We are probably over-indexed on the flashiest parts of AI in the consumer world and then under-indexed on the things that will just make hundreds of billions of dollars in B2B.
What's the coolest or most mind-blowing demo you've seen in the last three to six months?
I wish I could have my pre-ChatGPT wiring for every individual breakthrough, because I think I would just be like holy shit, this thing is witchcraft, like every day.
How do you think about durable value creation?
The things you probably want to do to like maximize your moats are.
Welcome to the Logan Bartlett Show.
In this episode.
What you're going to hear is a conversation I had with co-founder and CEO of Box, Aaron Levy.
We cover a bunch of ground related to the state of the enterprise today and what he and his customers are thinking about artificial intelligence.
Talked through a bunch of different topics related to AI, what different people should be thinking about, his views on closed source versus open source and the opportunity that presents going forward, as well as some of the considerations related to the business model change that seems to be occurring with usage-based pricing.
A really fun conversation that you'll hear with Aaron now.
Aaron, thanks for doing this.
Thanks for having me.
What in your mind is a model company in today's parlance as we sit here, March of 2025?
Wow.
Okay.
That's just jumping right in.
What is a model company?
I mean, I don't think you actually even have many model only companies.
And so maybe the only distinction is you have companies that have models and then other things.
So if I'm being like extremely technical about the definition, there's probably not that many model companies.
I think you have AI product companies that have research divisions that build models and then they have APIs into those models.
So I mean probably to this point.
It's like Mistral is probably the the only true model company where most OpenAI I would not call a model company.
I wouldn't call Anthropica a model company.
I'd say those are AI foundation frontier lab companies that have products and models and services and just all the things.
Do you think that serving enterprises with some type of model offering but ultimately selling whatever the core things that enterprises mostly buy?
Do you think a company can build long-term value around that?
Or do you think that will ultimately, whatever that is resides within an AWS or a Microsoft Azure or a GCP or something like that?
I think it would be hard to underwrite that alone as a massive business, because you basically are dealing with this issue where all of the hyperscalers at a minimum, and then a few other companies that aren't kind of natural hyperscalers but they have, they're adjacent enough to the hyperscalers.
They need to make sure their models get used at an unbelievable scale.
And because they have to win an infrastructure war, they need the compute to come to them.
They have other software they have to sell.
So it is so existential to them to have their models get used at massive scale that they will constantly drive down the price of AI and the quality up in AI that you would just get squeezed into such a difficult position that I'm not sure what your value proposition would be.
We ran these tests on Gemma from two weeks ago and the new open source model from Google.
Yeah, not to be confused with the Severance character, Gemma.
Yes, exactly.
So we ran these tests and it actually performed only like a couple of points worse than Gemini on our evals for accurate data extraction from enterprise documents and data.
And so the fact that you have an open weights model from Google that is almost at par with their commercial model that is, only was already kind of state of the art two or three months prior as as a set of breakthroughs means that, like you're gonna, as an enterprise, you're gonna have a choice between an open weights model.
You're gonna have a choice between a commercial model.
They're gonna all perform extremely well.
And you're really only gonna be making a choice of like do I want all of the scaffolding managed for me?
Or do I wanna just run this in my own data center or GPUs or my own cloud?
But imagine being a complete standalone company trying to do that when you have Alphabet and Meta and Microsoft and then OpenAI and Anthropic and X Everybody's doing that.
I don't know how you'd sort of slip in there.
So you know there's a case to be made for obviously an extreme like deep vertical um, you know use case or or some you know kind of of of niche.
But but i'm taking your question literally.
I think you'd still eventually then need to build the software around it.
You'd have to build, You have to get really good at all of the technology that surrounds delivering that AI model to the customer in a way that lets them actually use it, as opposed to just selling tokens back and forth.
I don't think there's a business model for kind of an independent player at scale to just own kind of monetizing tokens.
If you were running OpenAI, would you keep pushing the boundaries of frontier models?
Or does that just feel like such diminishing returns?
And why not redirect all that energy to the consumer product that is ChatGPT?
Yeah, so I would definitely do it.
And I don't have that much kind of critical feedback for OpenAI or kind of any of the major players at the moment.
The areas of critical issue are, You know, what does Google do about core search?
Like that's a super interesting question.
What does Apple do about you know kind of the Siri problem and and like how much do they need to own themselves versus you know kind of let their, let their, let let other players you know kind of you know power, some of that functionality?
So, you know, there's some interesting strategic questions that I think emerge, but in terms of open AI, their execution, I think it's exactly what probably I'd be doing, which is, let's kind of, you know, it's not even hedging your bets.
There's a nice flywheel.
So they're not like different work streams.
It's like let's have the best models in the world.
Let's then incorporate that into a consumer and prosumer and then business set of products.
And then there's now a new layer which is like let's make sure that we have like applied use cases in these agents, the rumored engineering agent and all this stuff.
And so they can kind of, I think they can have their cake and eat it too in a lot of this.
But it's critical that they are always at the forefront of the state of the art models, because that's the calling card into the business conversation which is we also have the best performing models on these dimensions.
