May 22, 2024 · 59m · news
Aaron Levie: How the Business Model of SaaS Changes Forever & Startups vs Incumbents:Who Wins?|E1155 · 20VC with Harry Stebbings
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of 20VC, Harry Stebbings interviews Box CEO Aaron Levie to explore the massive architectural shift of generative AI, examining how it transforms legacy SaaS pricing models, redefines the competitive battlefield between startups and incumbents, and reshapes corporate organization structures.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Harry holds 16.4% of the talking time here. How this is scored →
Harry as informed peer 4.4 Guest teaching 4.3 Guest disagreement 2.1 Harry pushing back 2.9
speaking balance: gold is Harry, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 14:46 Aaron interrupting to defend org structure stability
Aaron forcefully interrupts Harry's premise ('Sorry, sorry, sorry...') to reject the idea that AI immediately eliminates entire functional departments like outbound sales.
Hardest push from Harry ▶ 57:03 Harry confronting Aaron on Box's valuation multipleHarry bluntly refuses standard PR answers by pressing Aaron on Box's $3.89B market cap relative to its $1B revenue, arguing the multiple should be at least double.
Biggest teaching moment ▶ 7:27 Aaron clarifying context window scaling vs Moore's LawAaron educates Harry on the difference between chip density and context window scaling, highlighting a 500x improvement in 18 months that eclipses traditional compute laws.
Harry holds his own ▶ 41:14 Harry citing Accenture vs OpenAI financial metricsHarry cites his own viral thesis backed by real revenue numbers ($2.4B for Accenture vs $2B for OpenAI) to prove services outpace foundation model revenue.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Harry as informed peer | Guest teaching | Guest disagreement | Harry pushing back | Why |
|---|---|---|---|---|---|---|
| Warm Welcome and Twitter Fanboying | 1 | 3 | 1 | 0 | Harry opens with casual praise for Aaron's Twitter presence before asking a broad question about surviving tech transitions. Aaron provides a structured overview of historical platform shifts every decade. | |
| Will Startups Win the Foundation Model or Application Layer? | 3 | 4 | 1 | 1 | Harry asks whether value will accrue to foundation models or application layers and inquires about Box's model-switching architecture. Aaron explains why model personalities prevent total abstraction and why hyperscalers commoditize the model layer. | |
| The Death of On-Premise: Why AI Forces Data to the Cloud | 4 | 6 | 4 | 3 | Aaron directly counters Harry's premise that enterprises are returning on-premise out of AI security fears, stating he sees the exact opposite. Harry also asks if rapid context window gains discard Moore's Law, leading Aaron to clarify model algorithmic progress versus GPU chip density. | |
| AI Agents: Moving Beyond Chat to Autopilots | 4 | 6 | 3 | 4 | Harry pushes back on Aaron's AI agent definition by asking if it is simply traditional RPA. Aaron validates the functional similarity while detailing how LLM general intelligence overcomes traditional RPA fragility. | |
| How AI Redefines Organizational Structure and Headcount | 6 | 5 | 4 | 7 | Harry forcefully challenges Aaron's assertion that org structures won't change, citing Aaron's prior podcast comments and arguing team sizes will collapse from 20 to 2. Aaron interrupting to defend his view that productivity gains lead to reinvestment rather than permanent headcount reduction. | |
| Experimental vs. Production AI Spend in Enterprises | 4 | 3 | 2 | 2 | Harry asks sharp VC questions regarding enterprise experimental versus production AI budgets and the disruption of seat-based SaaS pricing. Aaron admits pricing units are uncertain and outlines emerging consumption-based experiments. | |
| The Battle of Incumbents vs. Startups and the Innovator's Dilemma | 3 | 4 | 1 | 1 | Harry asks how incumbent software vendors will fare against vertical AI startups. Aaron outlines the Innovator's Dilemma, explaining how seat-based business models leave incumbents vulnerable to AI-native labor replacements. | |
| Fearing OpenAI: How Startups Survive the Platform Threat | 5 | 5 | 3 | 4 | Harry quotes Sam Altman's threat to steamroll startups and cites Duolingo as an example. Aaron advises incumbents to abandon inferior custom models rapidly in favor of superior third-party LLMs to survive. | |
