Transcript
Speaker: You're listening to From the Horse's Mouth, intrepid conversations with Phil First.
Speaker: Hello and welcome to the latest edition of From the Horse's Mouth podcast. I'm your host, Phil First. And joining me today is a very special guest. I'm delighted to get him on here. It's Aaron Levy, the CEO of Vox. Welcome, Aaron. Great to have you.
Speaker: Hey, thanks, Phil. Good to ah good to be here. Great. And just for the benefit of our audience, would you just share a little bit about yourself and your background? ah Sure. Yeah. So I'm a co-founder and CEO of of Box. um we yeah We started the the company back in 2005. And this was after a number of years trying lots of different startups. And um we actually dropped out of college. So so my my startup experience is more like in high school and ah in in college, building different ideas with with a variety of friends. and And one of those ideas that we landed on was was Box and ultimately you know what we're what we're doing today.
Speaker: Wow. like You remind me of a a guy I met recently who was a CEO of giant startup. He was about just turned 20. And I'm like, dude, did you ever go to school? yeah i ah it is It is interesting that that everybody does seem to be getting younger. like i'm i'm meeting like I'm getting emails from 16-year-olds that are looking for funding. So um i would say i we we now we we sort of have built the formula for like how to drop out of high school or college, how to how to build a startup, how to get funding. So it's been fun to watch ah the next generation.
Speaker: Great. I mean, let's see, look, you've come through the most exciting time in the technology industry. So who've been your biggest influences in your career and are any defining moments that have shaped how you think as an entrepreneur?
Speaker: Yeah, I've been i've been fortunate. you know when When we moved to Silicon Valley, um it was in the kind of, again, mid-2000s. So we were in the ramp up of the SaaS wave and kind of early days of cloud.
Speaker: um And so I've been you know ah been know very honored and and fortunate to be able to get feedback over the years from you know folks like Mark Benioff were extremely helpful in in building up Box. um you know some Some OG names like Tom Siebel or Michael Dell,
Speaker: um you know giving fantastic advice over the years. And um and so so just, I've had amazing mentors and and kind of helpers along the way. And then I'm very religious about trying to learn as much as as as possible from the history of of software and technology. So I got a lot of benefit of just you know all of the all the literature and content and books in the in the space that have documented you know, all the all the various strategies. So we've tried to put all of that information and advice to good work and it certainly helped us scale over the years.
Speaker: Right. and So you made a point um quite recently that really stuck with me. And that was we haven't removed humans from the loop that, we've just changed where they enter the loop. So as AI becomes more autonomous, where do humans create the most value and how should leaders rethink talent, accountability and decision making?
Speaker: Yeah, so um so I think what's interesting is the first the first phase of AI is isre we're kind of augmenting the work that we're already doing. you know we're we're yeah You know, that was the kind of chatbot era. um This is where, you know, you're writing code and it's kind of typing ahead a little bit faster. um You're writing an email and it kind of helps helps generate the content.
Speaker: that That leads you to maybe like a 10% or 20% productivity gain inside of a workflow, um but it's sort of less likely to completely transform you know what is the the kind of business outcome that you're getting.
Speaker: um the The closest parts of work where we've seen those business outcomes emerge is is in coding because we've moved from agents that can you know add value a couple hundred lines of of of code to now agents that can go and do you know thousands or tens of thousands or even millions of lines of code in a single run or a single session.
Speaker: And that breakthrough was basically, again, letting agents actually go off and do lots of work for you and make decisions and and execute the full plan. um But that type of work is so far, I'd say, largely bottled up in the engineering department.
Speaker: And there's some interesting kind of sort of reasons why i think that's been captured most by, but those gains have been captured most by the engineering org. um Engineering is an area where, you know, your kind of direct value creation um yeah often can be correlated to to the amount of code you write. Obviously, it has to be good code and and useful, but but it can be it can just be a lot of text that that comes into a single file and that actually correlates to value generated.
Speaker: There's a lot of work in the enterprise that doesn't have that same correlation. um Like if I wanted to go and apply automation to our sales team or our marketing team or our HR organization, that's much more about workflows that have to be executed.