And if there was only you know, the chat gbt play, i'm not sure that would be.
You know that that is not why their brand got to where they are um, and i think you want both those things to kind of be interlocked.
Now there was an interesting thought experiment that somebody mentioned recently, which was for at least the consumer use case.
If tomorrow you swapped in Gemini or DeepSeq, you know, just like and just you know, had a person go to ChatGPT, would they notice the difference?
The answer is, you know, obviously probably not at this point.
Like we've reached.
We've reached a convergence on model quality for most consumer use cases to be satisfied for, but most as in like.
As in like, you know, the majority but not necessarily the most important ones.
And so there's going to be an 80-20 rule, which is we probably have solved like 80 of like hey, what's the score to this sports game?
Tell me about World War II.
You know, what should I wear in this meeting?
Like, we've solved that.
But the 20 is going to still be like orders of magnitude more quality that we need, which is like go out and do the classic.
I need you to book a flight type problem.
And those were nowhere close to solving.
So we kind of got to the 80% really quickly.
And now the 20% is going to take years and years to kind of keep cranking through.
I get the feeling you're excited about agents if I read your tweets.
Is there an agent that you think has product market fit today?
Yes, but a couple of things.
I don't know if everybody's equally defining agents in the same way, but I'll give you three or four.
One company that I happen to be an investor in but I think they just independently, you'd probably agree is this sort of space around outbound selling.
So one company is 11X, I think there's others that are good.
And I think they've tapped into, maybe the product market fit right now is like lead gen and then it kind of expands from there.
But for all intents and purposes, people are using this to augment their sales capacity, generate more demand.
That seems to be working.
I think like writ large coding agents seem to be working.
Again, maybe we could define different levels of agenticness, but what you're seeing in Cursor Windsurf Replit, you know love, a little bolt, et cetera.
Like obviously you know probably different use cases, like probably 80 of the vibe.
Coding is more fun and it's interesting to watch and not going into production.
But I could see a lot of, you know, product market fit for, hey, I have this internal IT tool.
It's not, you know, not the most strategic thing we have to do.
I can have an agent kind of write it or modify it and then boom, you know like I've just like solved that problem that used to maybe go to low code tools or you would have to, like you'd either never solve or you might hire a contract firm and it would take six months to go build.
So I'd say that's product market fit.
On our end we have one of our first kind of agentic use cases, in just the fact that it's not kind of a single one shot in the model is just we extract data from documents.
So then you can say, hey, here's a contract, here's an invoice.
I want to read the document, put that into a database and then automate a workflow around it.
That's clearly got, at least in our world, product market fit.
Mostly our problem is we can't keep up with all the range of use cases customers want.
And so there's more demand than the ability to satisfy everything that the customers are asking for.
But those are a few.
On the consumer side, I'd say probably not a lot of PMF on agents.
But on business, my bet was this two years ago.
My bet will probably be it in five years from now.
We are probably... um, over indexed on, on the flashiest parts of AI in the consumer world.
And then under indexed on the things that will just make you know hundreds of billions of dollars in B2B.
Um, and like, I'm sure like Nikita is like, uh, it's good.
You're talking my book here.
Uh, same.
Yeah.
Yeah.
Yeah.
So uh, I guess one question the business model of cloud SaaS, if you will, I think has been historically probably underappreciated versus obviously the delivery mechanism and all that.
I think people have a lot of appreciation for, but just purely how the pricing was able to disrupt a lot of things, be it Salesforce or whoever.
I'm curious do you think that usage pricing is an opportunity, with AI, to have that similar effect, or is it something different in your mind for the broad transformation or disruption, or on which axis?
Yeah.
I guess, as you think about who your historical competitors were, I think historically people underestimated the on-prem to SaaS transition and the number of customers you could reach was probably one of the big things.
But there was a lot of innovation that ended up happening around seat-based pricing and elements of fair use pricing, like Slack was early in that.
And I think all of that stuff.
There was a bunch of different elements of the SaaS transition that ended up being user-based in some way.
That I think was an accelerant of this transformation and transition that ended up playing out.
Do you think usage has a similar opportunity for existing businesses, or is it a little orthogonal and ultimately it might not be as broadly applicable?
I think it's very relevant.
My characterization of SaaS and cloud from a business model disruption was it took something that you used to spend upfront, like maybe millions of dollars, on to now you spent thousands of dollars on.
The CapEx to OpEx transition.
CapEx to OpEx. sort of like perpetual license to, you know, buy the drip.
You know anything that was just like okay, I have to wrap my head around this major outlay of capital system, integrator fees, infrastructure.
I have to build out team.
Like you know, if you, if this is more anecdotal but if you were like to deploy a CRM system in the nineties, you know, just think about what had to go into that like like, data centers, system integration.
Like you bought software for the CRM system but you still had to buy hardware.
You had to buy all the servers, you had to have a network.
So think about how many steps from the moment you said I want a CRM system to the day that you're actually gonna log into it with your data.
I don't know year, year and a half, you know millions of dollars.