| The Speed Advantage: Scaling Alignment and Avoiding Google's Mistakes | 4 | 4 | 1 | 1 | Harry asks how Aaron maintains startup velocity at Box's scale. Aaron details the challenges of organizational alignment and uses Google Gemini's flawed launch as a case study in large-company product execution gaps. | |
| Is Google Smashed It? Assessing Google's AI Strategy | 5 | 3 | 2 | 2 | Harry critiques European AI regulation with witty sarcasm about taxing residents, then asks about regulatory risks. Aaron outlines why surgical IP and national security regulations are healthy while dismissing extreme pause-AI movements. | |
| The Future of RPA and Enterprise Change Management | 4 | 5 | 3 | 2 | Harry asks if legacy RPA players like UiPath are rendered obsolete by AI agents. Aaron reframes the premise, arguing RPA incumbents are well-positioned to capture an expanded automation market if they execute well. | |
| Consulting and Services: The Looming Boom in AI Implementation | 7 | 4 | 1 | 2 | Harry demonstrates strong domain expertise by citing specific financial figures comparing Accenture's $2.4B AI revenue to OpenAI's $2B. Aaron validates Harry's insight while adding that long-term software/infra spend will eventually eclipse implementation services. | |
| Unlimited Supply and the Threat of Digital Noise | 4 | 4 | 2 | 2 | Harry raises concerns over infinite AI supply eroding value across outreach and media, revealing his podcast intros use synthetic voice clones. Aaron points to historical creative destruction and power-law distribution dynamics. | |
| Quick-Fire: The Art of Delegation and Micromanagement | 5 | 3 | 2 | 4 | During quick-fire questions, Harry challenges Aaron's delegation struggles by bringing up Jensen Huang managing 60 direct reports without delegating. Aaron defends selective micromanagement in core product areas. | |
| Founder Personal Branding and a Focus on Software | 4 | 4 | 2 | 3 | Harry asks about founder branding, board wishes, and Apple's AI positioning, challenging Aaron with rumors of Apple partnering with OpenAI. Aaron explains why partnering fits naturally into Apple's platform architecture. | |
| Box Regrets: Embracing Capital Constraints and Cash Flow | 7 | 5 | 2 | 8 | Harry bluntly confronts Aaron regarding Box's public market performance, pointing out that $1B in revenue resulting in a $3.89B market cap represents a depressed 4x multiple compared to 8x standards. Aaron accepts the critique, acknowledging growth transitions in public markets. |
Statements from this episode (41)
Levie: Incumbents will capture AI value as effectively as startups
“That is as much going to be an incumbent's game as a startup's game this time around.”
Levie: Tech platform shifts are sole window for new franchise companies
“Each one of these moments, there's an architecture shift that happens in tech that creates Really the only window of opportunity you have for new kind of platform scale, you know, large franchise kind of companies to emerge, because you need a change in the te…”
Levie: Zuckerberg is spending billions to commoditize the AI model layer
“And much of that is, is due to the trends we've already seen, which is, you know, the moment you have, you know, people like Zuckerberg that are literally willing to spend billions and billions of dollars commoditizing the model layer,”
Levie: Only 1 to 3 independent horizontal LLM startups will reach scale
“But I think for like the pure play horizontal LLMs we, I think those will largely be subsumed by the big players with maybe room for one, two, three independent companies at scale that are not in the hyperscalers, but there will not be room for 50 companies. T…”
Levie: Users Will Wire Into Specific AI Models Rather Than Abstracted Selection
“I don't think you're going to have complete commoditization of sort of the personality of the models for the sort of style and the response to the point where then, you know, if you're a user, you know, any given response could come from Gemini, Versus GPT-IV …”
Levie: AI is the final death knell for on-premise data storage
“AI is kind of the final I would say maybe death knell to that particular approach. If you're, if your data is not in the cloud or in a cloud ready form, I think it's gonna be extremely hard for you to get the full value of AI on your information.”
