Apr 20, 2026 · 54m · 20vc
Aaron Levie: Everyone is Wrong; We'll Have More Developers in 5 Years · 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, host Harry Stebbings interviews Box CEO Aaron Levie to demystify the realities of enterprise AI, highlighting why software engineering and professional headcounts will grow, how businesses will restructure their architectures to support AI agents, and how leadership must adapt to survive.
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 18.1% of the talking time here. How this is scored →
Harry as informed peer 4.5 Guest teaching 5.3 Guest disagreement 3.1 Harry pushing back 4.3
speaking balance: gold is Harry, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 26:25 Defending the OpEx budget shift against host mockery
Aaron forcefully pushes back when Harry dismisses his point about AI spend moving to OpEx budgets as a cliché VC blog post trope, insisting it represents a major structural shift in enterprise software.
Hardest push from Harry ▶ 16:58 Interjecting to challenge Box's value retentionHarry cuts in mid-explanation to directly challenge Box's business model, asking how the company can retain software value if it becomes a simple database crawled by third-party agents.
Biggest teaching moment ▶ 5:22 Exposing tech industry myopia on developer headcountAaron corrects popular narrative assumptions about AI reducing engineering jobs by showing that 85% of the non-tech economy will finally gain access to software engineering talent.
Harry holds his own ▶ 7:47 Challenging optimism with junior legal job displacement dataHarry brings concrete industry observations about the elimination of junior legal positions to counter Aaron's argument that total legal headcount will increase under AI.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Harry as informed peer | Guest teaching | Guest disagreement | Harry pushing back | Why |
|---|---|---|---|---|---|---|
| Analyzing Jensen Huang's AI Outlook | 2 | 5 | 3 | 2 | Harry opens by asserting that Jensen Huang performed poorly on another podcast, but Aaron reframes the debate around commercial realities and software upgrade timelines. Aaron challenges the premise that short AI leads create permanent existential advantages. | |
| The Future of Software Engineering Headcount | 4 | 6 | 4 | 5 | Aaron argues that non-tech industries will hire massive numbers of software engineers, calling the tech industry's views myopic. Harry pushes back by citing the eradication of junior legal roles, forcing Aaron to clarify how automation shifts operational bottlenecks. | |
| The Rise of the Agent Operator | 3 | 6 | 2 | 1 | Aaron defines the emerging role of the agent operator needed to re-architect business workflows in regulated enterprises. Harry maintains an encouraging, conversational tone with lighthearted banter. | |
| SaaS Value, APIs, and Unstructured Data | 6 | 5 | 3 | 7 | Harry challenges whether SaaS platforms become commodity databases when agents interact via APIs, interjecting to ask how Box retains value. Aaron defends Box's moat by pointing to its high API volume and proprietary business logic layers. | |
| Cybersecurity Risks and Agentic Solutions | 5 | 5 | 2 | 2 | Harry quotes Jason Lemkin on AI triggering a golden age for cybersecurity. Aaron agrees and details the dual threat of AI-generated code volume and automated offensive vulnerability scanning. | |
| Compute Spend and Enterprise Token Budgets | 4 | 5 | 3 | 6 | Harry presses Aaron to give public financial projections regarding Box's long-term compute spend relative to employee salary. Aaron avoids SEC compliance issues before explaining enterprise token allocation models. | |
| OpEx Budgets and Non-IT Funding | 6 | 7 | 4 | 6 | Harry mockingly dismisses the idea of token budgets moving to OpEx as a generic VC blog post trope. Aaron forcefully defends the thesis and explains why real-world enterprise adoption takes years due to compliance and legacy data fragmentation. | |
| Liability, Ownership, and Open Models | 5 | 4 | 2 | 3 | Harry notes that enterprise software reliance stems from needing organizational accountability when things go wrong. Aaron agrees, emphasizing that AI speed does not bypass legal regimes or corporate liability. | |
| Why Public Companies Struggle with Agents | 5 | 5 | 4 | 5 | Harry presents Jason Lemkin's claim that Palantir is the only public company building good AI agents. Aaron rejects the assertion, citing Box's native agentic capabilities and describing the intense, fast-moving technical demands on current CEOs. | |
| Monetizing Agents and Public Market Valuation | 5 | 5 | 3 | 5 | Harry questions whether agent features actually drive new pricing power and revenue re-acceleration for public SaaS companies. Aaron provides Box's specific monetization data while offering a balanced view on depressed software market multiples. | |
| CEO AI Readiness and Technical Leadership | 5 | 5 | 4 | 5 | Harry quotes Jason Lemkin implying Aaron is struggling, leading Aaron to playfully object and defend Box's performance. The segment transitions into rapid-fire questions covering Atlassian, frontier models, and personal advice. |
Statements from this episode (24)
Levie: AI shifts where humans enter workflows rather than eliminating them
“We haven't removed humans from the loop. We've just changed where they enter the loop.”
Levie: There will be more lawyers in 5 years than today
“There are going to be more lawyers in the next five years than we have today.”