Salesforce shows up and they're like literally like tomorrow you can just have a CRM system.
And oh, by the way, you could use it for 20 people in your company.
And there's no, you don't need some kind of like critical mass to make the economics work like, just like start using it.
And so you multiply that over two decades or two and a half decades, and it's like every SMB on the planet can do it.
Every small team can do it and the TAMs of these markets just basically grow, because now everybody can afford a CRM system.
That couldn't.
So that was.
It was because it was a seat base, it was because it was a subscription, it was also just because literally, like your day, one upfront bill was just a hundredth of what it would have been.
You know, because you don't have all the other, all the other things wrapped around it.
Cloud did the same thing, obviously more on an elastic basis with compute.
So I think what agents do is that, I think there's gonna be two big areas of upside.
One is you're doing that now for effectively a form of labor.
And so that kind of has this interesting impact of it expands the TAM of software from only being able to sell the software to people on the other end.
So you were kind of capped by the number of seats to now you have kind of like I can think of categories where the category of AI agent spend will be an order of magnitude more than the spend of the software in a kind of a non AI world.
So take a relatively simple example of like contract management software.
Contract management software as a category is like a couple billion dollars, like globally in the world.
Legal services is a couple hundred billion dollar category.
And so now can software take 10 of that spend through AI and pull from the legal services spend or expand the legal services spend.
And then all of a sudden this category that was relatively niche Couldn't make that much money because the amount of people that wanted to buy contract management software is kind of low.
Could that now be a $20 billion category as opposed to a couple billion dollar category?
So you could do that probably 20, 50 times across software and already underwrite master TAM expansion after going after a whole bunch of new use cases.
That used to be kind of labor essentially.
And then the other part is AI has a way of just completing the things that people have kind of always wanted to do in the products themselves but just were never possible, which then also unlocks a whole bunch of new budget as well.
So I think you're gonna see.
So this is the new version of disruption, is it just will?
It makes it, lets you go and complete the outcome that the customer was looking for in a way that they are not required to go and deliver it themselves with their people.
And then that means that more of that net value accrues to you.
And so that's the usage-based or outcome pricing I think we're gonna see in AI.
When you extrapolate that service budget there forward and think about the value to the end customer, totally makes sense that they'd be willing to pay.
Hey, There was this person that I had to do this and now I can get software to reach things that weren't even possible.
Or do it?
Do it more?
I worry about the deflationary element a little bit of like the competition.
And at the end of the day, are you beholden to what your next challenge, customer or sorry, competitor is willing to charge?
And so we might not be able to actually tap into the service budget quite as much, not because the opportunity is not there, but because we're.
There's tons of venture companies popping up and competing with one another.
So I guess one, we'd love to get your reaction to that.
Two, how do you think about durable value creation?
Because I think everyone agrees.
For example Zoom, I would pay anything for it basically to work if there was no other alternative.
But there's a lot of other alternatives, so there's an anchor on what I'm actually willing to pay.
Yeah, so I don't think that any law of market dynamics changes because of AI.
But I would say that has probably been true in venture startup land kind of forever.
Which is the CRM?
I bet there was a CRM market map from Gartner in 96.
That was like 47 companies that did CRM and somehow only Siebel and like one or two others made it to the other end.
And so did they temporarily cause a deflationary impact on the price of maybe what Siebel could have charged if they were left alone.
For sure. did still create a great outcome for Siebel.
And there's a version of the world where it doesn't sell the Oracle and they pivot the company and now it's a hundred billion dollar company.
So I guess uh, i guess it's 100 true, but but nothing about ai seems to follow a different law of of how these things play out, which is competition will suppress you know what, what you could like, maximally charge, and it keeps kind of all the players in check.
But eventually you get like some degree of network effects, some degree of like you know reference customers that tell each other you know that you can all use you know X thing and so on.
And, you know, like just apropos the timing of this conversation, like, um uh my my dad was asking like what how did whiz get sold for you know 30 billion dollars and we're whiz customer and i i think the technology is fantastic um how does that happen in a world where like obviously there's like 500 security startups like well how does one break out and it's there's a little flywheel where like your product's just 20 better then you get a few more reference customers and then and then you learn faster and then you're product is another 20 better faster and then you get slightly better you know people working on it and and it kind of spins out and then all of a sudden you have one with escape velocity you know out outpacing the rest of the category but yet probably there's still a second and third and fourth player that are pretty good and they'll probably also grow and then they'll get acquired so i think ai i mean i don't i don't see a different dynamic playing out in ai for for any reason And so then the things you'd probably wanna do to like maximize your moats are, again, the same lessons probably from 35 years ago in enterprise software, which is get data, do workflows, make sure you've got kind of enough happening within your ecosystem, that keeps customers to have some kind of there's like a like a compounding value more data that comes into the system you know the next run of the ai agent you know gets just a little bit smarter you know there's going to be a lot of network effects on the embeddings of your data so so the the You could, as a single programmer, you could probably switch between Cursor, Windsurf and Replit and a few others.