Levie: AI context windows grew 500x in 18 months
“The token context window was more or less, I think, 4000 tokens. So you kind of give it like a blog post to analyze just in the past, you know, week at Google IO Sundar announced a two million token window model or version of Gemini. So think about that. That'…”
Levie: GPU performance progress is outstripping Moore's Law
“We have seen faster than Moore's law improvement on GPU performance.”
Levie: AI model progress shows no signs of plateauing
“But I would say I, I'm seeing no signs of AI models, slow down, anything plateau in terms of the innovation curves that that we're seeing.”
Levie: Chat interfaces merely turn graphical user interfaces into command lines
“Chat is more of like a UX paradigm shift. So it takes, you know, what used to be a graphic, you know, graphic user interface. And turns it more into a command line.”
Levie: Traditional RPA is frail and fails to handle business variability
“The challenge with RPA is, you know, RPA is relatively frail. It's often sort of, it's like looking at your computer screen and, you know, performing some kind of wrote, you know, routine actions doesn't handle variability very well. Because there's again, doe…”
Levie: AI agents will act as enterprise autopilots, not co-pilots
“When you think about now an organization that has, you know, effectively AI agents that augment the labor or the people in the organization, they're not co-pilots. They're literally autopilots that are going out and doing work for you. That's a, you know, Pret…”
Levie: Corporate org structures will survive AI disruption intact
“I think org structure stays the same. I think what happens is, is you now have the ability to sort of drop AI labor into any part of that org chart, And in fact, in many cases augment, you know, just whatever you would normally have had people go do.”
Levie: Box reinvests in headcount whenever AI boosts function productivity
“At least so far empirically, every time that we've done that, we will actually reinvest in that area of the business.”
Levie: One-person billion-dollar AI startups will be rare novelties
“There's a lot of belief that like, oh, well, well, you might have these solo companies that, you know, become multi-billion dollar companies with just one person or, you know, three people and a bunch of AI. I think that will be done, you know, just by virtue …”
Levie: Most enterprises have both active AI experiments and production deployments
“The vast majority of companies have a meaningful number of AI experiments happening with lots of different, you know, kind of areas of their business, lots of different applications. But the vast majority also have, you know areas that are in production alread…”
Levie: Bespoke value-based pricing is nearly impossible to scale in software
“That one's tough to scale because everything's bespoke. So let's kind of throw away the value, the kind of bespoke value oriented approach, because that does never, you can, you know, Except for maybe like a Palantir or something, it's very, very hard to kind …”
Levie: AI software pricing will standardize around consumption-based volume metrics
“So so eventually you're gonna have some unit that, that we all agree with that, that you can go off of. So is it sort of a consumption, you know, model of what is your unit of volume?”
Levie: Dozens of category-winning AI labor startups will emerge within five years
“And this is a, this is again, one of these windows Where 10, 20, 30, 50 companies will get started that, that were like the window in two, in the mid 2000 where like all of today's SaaS companies basically emerged in like a five year period essentially. And we…”
Levie: AI startups will win where incumbents face Innovator's Dilemma business model traps
“Similarly in AI, I think there'll be some categories where incumbents have the natural advantage, but then there will be either blind spots that those incumbents don't kind of go after, or there will be things so different from their existing business model in…”
Levie: OpenAI will become a universal assistant interface across audio, video, and text
“ChatGPT is going to be this universal sort of assistant interaction, you know, interface, and all the, you know, subsequent tools to manage that and interact with different AIs that kind of come together there. And then they're going to have an API business th…”
Levie: ChatGPT will not replace dedicated language tutoring apps like Duolingo
“I don't think ChatGPT itself is going to become a language tutor. I think it'll have the full capability to be a language tutor at the model level.”