Speaker: That's actually like many, many stages of a process where at different points in time, ah another person has to enter the process and make a decision or add value, or maybe you need feedback from a customer. And so those are not workflows that are as as sort of prone to being able to automate in the same way that we've seen with, let's say, agentic engineering as an example.
Speaker: um And so I think we're in for now a ah very long journey of diffusion of AI into the business. And some of the lessons will be learned directly from looking at what happened with engineering. But in many areas, it's going to actually take really reengineering the underlying business process and workflow itself to get the full gains of agents. And that's, that's I think, the the the the area where most enterprises are kind of running into into now the next phase of agentic adoption is is, okay, I actually have to go and re-engineer the workflow. I have to get data into a new environment. i have set up agents in the right part in ah in a process where they can be useful. I have to figure out where the human now is in that workflow so they can review the work and make sure the agent kind of doesn't go off the rails and and execute the wrong task or produce, um you know, yeah yeah maybe less useful outcomes or or or work product.
Speaker: that's a lot of work ahead for probably realistically like the next decade in most enterprises right right so i mean you've argued and i'll agree with you here that um ai shouldn't be funded purely from the cio's technology budget because the real value really comes from transforming business operations but From your vantage point, are you seeing enterprises really moving AI investment into operating budgets?
Speaker: Or are you still seeing CFOs, CMOs still constrained by the traditional funding models? I think i think what what happens is like, Probably the first few years of AI, it was still largely a CIO budget because it it sort of was ah ah an output of the technology organization. And so everybody sort of thought that, okay, well, the funding of Chachabuti and the funding of Claude and the funding of these agents is going to come from it t The challenge is as as you reach a certain threshold of expense where you know you're no longer paying a couple of million dollars for ai but now you're paying tens or hundreds of millions of dollars for AI, it's just unreasonable to kind of trade off ah you know that type of expense envelope with
Speaker: other more deterministic parts of your your software stack or or technology stack, like having to decide between your networking gear spend and your inference and and AI, you know, token spend is sort of, those are completely orthogonal, you know, investment areas of of a business.
Speaker: And so what what naturally happens is yeah it starts to flow into the line of business because the line of business, i.e. the marketing department or the HR department or the R&D department actually has to make trade-offs between what do they want in their organization to be agentic and what do they want to be either human labor or other areas of expense, you know, outsourcing to partners or or other organizations.
Speaker: So ultimately, this ah age agents look a lot more like how you would have funded like a bit a BPO firm in the past. then it then it looks like buying software.
Speaker: And so you you you know, the BPO firm wasn't funded by the IT department unless the the BPO is IT work. It was funded by I'm a I'm the customer support team and I want to be able to have, you know, customer support, you know, tasks be done by some agency.
Speaker: Then the customer support organization ultimately is the holder of that line item in the budget. Agents kind of have that similar product, which is, I'm the head of engineering, I have to determine, you know, given $100 of organizational spend, do I want $90 to be on people? want $95 on people? Do I want $85 on people? And then what do want do the $5 or $10 or $15?
Speaker: Do i want to put that into tokens? Do I want to put that into you know training, you know up my engineering team in different ways? That's actually who needs to like sort of hold on to the budget once you've reached a certain scale of spend, which you know agents clearly have have ended up um you know kind of hitting.
Speaker: It's really interesting. I i was with a pure play BPO firm yesterday, and I were talking about exactly this, that They were getting a lot of increased dollars coming in in areas particularly tied to revenue growth, not necessarily back office.
Speaker: And they were looking to hire like Google engineers and anthropic engineers to support business processes. The difference they're they're seeing is the business side like to pay on a as a service model.
Speaker: So they'll they'll they' like to buy on a subscription. What they don't want to do is buy millions of dollars of billing rates and consulting transformation projects. It's like, what can we plug into that's quick, that can pick up process, that can really take us forward? Are you seeing a commercial model evolve at this point or do you still think it's a kind of big mixed bag out there?