But as a network of 500 engineers in a company and I've done embeddings on the whole code base and I'm talking to all the different systems it's wired up to wow, change management on that, you know, becomes a lot harder.
So I think we're gonna see a lot of very similar.
You know, maybe this will be bad news for everybody in AI, but it's gonna look a lot like SaaS and the lessons are gonna look a lot like SaaS.
And so you want data, you want workflows, you want strong customer references and you wanna build a network effect around your customers.
And I think it's gonna play out like that.
And then maybe the one thing that we just have to be prepared for is there might be some pricing dynamics right now in play in AI that won't necessarily last forever, which is like Right now you can kind of comp it to labor in some areas.
And that probably feels like a little bit of a temporary advantage that you might have when ultimately it'll probably be more comp to like typical software margin structures.
And that'll just be a journey that the industry goes through as we kind of land to the terminal business model.
Yeah, I guess we don't need to belabor this point.
The two things that I do wonder about with AI specifically are the solutions that are maybe less workflow UI centric.
In some ways there's like the SaaS world.
Traditionally we had all these screens and logins and you know the workflow that existed on top, and so maybe security in that way is actually the right uh analogy, because there's definitely dashboards and integration points and all that with security, but ultimately the work is getting done oftentimes outside of those screens, right.
It's pulling in all the data in some ways.
And so, yeah, it's just something we've been noodling internally.
And what do you conclude because of that?
I think the things that are just tapping into the work automation be it customer support or SDR or whatever it is, Ultimately because there's less workflow embedded into it, the change from one to the next.
I think there's less friction associated with that change.
And so you can swap it out.
You see, it's almost like a meme on the Internet of people going from you know cursor to codium or whatever it is.
And that's not the right analogy because that actually has some workflow around it.
But it does feel like people are a little more willing to switch than at least I had historically experienced.
Where I mean, in truth, what's the difference between Asana and Monday and Trello or whatever?
Well, the difference is like you got to switch your entire team onto something different uh, in a meaningful way, versus when the work is actually being done outside of the ui.
It might just be one click change which just leads to a little bit um, i don't know uh, more ability to price shop, more ability to do other stuff, but i guess that it ties back to the integration.
Similar yeah yeah, so that's the thing.
So so, if you so.
So the question is so it used to be the network.
The stickiness was I had a hundred people that all learned how to use the buttons and then there was data behind the buttons.
And so, man, that's a lot of change.
But the question would be like is there gonna be an analogy where it's inverted, where the AI it's APIs, it's APIs into your other systems, it's APIs into your e-commerce thing?
It's then the logic around when to make the decision about the e-commerce thing.
And it's self-reinforcing and learning, right?
And so at some extent you get all the data, get all the integrations, but then there's ultimately some algorithm or whatever, some AI thing on top that's learned on top of itself.
So sure, you could switch, but then it's a cold start to start over.
Because I could underwrite the exact opposite of this point, which is that when you have a learning system, that learning system will probably perform better than your people did.
And so previously you could change out, you could swap out software and there wasn't like so much IP in the heads of the people about how that software worked.
Cause we're just like, okay, click the next button and whatever.
But, like the AI is gonna track every event it's done in history and just get a little bit better and the signal will just get a little bit better and it'll have a little bit more fine tuning or it'll have a little bit better instructions to the model.
And then that becomes some opaque know data source to you, the customer that is inside this, you know ai product that just makes the thing always perform five or ten percent better in a way where you can't just go to people and say here's the new sheet of paper for how to use this new product we just implemented.
Like you almost have now created your own, for all intents and purposes model for doing customer support or coding in your product.
So the more black box or obfuscated or non UI oriented it is to me doesn't necessarily correlate to its stickiness.
And in fact, maybe it could be the opposite.
I heard you say that cloud was pure efficiency, and maybe there goes back to one talk you were giving at some big bank that everyone's eyes were glazing over.
I think it was like a conference or something that maybe the most boring or least interested you've had a crowd.
But it wasn't like tangibly output based in some way that, like the rank and file within an organization, didn't care about the back end infrastructure.
And ultimately you're speaking about some elements of it.
Can you give that quick anecdote but then also talk about how AI is different and how that translates to the enterprise and the conversations you're having?
Yeah, so if I recall the anecdote properly, it was I think I was just referencing that, like in the peak of cloud adoption, I did this keynote on, like the future of you know running your bank or you know financial services.
And it was mostly just the pitch cloud.
And it was, you know the, the.
I found it very exciting, but the audience didn't, because it was because it was just like yep cool.
We get it, man.
Like the servers are not in our data center anymore.
They're in now a cloud data center, like really like super cool.
And and so so that that to me was kind of the.
It kind of captured the cloud transition, you know, at an industry level.
In a nutshell, which was like it was it made your IT operations faster.
It meant you could store more data so you could make better decisions.
It makes your employees more efficient because they have better software, but it didn't radically transform like entire industries or sectors.
And AI conversely, has the opportunity to actually radically transform how you run your business, how you make decisions, how you serve your customers, how you build products.
