Aaron Levie's advice for founders and startups

On numerous occasions, I’ve seen funny tweets from Aaron Levie, the CEO of Box on my Twitter feed, but I had never heard him speak. I heard this podcast where he talks about his views about AI and the use cases at Box. While that was interesting, the highlight, for me, was his advice for startups and founders. There are 100 ways to heaven, and you might disagree with it, but I still found it pretty sensible and something to think about

Here’s some loosely edited (please don’t sue me) snippets from the show.

How would you start a company as a young enterprise founder today?

Aaron Levie of 2006 is in 2023, and you have the urge to go start a company in some way. How do you even go about doing that?

Aaron Levy: I’m pretty confident it would be an AI. But I mean, this is how we did it for Box. As rigorous as 19 and 20-year-olds could be on sort of business strategy, we did the work. We looked at the market, we analyzed the market, and we tried to figure out our competitive advantage. We attempted to assume what incumbents might do in the market and how we would compete with them.

I think that there are some situations where you just totally capture a Zeitgeist moment, like Facebook or a product like that, where you build something for fun and it works and it takes off and takes over the world. There’s that category, and I would say, at your own risk, pursue that approach. Then there’s the category which is like “Hey, we analyzed the market and we think there’s an opportunity if we build better software for this,” and then let’s go and create this thing. That that’s the approach we took.

I think AI probably requires a little bit of that as a result of the incumbent advantage that does exist in this space, so like, if you just go and say, “Oh, I built this really cool thing,” and obviously that it’s amazing, and I’m gonna pursue this as a startup opportunity, well that cool thing again is like right head on in front of Salesforce and their AI strategy. I would just say that’s going to be very, very hard to scale because Salesforce is going to get very good at this, and the same would be true of, say, Atlassian, ServiceNow, or whatnot.

That being said, I think there are a lot of spaces that are either in the blind spots of those incumbents or areas where the business model of those incumbents is in sort of direct conflict with pursuing that opportunity. You know, there are a lot of use cases where you have an enterprise software company that sells seats of software, and they’re not going to be as good at going after use cases that either are just that you only need one seat of the software to do the whole thing, or they actually reduce the seats of the software by virtue of using it, and so find opportunities where there’s some kind of inherent innovator’s dilemma challenge with that incumbent.

Operational advice

I think we’re always kind of improving many dimensions of it. So you know, there’s like Apple-level discipline, and then there’s like some large gap, and then there’s us. So there’s a lot of distance from, let’s say, that level of intensity. But if you distill it down to the component parts and you say there are KPIs that are important KPIs to the business, like cost per employee, your pipeline, how much you’re spending on R&D as a percentage of revenue, and 30–50 other important KPIs. We have a very good handle on, I believe, all of them, to the ones that we can forecast and understand. So having very, very good people in finance and financial planning and analysis—that’s a function that, when we started the company, we wouldn’t have been able to spell. Now, it’s like we run the business with with that kind of discipline. So having a very, very, very strong Finance team is incredibly important for literally any company that has probably passed like $10 million in revenue and can understand all these KPIs.

I think being extremely rigorous on your investment decisions, and there’s only like, I think, only two major problems financially that companies really can get into in sort of startup software land that have any remote success.

  1. Either you over-invest, and that could be too many growth initiatives, entering too many markets, too many product priorities, or any form of over-investing with the hope or bet that it’ll all work out. That’s like one big category of mistake

  2. Another big category mistake is that the thing you’re doing is just literally not working, the product is failing, you have to pivot, or the market has changed. Then what will happen is your revenue goes lower, your costs have maintained, and then you’re kind of screwed. So those are the main failure modes past product market fit.

So operational discipline, I think, is just some version of avoiding those two outcomes.

How do you not overinvest? You have a sequence or have a set of heuristics when we enter a new market, like:

  1. How profitable does our core business have to be before we enter new product categories or new geographies?

  2. How many new product categories can we afford to manage at once that are net new?

How do we make sure that we don’t dilute our efforts by being in five product categories instead of two?

It’s all those kinds of things, and that’s why it’s hard to be generic because every business has a different version. It’s usually stuff like that. You need to have a deep, deep understanding of value creation in your particular category. Where we think a lot about and where our actual literal points of leverage are. So you can have 10 engineers build this thing that enables the whole business to improve, instead of 50 engineers going off and building all those other things where we don’t get any leverage from that kind of investment. So we’re thinking a lot about where the points of leverage are. What things require the smallest investment to the greatest output, and we’re always sort of squishing all that information together, and that’s our strategy basically.

What’s a piece of startup advice that you hear out there that makes you want to pull your hair out because you think it’s so wrong?

That’s really funny. I like that one. So I think probably the one that I hear more often, and then I get confused by is, and I don’t know because I think this stuff goes both ways depending on literally who’s sending the advice. But you know there’s all this fake nothing. I’ll give you two categories. I think there are a lot of ways that we simplify these points of scale, like from zero to 10 million, 10 million to 100 million, or this is definitely going to change at this level. And I’m like, “No!”

We’ve had employees that went from exactly zero to like $500 million in revenue. We’ve had processes that went from zero to $500 million in revenue. We’ve had processes that went from zero to a billion dollars in revenue, and then we’ve had things that have had to change at like 50 different steps along the way. So I think there are all these magical, very simple, neat, sort of like step functions that we kind of simplify advice in, and I tend to disagree with a lot of that. The first person we hired, the first human that sold software at Box, took us from zero to 100 million as the head of sales, and so that advice is actually just not true.

Then some people will say things like founders need to be responsible for the first X number of million in revenue or whatever. You’re like, “How do you do all the selling yourself?” or whatever. Well, we actually had a head of sales that was way better at it than I was, and very quickly I realized I shouldn’t be in any sales conversation because I was literally 22 years old, and I would have been a liability for the deal if anything. So yeah, more of those kinds of things where you try to oversimplify and overgeneralize.

I’ve just seen too many different ways of succeeding and failing, just from friends. You’re like, “Wow, that person has never been on a sales call in their life, and that business generates billions of dollars in revenue, and then that person sells all day long, and it’s not working.” Ironically, my solution is to tell everybody to read everything possible because you might as well at least train your human large language model (LLM) on everything possible because you might see scenarios emerge where you could be like, “Oh, like that’s when Salesforce decided to invest in this strategy” versus that’s when you know Twilio or Stripe decided to do this approach. Learn everything you can, but like the idea that one of the paths is the right one, at least a priori, is not true.