Speaker: I think it's still pretty mixed. um ah you know i think you could you could probably get in a room, let's say, with 10 lawyers and you would say, how do you want to pay for your AI? And you know five of the lawyers would probably say, I want it to i want it to be per...
Speaker: per unit of task. I want it to be per contract reviewed or per contract generated. And then the other five would say, I want this to just be the amount of usage of the system because I want to be in control of my level of efficiency and how I'm using your tool to save money or spend more.
Speaker: um and And so I think you it a lot of it is is sort of either the business model of the end customer and and how they build their customers and how like, do they do cost plus as a pricing model for their customer base? Or do they want everything kind of folded into, um into an hourly, you know, into their, into their overall service fee?
Speaker: um I think, I think there's just so many different dynamics at play here. In general, um My rough sense would be that we should expect that AI will largely converge with a a sort of a volume-based resource, you know type-based pricing model. I think it's very hard to do these kind of fully packaged up outcomes with ah except for areas that are relatively uniform across large sets of customers. So like responding to a customer support ticket is relatively relatively uniform. You know, there can be difficult tickets, there can be easy tickets, but like the economy has kind of figured out roughly what that should cost.
Speaker: Whereas, you know, going and doing a large software project is completely bespoke and a snowflake for every single company on the planet. So it's very hard to have kind of uniform outcome pricing on software projects, which means eventually it converges on some form of cost plus or volume oriented approach, um which is why I think you're going to spend your tokens.
Speaker: I think your your token pricing will largely be some kind of volume orientation on just what are you trying? What what are you trying to tackle with agents ah inside the enterprise? Interesting.
Speaker: And in terms of figuring out what to move on to a third-party model, shared model versus a private model, we heard Alex Cart coming out.
Speaker: saying, you know, you're insane if you're going to put your core intelligence out on these public LLMs. Like what Figma, you know, they got really screwed by Anthropic, didn't they, when they partnered. Anthropic figured out their business said, hey, let's go after this.
Speaker: What's your view on on that whole sort of multi-model enterprise thing? And what's your advice to CEOs on what should be going out onto a, you know, shared LLM versus something more private and sovereign to themselves?
Speaker: Yeah, i'm I'm a little bit – I'm mixed on this one. i I would say to me the issue is less about about the sort of shared nature of kind of learning specific things and it's much more about do you need extra performance on a particular task or dimension that the that the horizontal models or the closed models are not able to support?
Speaker: So I would be less sort of worried about the models themselves kind of training off of you because you can you know generally you know sign contracts with all the model providers to prevent that from happening.
Speaker: um So so i i i like even the even the Figma you know kind of issue is a little bit unique because um there were dynamics where just a board member was was sort of overlapping um as it was less about the model itself kind of being trained on the data.
Speaker: So so i I think the way I would think about it is um If you are able to to get high quality outcomes at an an affordable price from frontier AI models, whether the closed or open kind of, you know, doesn't really matter.
Speaker: i would still currently bet on that because because what's going to happen is you're going to just going to see, you know, continued amazing progress from those these models. um And they're just going to get better and better and better and better. And so you want to kind of ride out that improvement curve.
Speaker: If there are some areas where the cost profile makes it just completely impossible to use that frontier and intelligence, or where you need the model to be trained on your set of data for some very specific type of task, like you're doing fraud detection at a large bank, um or you're doing clinical life sciences research and you need to give the model a lot of bio data, those might be areas where training your own model or having some kind of sovereign intelligence can be useful.
Speaker: But you you're making ah you're making a a judgment call, which is you actually want to now sort of fork the model and go off on your own direction, which means you're not going to get the same benefits of the concept model progress that that's been happening.
Speaker: And in a lot of times, that makes complete sense. There's a lot of software providers where if you're Sierra or Decagon or or kind of one of these customer support agents, there's a lot of use cases where you have so much data about your tasks type that it makes more sense to train your own model as opposed to use something off the shelf because you can get it to be both cheaper and higher performance.
Speaker: But then there's a lot of industries where it's just it's just premature to to go in and sort of veer off from this frontier set of progress because you actually still need the ultimate gains that are left to come from these AI models.