And I think it's.
I think it's always useful to learn, figure out what things will be similar to a prior technology transition, which things will be different.
I think the thing that will be similar is that that adopting AI versus cloud will take all the same change management.
It'll take all the same meetings.
It'll take all of the same privacy and policy and governance and compliance review people.
It's gonna be all of that.
But I do notice and see one big distinction in the two waves, having kind of been at the similar timeframes for both.
In the very early days of cloud, you'd go into a meeting with a bank, a pharma, a government agency and you'd say I'd love to talk to you about moving your data to the cloud.
And it was sort of like not interested shock, Like definitely not gonna happen, not gonna happen here.
Maybe there'll be some small little workload that's like a dev test workload.
Maybe some department or fringe group will use it.
And so it was mostly like years and years of just like selling and selling and selling, and trying to get you to understand what was so much better and all of this.
Conversely.
And so we're only about two and a half years into ChachiBT, let's say so.
That would easily describe two and a half years in the cloud.
If I go back to the late 2000s, let's say
If I look at AI right now, that is actually meeting with a customer or set of peer customers is like is like we know this is going to change everything.
Everybody's clamoring for it, every employee needs it.
The new generation coming into the workforce is asking for it.
My boss is asking for it, the board is asking for it.
I need to figure out a strategy and and like it's not when it's it's how It's not.
If it's when it's how it's like, what can we do?
I need to do more yesterday.
And so totally different universes.
And it is so.
Basically, what that means is that you've bypassed all of the normal resistance to this happening that normally happens with one of these big changes.
And now it's purely in the mode of like, how do we make it happen?
Which functions are going to use it?
How do we make decisions when an AI generated the response?
How much human in the loop do we need?
So we've jumped all into just now that mode, five to seven years faster than we did with cloud.
Now that doesn't mean that the adoption will follow a different trajectory, because again it's really people, it's change management, it's systems.
But the reluctance and emotional kind of consternation is very different this time.
And so what is the resistance when you're going into the organization?
Or how would you describe beyond the enthusiasm and trying to figure out the how and the when?
What are the long poles in the tent?
How much are on these organizations versus how much are on the commercial vendors?
Yeah.
Maybe also this might abuse the analogy too much, but if I compare to cloud, cloud could do more than what customers were asking for because they were resisting it.
So you could have always implemented cloud 50 more than you actually were at the time, or hundreds of percent more.
Maybe actually the reverse is true, which is, I'd say, most conversations now.
The customer is actually like they're jumping like three steps ahead once they see the demo and they're like oh, that's great.
Like, we'll just like totally automate that entire thing.
And you're like okay well well, your data's not yet in the environment that it's gonna need to be to do that use case.
Or we're still waiting on some reasoning model breakthroughs from OpenAI or Claude or something to just get that extra 3 of accuracy that you're gonna need.
So in this case, I think customers have pretty much jumped ahead to way closer to the end state, which is like yeah, great.
Like, let's give all the employees a productivity boost.
Let's figure out where we can deploy agents to the organization.
Now, you know, I wanna do a massive asterisk.
I'm talking to the people that are talking to us, which means that, like by definition, they're leaning in, they're a little bit more forward thinking.
So some of the anecdotes I'm sharing are from those types of customers.
I think there's still an 80% of customers that are like, whoa, my head's exploding.
Where should I start?
Where should I go first?
But there are a lot of customers where they're now ahead of the technology, based on the set of use cases that they have.
And so I wouldn't say it's.
The hurdles unfortunately, are much more like practical considerations like literally, like I can't just, you know, turn on an AI thing that looks at all my data and then an employee can ask the question, because guess what?
It's all of a sudden going to reveal, you know, secrets in the organization that they shouldn't have had access to, but they were over provisioned for access, like five years ago.
Or hey, AI agents are really good at doing code auto-completion, but am I ready to put an AI agent in production for generating a full?
You know multi-file, you know output.
And then how do I review that, when the human actually wasn't along the ride for the actual editing process?
Are they gonna really understand the code?
Are they allowed to press submit and that goes into production?
These are the big questions that people are dealing with.
Imagine how unhinged we are online, where we're like oh man, we just made a video game in three hours and it does all this stuff.
And then you're an engineer in a banking system.
Are you supposed to go implement that right now?
Like.
Is that like?
Are we ready for like?
Are we ready for like your bank account to be, you know, driven by vibe coding?
Like, we don't know.
So these are the kinds of things that I think every enterprise is trying to figure out like okay, so what page should I implement this stuff?
You know, is my data ready?
Are my systems ready?
A lot of it is about, wow, the industry is moving so fast.
How do I make, ironically, the industry is moving so fast.
How do I make decisions right now that I'm going to be stuck with in a fast changing space?
So a lot of it is actually like Can you get your architecture set up for like optionality in the future?
Because maybe the vendor you work with today is not going to be the one that's the best in the future.
You need to be able to swap out different parts.
These are the kind of conversations that we have.
To that end you cautioned.
