the
only thing that matters.
And so, so I'd say it's very hard to predict anything at this point.
Um, you know, that we're on an exponential curve and, and that, that, you
know, you can kind of try and extrapolate from an exponential curve, but
then there's crazy step functions that sort of exceed even the exponent, um,
that, that you were on.
So, um, so if, if I just take all of the, if I take today's curve without,
you know, any knowledge of a step function or kind of complete asymmetric shift
then you can, you know, at least a few things are true.
Your AI models get vastly more intelligent.
That matters because that means that it can make, uh, it can, you can bring
more
reasoning to more basically workflows.
Um, and so, you know, let's say you have a, uh, a case of a, of a, of an AI
agent
answering customer questions wrong 5% of the time.
Well, as AI reasoning goes up, you can just see that that, that number will
shrink.
And so, you know, probably what, if that number is, is 5% today, it'll probably
be like 0.1% in three years from now.
So just that's where the exponent, that's where the exponential shift in AI
quality of the models goes.
Um, you have another component, which is the cost of compute continues to drop,
mostly because of GPU performance going up and AI model efficiency improving.
And so that means that if today a task with AI costs, you know, $2 to complete
that task, you can just kind of, you know, squint and see that it'll probably
cost like, like 10 cents in three years.
So, so just think about what that opens up when you have a 90 or 95% reduction
in underlying cost of, of what you're doing.
So you have a, let's say, a, you know, a 20 X improvement in costs, uh, an X
percent increase in model quality.
That just opens up way more use cases where AI right now might look like a toy,
but, but you just continue to merge up that, that, that curve of, you know,
more
and more complex use cases because the technology is cheaper.
It's more robust and it's in its higher quality.
Um, and I think that that we're already, you know, we're already on that curve.
And so in three years, you'll just see, I don't know, a hundred times more
agents in the world doing work than what we have today.
Well, the PC movement was exciting, the cloud, SaaS, mobile, social media.
But I think every, you know, all that pales in comparison to how exciting it is
in the tech industry right now, Aaron, I want to thank you so much for joining
us
today at the, I work for summit.
I really appreciate it being here.
Good to be here.
And this is like the old, there's like the final movement I can participate in.
So like I can't have any more technology shifts in my lifetime.
So like I hope this is just the final one.
Let's make a count.
It's been crazy.
So all right.
Thanks a lot, Aaron.