Speaker: um And you want to be able to ride out that that progress curve ah you know quite a bit more. So I think there's a judgment call that has to be made on each individual organization. The thing that that I think is is critical and non-negotiable is you should find ways of separating your corporate intelligence and your corporate knowledge and your corporate intellectual property from the model.
Speaker: So that way you have the ability to swap out models as new use cases arrive or as new models you know kind of are are able to have greater capability. so There's this really interesting new era and art of of of basically IT t architecture in in a world of AI, which is how do I ensure that I get all of the ongoing gains from technical progress in the AI space and not sort of accidentally sort of force fit my workflows or my business process or my knowledge work into one particular model paradigm? Having that abstraction layer ends up being incredibly important, I think, for most enterprises.
Speaker: Yeah. Interesting. So your company, Bucks, you've spent what a couple of decades organizing enterprise information. So do you see the next decade just being around this enterprise intelligence organization where content you know becomes organizational memory and agents continuously reason over it?
Speaker: Do you think that's that that's going to be the conversation or do you think this is going to evolve into something different? um Yeah, i think I think that's exactly, i think it it would be the former of what you just proposed, because what's what's happened is we spent a couple of decades trying to to manage and govern and store and secure all of this corporate information.
Speaker: And we've never really quite known what's inside of this data, ahss what's inside of our contracts, what's inside of our financial documents, what's inside of our research materials. The only time you really get to find out is when you load up a document, you read it, and you search and you find the thing that you're looking for.
Speaker: But for the first time ever, we can now throw computers at all of this data and better understand it and help us reason over this information and make better decisions and use all that knowledge to onboard an employee or onboard a customer far faster.
Speaker: So it's kind of the first time ever where we have had this breakthrough where we can finally tap into this gold mine of information inside of our enterprise, which is all the digital data that we've been creating for decades.
Speaker: And now we can finally actually ah treat it like like you know or an actual resource that we can use computers to to better understand. So I think the next couple of decades is all about now how do we finally tap into the full value of this information? Now, there's a little bit of a gotcha, which is most enterprises don't yet have their data in either the right format or right environment to fully take advantage of that. We go to a lot of organizations where there's lots of legacy technology, lots of legacy infrastructure, lots of systems that don't connect well to agentic systems. They don't support MCP.
Speaker: They don't have modern APIs. So there's actually a lot of modernization work that has to happen in most organizations. what's What's going to be interesting is it's going to be akin to what we saw in the structured data world.
Speaker: So the same kind of wave of growth and and kind of um dynamic environment that we've seen in places like Databricks and Snowflake, I think it is going to happen in the unstructured data space because companies are going to realize that actually all of that unstructured data has a tremendous amount of value associated with it. And we need to be able to secure it. We need to govern it. But now we need to bring intelligence to it.
Speaker: um And that's going to be ah just a ah huge opportunity going forward. Right, right. So on that line around software applications, um you know, we've been working through our apps for decades, and and now ai AI agents are increasingly interacting with APIs and business logic.
Speaker: Does the app itself become less important? And and what separates those software companies that will become indispensable AI platforms from those that simply become infrastructure, Aaron?
Speaker: Yeah, i think um I think what's going to happen is um if if your core value proposition was was sort of maybe like too much at the user interface layer, and it was about the knobs and the and the and the graphs that you you interacted with, I think that's going to get kind of compressed by agents because agents will just generate a lot of that on the fly.
Speaker: But if your business if your value proposition was you know the the workflow, the business logic, the the management in a secure and governed way of the data in the process, I actually would i would kind of flip it. And i would I would argue that agents actually make all that more valuable because agents are going to need to tap into that data.
Speaker: They're going to need to treat that data as as their natural resource to work off of. And so the systems in can provide the guardrails and that that can um effectively kind of traffic cop access to that information. I think those will be the most valuable systems in the future.
Speaker: um And so so this is why I'm i'm i'm you know pretty bullish on things like your core CRM system, your core yeah ERP systems, um you know core systems of record that have house and manage and maintain the the structured or unstructured data you know relevant to a critical business process.