I forget if it was a blog post or a tweet or something, but that being early to AI in some ways can be a liability if you don't set up the architecture with the right abstractions in some ways.
I guess, as you think about the applicability of this for Box as an organization and therefore the advice you might have for others, How do you think about being flexible with this pace of change?
And is it more about being model agnostic or modular in elements of the design?
What would you share on that?
Yeah, all of the above.
So there's how we build and then how we even implement AI internally.
How we build is basically we try and figure out the parts of the stack that.
There's probably a few different categories and I'm making up some of them, maybe the nomenclature on the fly, but there'll be parts of the stack that just will never change.
A model breakthrough will happen and it doesn't matter.
We're not gonna change that particular interface with our system.
So that you can just lock and load and start executing.
There's other parts where you're like we think there's a best in class technology.
Now Maybe we're kind of like we're shaky on it.
We'd probably prefer there to be an open source version.
So that's like a spot where you probably need to be really thoughtful about how much do you lean into a particular part of the architecture.
I'm thinking like your vector search, uh your you know your rag uh you know kind of you know execution you know these are spaces where there's a bunch of first movers you know people started adopting them it's not it's not exactly clear where it all ends up in five years from now so you kind of want to be you know thoughtful about that and then there's the the final part on on your modular point uh which is like This base is changing every three minutes.
You need to decide, do you want maximum flexibility?
So these are the model providers.
Do you wanna be able to have a way of swapping in different parts?
There's multiple parts of Box where we use multiple model providers for the same workflow.
And so you just need to then have like a high degree of optionality, a high degree of flexibility in that part of the stack.
And then you have different like abstraction layers between those different interfaces.
But I think basically anybody in AI at this point has been building with that pattern.
You go to perplexity, you change your model, you go to cursor, you change your model.
I think we're all realizing like okay, you probably wanna let the user change their model, or at least you as the developer have some some degree of optionality there, because we know how fast that space moves.
And then you have to decide which other parts of the staff do you just kind of, you know, lock in on and don't worry about again.
Uh clarna uh sebastian who uh, we've had on before and a good uh, a good guy, great ceo has been.
Um, i think he tweeted something recently, i think he accidentally uh made news, uh in his telling of like, consolidating sas vendors and moving more things to agentic workflow, and i'll tell you, it set off a little bit of a firestorm internally to see if this is like a, you know, a long-term thing that we need to be existentially worried about investing in software because there was going to agentic workflows and systems were just going to do away with the, the need for the interface and therefore everything could be in a crud database, right in some ways.
Um, i'm curious your perspective on on that, as we zoom out and maybe is is what they're doing, you think broadly applicable to other organizations, or is this a one-of-one, maybe exceptional case that they're testing something and on the early frontier?
Yeah, I mean, I found it very fascinating and I enjoyed that it was happening.
And I mean, I enjoy that it's still theoretically happening.
I think some of his updates were that, like they didn't actually do as much on the customer support side, or they have some more you know kind of premium agent, you know human agents that they now have.
But but so I thought it was great from a provocativeness standpoint.
I think we do need people testing the boundaries.
I think it's the kind of thing which is it didn't cause me to rethink all of SAS because, because the the, you know, I don't remember the first time I heard this, but maybe it was like 15 years ago or something.
I was meeting with an IT leader at Tesla and they were like building I'm going to make up 80 of the story but they were building their own ERP system.
And I was like, holy crap, I cannot believe that you're doing this.
And maybe this is the future.
And then you like quickly wake up that like okay, like no like, like you know, this is sort of an Elon special where he needed to control, end to end the whole thing.
And they needed their own.
You know they, they didn't want any of the off the shelf stuff and the amount of fortitude and just you know shit, sheer entrepreneurialness you need to have to even want to attempt that is like kind of insane.
And so I think the client example is more in that category.
Like.
It's just like if you go to the average company and you're like do you want to be responsible for running your own HR system?
Like the average enterprise is not going to say yes to that.
And no matter how agentic know something is you, it does mean that you're responsible for the thing.
Like, like this the, the ai provider that has made the agent that wrote your you know your hr system is not taking responsibility for the ar hr system code that was generated.
So that means, by definition you, the customer, are now responsible for whatever the thing is that it produced.
It's not going to be the AI models.
Claude is not, Anthropic is not taking any responsibility for how your HR system works.
And if it leaks data and you get sued by the EU because there's some kind of data privacy thing.
That's not Claude.
That's Workday.
Workday will do that, not Anthropic.
So I think you'd have to be like.
You know which parts of software are people so unhappy about that they would do themselves, or which parts of software just like are so needed to be customized.
Um, or which parts of software are so unstrategic that you don't mind whipping up these like micro apps for like one-off things?
And so you're fine that it is built by an agent that you can just deploy quickly and turn it off.
And it's kind of, it's a three month, it needs to live in the world for only three months.
Those for the foreseeable future next two years, let's say that would be the only thing I think could be disrupted.
But I don't think your HR system is going away.
I don't think your CRM system is going away.
I don't think that really changes.