Levie: Incumbents must adopt superior third-party AI models to survive
“If what they have is better for the product that you're trying to deliver to customers, that is the only way that you're going to survive this.”
Levie: Google now has religion on AI
“And I think, you know, Google IO is merely the, you know, another moment of it manifesting that, that, that, that company has religion now on AI.”
Levie: A six-month AI pause cannot resolve irreconcilable AI safety philosophies
“Six months is not going to solve this. There are very fundamentally different philosophies in, in the land of AI that, that, that are irreconcilable. They will not, they will never fit together.”
Aaron Levie: Incumbent RPA vendors are uniquely positioned to win in AI
“Yeah, I mean, actually, I think being an incumbent RPA vendor is, is actually a great spot because if you already are talking to customers about literally automating business processes and workflows, and there's just a better way to do that and there's nothing…”
Levie: The workflow automation market could grow 100x due to AI
“So the I think the thing that will 100% guaranteed happen, like in five years from now, this will be the most obvious thing of all time. But when you look at like RPA, you had to be, you know, a relatively deep expert in, in RPA. You had to be like a midsize o…”
Aaron Levie: Human change management, not tech, is AI's bottleneck
“Yeah. So this is the, I mean, probably the long pole of all of AI right now is not going to be the technical breakthroughs. It's going to be the implementation and change management on the human side.”
Stebbings: Accenture posted $2.4B in revenue while OpenAI posted $2B
“Accenture posted 2.4 billion in revenue and OpenAI posted two billion in revenue.”
Levie: AI services and implementation spending will be massive over next five years
“So, so I think I would take the bet on, on AI services for the next five years, unquestionably. The amount of dollars that will go into the change management of systems, the implemented, the implementation of the technology is, is going to continue to be massi…”
Levie: AI software and infrastructure spending will eventually outpace human services
“At some point, the actual AI, sort of, the software services of AI, this, the infrastructure services of AI will eventually exceed the human services on the implementation simply because now once it's in production, you don't need that same change management 1…”
Aaron Levie: AI agents will lower startup barriers like AWS and Stripe
“If Shopify caused businesses to get started because it lowered the barrier to being able to sell online, if AWS caused applications to get started because I was like, oh, I could just build an app and run it in the cloud. I don't have to think about servers an…”
Stebbings: 20VC podcast intros use AI voice clone instead of real voice
“We don't, people don't know this, but for our intros, it's not my voice, Aaron. Like, not at all.”
Levie: Existing legal frameworks are largely sufficient for dangerous AI use cases
“I think there are real reasons we should, you know, sort of pay attention to some of the more you know dangerous use cases of AI. I just think we have in many cases sufficient Legal frameworks for addressing those things.”
Levie: Self-replicating AI escaping data centers is an unlikely near-term threat
“What I'm less worried about at the moment is probably the more fringe extreme things of the AI sort of self-replicates. It sort of, you know, jumps over outside the data center to another data center and then self-propagates. And I'm less nervous that that is,…”
Aaron Levie: Leaders should reserve micromanagement for core areas like AI
“I think that the real lesson is now is, is you choose the areas that you need to exert that level of involvement. So and in places like you know, let's say critical areas like AI or end user experience and some product, you know, strategy, you know, you, I sti…”
Levie: Zuckerberg is one of the best entrepreneurs of our time
“I think my read on this is you have you know, one of the best entrepreneurs of our time that has basically all of the right underlying resources for AI. So lots of compute, incredible engineers, you know, billions of users, i.e. Lots of data.”
Aaron Levie: Jensen Huang's 5-to-10-Year Tech View Is Unparalleled
“His window into, you know, five or 10 years out in, in tech is, is unparalleled right now.”
Levie: AI poses no threat to Apple and reinforces platform dominance
“I don't think Apple is, is immediately threatened, threatened by AI. It literally, most of my AI usage is on an Apple product that I have often paid money to Apple to be able to use. So, so like, if I, so like, just like on the most immediate basis, all of my …”