Speaker: I don't think those systems are going away anytime soon. And in fact, if anything, I think they actually become more valuable with agents because agents can now use this data at like 100 times the scale of what people actually did.
Speaker: So I share this example of ah you know a bunch, which is I'm actually now a much more active user of Salesforce because of agents and Because I just and i just entered MCP into Salesforce via Claude Cowork.
Speaker: And so so now I can talk to CRM data, where previously i actually rarely went into Salesforce except for to see like the dashboard of like how we're doing in a quarter. But now I'm like having agents go off and process and and understand certain sales territories and market opportunities and look at certain verticals and where where we could be doing better. i didn't do that before. So I'm actually more actively using Salesforce because of MCP and because of agents than I was previously. And I think there's a lot of areas of software where you're actually going to see the utilization just go up significantly because of what agents can do.
Speaker: Yeah, it's it's funny because I run a research company and our research readership has just gone off the stratosphere in the last year because of people accessing it through per perplexity and flawed and these LMs. And I'm like, I don't really care that much.
Speaker: if they're not going directly to our website all the time, as long as they're getting our research, because ultimately they'll call us up and they need help. And that's how we make money. so what Well, that's that's not right. And what's interesting is, um is you know, there's this there's this term now in software, which is build something agents want.
Speaker: I think that's going to be true of of all forms of business, which is agents are going to be mediating a lot of the information that gets discovered, a lot of the data that gets pulled, a lot of, you know, tools that get utilized. Yeah.
Speaker: And so so we are going to enter an era where you know if you're a consulting firm, you probably want to make sure that agents you know know about your existence and have access to your proprietary you know information because that's going to market the firm.
Speaker: If you're a technology provider, you probably want to make really, really good APIs that agents are going to be able to leverage because you're that's going to expose your tool set to agents while they're building an application or running a workflow.
Speaker: So it's a kind of a new way to think about building software or even marketing yourself, because now the agent is the one that's going to be doing the discovery. The agent is the one that's going to you know identify that the customer actually wants your your set of services.
Speaker: um it it It changes a lot of the underlying kind of marketing and and commercial models of information based services. yeah And on a similar vein, mean, terms of professional services companies, you know, there was a lot of doom and gloom in the services industry that AI was just going to wipe these guys out. But it really starts to seem like the need to integrate these systems, govern AI, have access to talent is is off the scale as well. How do you see consulting and tech services evolving um alongside this this trend? Yeah.
Speaker: Yeah, I think you know this one is is, it can be a little bit complicated because because there's many different kind of layers of the stack and in consulting and services. i I kind of lean by default optimistic, but I do know that there's going to be some transformation that happens you know in in in various you know kind of tiers of business models.
Speaker: um For instance, I think I'd be very optimistic that right now there's going to be a ton of work for the implementation, management, and transformation of workflows with agents.
Speaker: And that's going to often take consulting firms and and system integrators and technology kind of service providers to help with that journey. So i you know we we see, especially these kind of new, modern, like more forward-deployed, engineering-oriented SI firms ah that are actually helping actually you know go re-engineer the workflow for agents. You know each relying on every single enterprise to go and do this themselves and and and learn all the all the seller failure modes themselves and run into all the same problems. That's just not very efficient in the economy. So that's why services firms exist, because they can kind of get those best practices to be disseminated across a large body of companies.
Speaker: So I think there's going to be a lot of opportunity there. i think there's going to a lot of opportunity if you were a very forward-leaning services firm and you got very agent-pilled and you ensured that your engineers were using agents better than everybody else, then what you would be able to do is actually bring down the cost of projects or take on even more projects or do much bigger projects um than what you would have been able to do previously, which would make your economic value to the customer go up.
Speaker: So I think there's like, if i was if I was running an engineering services firm, and previously, you know, I used to have to say that that one project cost a million dollars, and it took a year, I would say, okay, that project is going to take three months, and it's going to cost a quarter of a million dollars.