To your point about deflationary impact, I do think that there's a chance that you could see you know you could see a new crop of like I'm gonna just like YOLO it and build like the AI.
You know, like I'm gonna build Oracle with AI agents and just give it a shot.
But that would still be a software company that emerges from that.
And you're really using AI purely to just add engineering augmentation.
I don't think the customer wants to be responsible for that software at the end of the day.
That was my conclusion as well, which I think is good for people investing in software companies.
Now, big asterisk, in 10 years, do I want to be on the record as saying this?
Let's see.
But just...
I just don't think companies wanna be responsible for every problem in the world.
Jeffrey Moore came up with this idea.
I think it was Jeffrey Moore that came up with it, but maybe he credited somebody else of core verse context.
And context is just shit that you just want done for you.
You don't wanna worry about it.
Core obviously being like, this is existential to my business.
And for most people, their core business is not their HR system.
And so thus you hire other people to do your HR system.
And so that's why you don't really want your internal IT team being responsible for rebuilding something that you can get off the shelf and just it's done for you.
That's the way I feel whenever technical founders want to reinvent sales comp.
And it's just like, you're not, it's hard enough to differentiate on one vector.
This isn't going to be the make or break.
Yeah, we can go through all the books and talk about it.
All right, I'm gonna give you some quick ones and you give me your your gut reaction on some of these things.
Ready?
Do you think the net of AI in some reasonable period of time, maybe five to 10 years, decreases the number of hours worked by your average knowledge worker?
Well, how many years?
Five to 10?
I'm going to stick with no for now.
Yeah, I think I agree.
Do you think AI will flatten org structures in any meaningful way?
Example, fewer middle managers, faster approval, more self-service, knowledge work?
Yeah.
This one I could buy.
I could buy.
I think actually layers have gotten pretty crazy over the past couple decades.
So I think yes.
I think yes.
And part of it is because I think that...
AI kind of lets you, it lets functions.
My latest kind of conclusion is that it kind of lets functions sometimes do their adjacent function.
And so it lets the designer write the front end code.
It lets the backend developer write the front end code.
It lets the copywriter generate the full white paper.
So I think what that means is that you will have more full stack workers, almost Like we don't really think about this outside of engineering, but I think you'll probably.
I think we've gotten like maybe overly disciplined in job functions in some parts of the economy.
And I think AI kind of lets you have a little bit more expansive responsibility.
So then that probably means that that's a little bit of a flattening when you don't have to hop through and around as many orgs to get things done.
So but we'll see.
Did you ever read the david epstein book range by generalist triumph?
No uh okay yeah, kind of.
It sort of talks about like they give the example of roger federer growing up and playing multiple sports versus tiger woods being specialized in one, and how many breakthroughs in society have occurred from people outside of a core field or domain leveraging the knowledge they have in that and going into another one.
And so it goes through scientific breakthroughs or company foundings or all this where, when you get too verticalized in some ways, you just don't have the context of the way other people do things.
And so oftentimes breakthroughs come outside.
I think work has probably gotten like a little bit too Adam Smith.
And we've, like you know, we're all just doing our little part of the paperclip factory or whatever, the you know, or the needle I think it was needle factory or whatever.
Yeah, the paperclip.
We've all, Andrew, what is it, Aaron Ross or whatever, predictable revenue.
We've done that across every element of society at this point.
Yeah, so I think probably AI relieves us a little bit of that.
Like division of labor may be maybe got just a little bit too granular over time.
And so I think maybe we get a little bit of a chance of a reset which is like no, I can just throw a little bit more of a general problem that somebody and they've got the ability to kind of go and fully solve it, which frankly I think is a hugely exciting thing and much more fulfilling probably for the knowledge worker kind of class that I think has a lot of upside.
What's the coolest or most mind blowing demo you've seen in the last call it three to six months, where you were like wow, I can't believe that's actually possible, that we can do this.
Yeah, I'm having a funny journey because it's like you do start to now, unfortunately build such high tolerance and desensitization.
It's like the hedonic treadmill in some ways, right.
You're just like, I always reference now with AI, the Louis CK bit about like airplane and wifi.
And you're just like, you're flying through the sky and you have access to the internet.
And that's the way I feel about a lot of the AI things where deep research like elucidates a little bit and I'm like well, this is bullshit, but it only saved me seven hours.
Was that 15 hours of work you would have done otherwise prepping for this interview?
Like, yes, you know?
Yeah.
So it's tough because I do wish I could.
I wish I could have my pre-ChatGPT like wiring for every individual breakthrough, because I think I would just be like, holy shit, this thing is witchcraft every day.
I remember yelling at a friend who was deep in LLMs a week into ChatGPT and I was like I'm sorry, I literally don't understand.
How did it write the business strategy on a thing that's never scanned from the internet and and like?
I was just like obviously an idiot at that point.
On on on all this stuff and um and like, but i want that, i want that like, like energy again, where i'm just like surprised by everything so um, so i, you know i'm, I think, from a pure utilitarian standpoint.
I still love the exact example you just did, which is I try and force myself as many times as possible to be like okay, deep research exists, deep research exists.