Levie: US-China AI race is commercial, not existential
“You know, the idea that we're in some kind of, you know, kind of race, existential race where a month or two of advantage is gonna, you know, change the total outcome of AI progress and what everybody does between us and China. I just don't agree with, I think…”
Levie: CS grads will join industrial giants over big tech
“Maybe if you're graduating, you know, name your computer science school today, you don't go immediately to Google. You go to literally John Deere or Caterpillar or Eli Lilly, but the skills that you have are going to be just as relevant in just a different dom…”
Levie: Automation exposes unseen downstream bottlenecks rather than reducing work
“Automation is going to actually just force us to see the next set of bottlenecks that are in all of these industries that we didn't perceive that we had before because everything was so slow and manual.”
Levie: AI agent operators will create up to 1M new jobs
“I'm not a hundred percent clear on that, but there is a hundred percent a role right now that there's going to be 500,000, a million jobs created for, and it's basically some kind of agent operator.”
Levie: Future software development will center on background systems, not apps
“There's a limit to, again, the number of software you need to build that looks like an app on your phone. There's an unlimited amount of software you need to build That looks like a background system process that's connecting different data sources, automating…”
Levie: AI agents will trigger an explosion of unstructured data
“Agents are going to be this incredible consumer and creator of your unstructured data. They're gonna Read through every one of your contracts and generate all of your contracts. They're going to generate marketing assets. They're going to write reports for you…”
Levie: Box API activity already dwarfs direct human user interactions
“The volume of API usage on our system is already enormous and already, you know, is outsized relative to any of the end user interactions in the system.”
Levie: A lot of money will be made in agentic security
“I think, you know, for better or worse, agents are the solution to the problem that agents have caused. And that's why there's going to be a lot of money made in agentic security as well.”
Levie: Companies will spend vastly more on compute per employee
“Directionally, the idea is actually pretty salient, which is you're going to be spending more on compute per person in the future than you ever thought, and that you certainly are today.”
Levie: Enterprise AI token budgets will shift from IT to departmental OpEx
“The budget of tokens will have to move out of IT spend and into regular kind of OPEC spend. This can't be treated like a, oh, you know, I'm going to trade off between Salesforce licenses or compute tokens. Like it's going to more be, I'm going to trade off You…”
Levie: AI agents tapping OpEx budgets will double global tech spend
“And I don't think it like, You know, 10 X's the size of IT spend or technology spend globally, but it certainly doubles it.”
Levie: Enterprise AI diffusion will take longer than Silicon Valley expects
“And then the only part why, why 18 months is like not even a relevant window to me is I think diffusion is going to take longer than Silicon Valley thinks.”
Levie: Enterprise AI preparation represents 10 years of work for Accenture
“That is like what I just described right now is 10 years of work for Accenture in every enterprise on the planet or the next gen Accenture that does this in particular industries or workflows.”
Levie: Enterprises cannot shift AI liability to providers like Anthropic
“I promise you, you're not going to be able to blame Anthropic when something goes wrong. And so if you can't blame Anthropic when something goes wrong, then at some point it doesn't really work to tell the company, like to tell your customer, well, That, that …”
Levie: Traditional tech media offers zero AI alpha compared to frontline engineers
“The information sources aren't the classic ones. It's not the roll up review two weeks later from your traditional news publication that is going to give you any kind of alpha. It's the practitioner who's the, you know, literally the engineer at the, you know,…”
Levie: Markets will differentiate software companies on AI agent impact
“I would say that we're in a period right now where basically the market is being treated roughly as in, as the kind of indiscriminately, you know, kind of bucketed sector. And the next year, two years or whatnot, you'll start to see some separation and parsing…”
Levie: Public company CEOs spend weekends coding with Claude Code and Codex
“I have Slack channels and WhatsApp groups where people on the weekend are just like working with Cloud Code or Codex building stuff, and they're public company CEOs.”
Levie: Software platforms must adopt a headless-first architecture for AI agents
“And in the past year, those capabilities have just absolutely accelerated. To the point where I'm fully convinced that you just, you have to be, you know, headless first as a software platform.”
Levie: Enterprise AI adoption will be multi-vendor rather than single-source
“Companies are going to adopt multiple of these systems. They don't want to be single vendor. They don't want to be, they don't want to have single vendor in this stack. One service goes down, or one changes its APIs, or one has a new commercial model. You're g…”
Levie: AI coding productivity gains will not quickly transfer to other knowledge work
“What they think is that, ah, that the outcomes that you're seeing in AI coding will quickly come for other areas of knowledge work, and that is a, ah, a slight misread on the other areas of knowledge work.”
Levie: Every enterprise deploying AI agents will require automated evaluation tooling
“And then I'm like, oh, actually everybody on the entire planet, if you're putting agents into an enterprise workflow needs evals. Because you need to know if all of a sudden your agent just stopped producing, you know, like you know, loan origination documents…”
Shorts cut from this episode
▶ "I just want to shake the industry" | Box CEO on 20VC · 20VC (@0:00)
▶ Why Dwarkash was wrong and Jensen was right on upgrading sys (@3:19)
▶ Is your software agent-ready? · 20VC with Harry Stebbings (@15:22)
▶ The future of job roles · 20VC with Harry Stebbings (@9:28)
▶ We will have MORE lawyers in 5 years not LESS · 20VC with Ha (@0:40)