Speaker: And immediately, you might say, okay, well, then you're going to dramatically shrink. But if you' if you're, you know, kind of executing this properly, you'll just find way more customers now, because you were previously priced out of being able to serve large amounts of the economy because only the biggest companies were ever able to work with the Accentures and Deloitte's of the world.
Speaker: But now actually way more people are able to go and leverage these types of services from these vendors. So I think there's going to be actually a lot of opportunity by by just really being good at using AI and then and then being able to deliver that on behalf of your customers in a way that has reduced their risk dramatically than them having to go and do all this themselves.
Speaker: Yeah. and And what's your advice to services companies in terms of creating the right pool of talent? Because, you know we've got some services firms, which are still very software engineering focused.
Speaker: Then you get others, you know, like your Accentures and Deloitte to have a mix of business transformation and tech transformation. Who do you think is going to win out in terms of providing talent that can take clients forward?
Speaker: i think um I think it's going to be super important that you are, you know, educating your next set of, of um you know, of workforce. We do a lot. We spend a lot of time really in in sort of this modern form of training and teaching and and kind of shadowing. And and it's it's sort of like...
Speaker: its Training is almost the wrong word for it because it's just constant best practice sort of dissemination on using agents and using AI. um There is some kind of training kind of in a structured way, but but a lot of it is just is building communities and and knowledge centers internally so everyone can kind of rapidly share with each other on on how they're getting better at ah using different tools.
Speaker: um And I think that if you're running a services firm, you probably have to do the exact same thing. Like, again, if if I were running a services firm, then the most important metric right now would be, can I get the entire organization to become experts in how to use agents to be able to execute on client work?
Speaker: Because the client is going coming to us for that expertise. So we better be the best in the world at being able to do that. and ensure that we are never falling behind on any dimension that that's happening in the market.
Speaker: That would be very, very critical. So I think that's going to be true of, again, any kind of services firm in the 21st century is you need to be an expert at these technologies. Terrific. That's great advice.
Speaker: So you've led, Aaron, you've led Barks through cloud, mobile, content, collaboration, now AI. So looking across all these tech shifts, What have you learned about separating genuine transformation from hype?
Speaker: And what advice would you give the next generation of technology leaders? Yeah, you know, it is interesting because there is, um for for certainly a few of those, and and it's sort of the ones you didn't mention are the ones to to kind of watch out for, of of you know, what trends in that 15, 20-year period but but didn't end up playing out. Those are the ones that did. ah And and the the question is, how do you kind of decipher what's working, what's not working, what's going to kind of cross the chasm versus not? you know i'm I'm a massive fan of Jeffrey Moore um you know coming up with the the idea of crossing the chasm based on the technology kind of adoption lifecycle and diffusion sort of model.
Speaker: and And, you know, every technology, there's every every technology that's ever been created, doesn't matter if it's like the weirdest technology of all time, there's always like 5% of the population that will adopt it. That's just guaranteed.
Speaker: There's an early adopter for every technology and tool on the planet because because there's just like a bunch of nerds like like me or us that that want to play with new things. and we we and And so it can be very confusing because you might see, like I buy every new technology that exists.
Speaker: But that doesn't mean that I then think every single new technology that that I buy is going to kind of take over the world. Like I very quickly eliminate out the things that that that I just don't think really make sense. And I often need to be like, I often need to sort of pressure test by like talking to like my wife or, or you know, friends that are outside of tech or family members to to be like, okay, am I...
Speaker: Am I being weird right now? or and And then I try and kind of like you know sort of ah calibrate myself based on my excitement and interest in in different tools.
Speaker: um And so i've I've done that just literally my whole life. And I and and i am the calibrator for other friends. And and so you kind of like build a... sort of almost a an intuition of, okay, this type of thing is is like really cool, but it's way too hard to adopt.
Speaker: And so in this current kind of package, there's just no way that it's actually going to take off. um or Or the opposite. It's like, i'm you know you play with something and you're like, I'm literally, I'm looking at the future.