Like don't do the thing where you open seven tabs and you go through all this stuff.
It's just like, just send the query, go back to it.
I had a thing 48 hours ago.
It saved me an hour and a half um and um and like i, but i had to catch myself because i was like about to just go to do my research, my normal.
Like you know, i've been doing this for 20 years, you know i know how to research things, did deep research and lo and behold like, as far as i could tell, no hallucination i i checked a couple of the answers.
Fantastic, because it's just like boom solved the problem for me.
So I'm now doing that probably five to 10 times a week on something which is just an awesome productivity boost.
I'm still probably incrementally surprised by and excited by front end design creation with AI.
So I'll pop in a screenshot into Cursor and it codes it, or I'll pop in a prompt into V0 Replit and it produces something that I would have had to ping an engineer or a designer about and be like can you go work on this?
Unfortunately my browser history kind of looks a little unhinged because it's like 1130 pm on Saturday.
I'm like on V0 doing designs.
But that still is pretty exciting, I think, to me.
Yeah.
What do you think is more likely to be impactful in the enterprise for AI multimodal capabilities like image text code voice, or the agentic workflows that we kind of talked about earlier?
If you have perfectly executing agentic workflows, that's the holy grail.
So the amount of the, you know, my...
My workshop analogy is like in data science, as an example, and anything obviously that you put into a structured database.
We've always had this ability to be like okay, I'm gonna go compute some large set of things.
Like go run the analysis on this customer base in this region to do a thing and like we used to, you know, we still do but we like we would talk about like okay, i have a job running and and i'm gonna, you know, i'll check back in in an hour and figure out what the answer was.
Like imagine, in all forms of unstructured knowledge work, a world where you just have jobs running, where you're just like I just sent these 10 agents to go and like, figure out and synthesize all of this clinical drug trial research to give me insights.
It's like we were only wired to think about sending off jobs and compute tasks for like 3 of all corporate work.
Now imagine if you do that for the other 97% or at least 80% of the 90%.
And you're just like yeah, like I sent out a job to go research this new region, I wanna enter as a business.
I sent out a job to go and figure out what what all of my product feedback was and then build me a roadmap that I should go look at and decide if it's validated or right.
It's just going to be crazy when you can actually do full agentic workflows, whether it's deep research or executing a task or automating a customer experience.
That continues to probably be the biggest area of upside.
Two years, five years, 10 years.
Do you think the best model in those periods will be closed source proprietary or open source?
Maybe go through each, like in two years, you say closed source or open source?
I think, for lack of having any imagination on this, I think that I would mostly be in the camp of there will?
It almost doesn't matter, because within three months of the breakthrough, you'll have an open source version.
So I think it's kind of doesn't matter whether the open source version did it first you know wow, China did it, or that open AI did it, because I think within three months we can expect that you'll have the open source version of the closed source breakthrough.
So you almost so then almost by definition, you can almost still expect then a large amount of the traffic and the use will be open source, because you know you're not trading off that much, because you'll still get the open-weight version of what just emerged.
What's the most exciting box AI feature application workflow that you've rolled out that you think is underappreciated by your existing user base and everyone should really be thinking more about adopting.
There's two features we're most excited about.
The first is we have this product called Hubs.
And what Hubs does is if you go to Box and you go to any file system, you create a folder, you put files in the folder.
You share that with people.
You have this problem, which is well.
Folders are kind of like unintuitive for the recipient because they're like I don't know which third subfolder down.
Did you put the sales material that I'm looking for to use for my customer pitch?
So we created hubs, which is an overlay onto folders.
So you can put as many of your files and folders into a hub and then you can go and just ask questions of the hub.
And so this solves actually a very, very nuanced, highly important thing, which was, if you go to a general purpose just AI chat window, You have to do a lot of work to guess what stuff is on the other end of this chat window.
If you go to a hub, in our case, you kind of know by hub what does it have access to.
So we have a sales hub, which means by definition that has all the sales information.
We have an HR hub. which means that it has all your HR information.
So you go to the HR hub and you ask the HR question and then we're basically doing RAG on your documents, for all of your HR documents.
And we're doing RAG on all your sales presentations.
We're doing RAG on all your customer support product documentation.
So that basically gives you these micro knowledge portals for every topic that you want.
And so that's a game changer, because you just again the user instantly knows what topic is gonna be covered in this thing.
So you dramatically reduce the hallucination, you dramatically increase the authoritative content that will be in the hub.
So that's pretty cool.
And then the final one that is just working out kind of way better than we had hoped initially, like a year and a half ago, let's say is just AI, data extraction from documents.
It sounds like way too straightforward for anybody listening, but it's insanely powerful if all your job is is like review contracts, review invoices, pull out data from resumes, you know, standardize my financial reporting documentation.
We could just now automate, you know, basically most of that.
That's great.
Well, Aaron, thanks for doing this.
I think we've covered a lot of ground, so really appreciate it.
Thank you, Logan.
Good luck on the valuations.
Thank you.
Yeah, it's good.