Speaker: I remember the day that the um iPad came out um yeah i I went and bought an iPad and I was just like, oh my God, this is this is obviously you know going to be incredible because it's just like this magical device. And ah same with the iPhone because I had owned a BlackBerry and BlackBerry was like, this is amazing. But like but like you know you run into all these problems with the BlackBerry. It doesn't you know do apps very well. It's kind of clunky to use. And the iPhone was just like, oh my God, this is going to be totally world changing.
Speaker: And, you know, Steve Ballmer was like, oh, it doesn't have a keyboard. Well, as an active user of the iPhone after a week, I was like, it doesn't need a keyboard. It's fine. Like people will be fine. on So, so I use that to say like, I just, I use every technology and, and with AI,
Speaker: that that, you know, AI has these versions of what things are being hyped and what things are real. And it it is kind of hard to process. I think there's a lot of things that are happening in the engineering world with ai that probably won't be able to kind of cross the chasm in the same way that that that it's worked for engineers.
Speaker: um A lot of the way you set up agents to be able to do sort of long running work, you know, the the sort of open claw style. I have an agent that has access to everything and I can do whatever it wants. These things are very, very hard for kind of like average people to go in and administer and govern and set up in a secure way.
Speaker: So they have to be packaged far simpler than than they've been packaged so far. um Even things like skills files and agents.md, these are things that are that are actually quite esoteric for most people.
Speaker: um and so And so we're going to have a hard time. It's going to take a while for heat for like regular knowledge workers to to kind of grasp this idea that that your agent is not just an extension of of sort of you that is sort of just doing small little things that you're asking it to do, but it's kind of its own entity and it has to keep its own set of notes and its own set of instructions and its own set of decisions.
Speaker: That's kind of a mental frame that most of the economy and most workers don't yet really have. um And so it's going to take a while for that to diffuse across the across the ah you know kind of economy.
Speaker: um So I think we're in for, again, years and years of this sort of diffusion of AI, um as as we go from these kind of more engineering centric use cases to the rest of knowledge work.
Speaker: Yeah. Yeah. I think it's only a matter of time before companies who are getting really experienced with AI will just say, this is how work gets done. not going to call it AI anymore. This is how work gets done.
Speaker: So look, final question. You've been at this for two decades, right? Multiple technology waves, market cycles. What advice would you give to young entrepreneurs today? What qualities or habits do you think matter most if they want to build an enduring company rather than simply ride the latest wave?
Speaker: Well, i you know it it in our case, I'm i'm a little bit biased because it helped that we we focused on something that was quite um ah quite timeless as an idea. like what what The reason we started Box was we wanted to make it so you could securely access your information from anywhere. And that that's that's quite timeless. like There's nothing that can that can truly end the need for that type of technology. There's been lots of competition over the years. We have to kind of fight in these mega battles with other companies, but it's a timeless idea.
Speaker: ah And so I think i think you know trying to focus on things that that have very, very long time horizons that are technically extremely difficult.
Speaker: So that way you have time to kind of build up a strong moat and um and have have something that really is kind of defensible. Yeah. Doing something, obviously, that you're excited by and passionate by. I think there's a lot of companies and and founders that get burned out by what they're doing. and And you kind of ultimately diagnose it that it just wasn't that interesting of a problem for them to go solve.
Speaker: um And then ah you know I got lucky because because I got to found the company with a ah handful of friends. So choosing your your founding team ends up being really, really important. um but But I think you know very big problems, ideally those that are timeless, timeless ones that have very strong, hard technical notes associated with them, and being able to build a team that you really want to go work with, those are you know probably four of the biggest things that I've i've found to be very important.
Speaker: Fantastic. Aaron, I can't thank you enough for the time today. It's been wonderful hearing you cross the chasm from services and software and AI and everything. So I really appreciate the time.
Speaker: Look forward to airing this with our with our network. It's been great having you Thanks, Phil. Really appreciate it. Yeah, yeah. Thank you for listening to From the Horse's Mouth. Don't forget to subscribe and like wherever you listen to podcasts.
Speaker: Got something to add to the discussion? Drop us a line that from the horse's mouth at at hfsresearch.com or connect with Phil on LinkedIn.
