a16z Podcast: “Software Finally Eats Services” — Aaron Levie

Episode Date: September 24, 2025
Host: Andreessen Horowitz
Guests: Aaron Levie (CEO, Box), Steven Sinofsky (a16z), Martin Casado (a16z)
Theme: Exploring how AI is fundamentally changing software and services, labor markets, productivity, and the tech startup landscape, including a deep dive into H1B reform, AI-native startups, and how software is now “eating” professional services.

Episode Overview

In this wide-ranging conversation, host Andreessen Horowitz is joined by Aaron Levie (CEO of Box), Steven Sinofsky, and Martin Casado to dissect the seismic changes happening as AI and software transform both the technology industry and professional services. Central topics include US immigration reform (H1B visas), the real productivity impact of AI coding agents, the unprecedented velocity of new startups, and how both incumbents and new companies are navigating the current “platform shift” ushered in by artificial intelligence.

Key Discussion Points & Insights

1. The H1B Debate: Who Wins With Reform?

[01:28 – 11:54]

• The discussion opens with reactions to recent immigration policy proposals about pricing H1B visas, including whether it would favor tech giants or startups.
• Martin Casado argues that setting a price is a practical way to allocate a scarce resource:

  “It’s very hard for startups to hire because of the lottery system… a reasonable way to do it is to set price, because you’ve got a market and you need to allocate supply.” (01:39, Casado)

• Aaron Levie pushes back, warning that pricing could privilege dominant players:

  “You’d have a situation where the Amazons and Googles would probably actually capture the vast portion of the talent in this situation.” (02:33, Levie)

• They discuss the lottery system’s inefficiencies, the “body shop” phenomenon, and the potential for minimum salary bands to solve wage dilution.
• Steven Sinofsky laments the wasted effort:

  “The incredible amount of resources… big companies… have enormous teams that spend all of their energy literally as lobbyists working this system.” (06:41, Sinofsky)

• There’s consensus on the need for policy that promotes high-merit immigration without squeezing out startups or certain job bands.

2. AI and Labor Productivity: Is the Hype Real?

[12:23 – 27:44]

• Levie shares data and anecdotes from Box:

  “About 30% of our code right now is coming from AI.” (13:39, Levie)

• Reports of productivity gains vary widely:

  “Some people say I’m getting a 20-30% productivity gain, other people will say 75%... I haven’t been able to quite figure out a pattern.” (13:12, Levie)

• The panel observes superhuman productivity, especially in senior, small teams using AI:

  “The more senior small teams that use AI are superhuman. It’s like they woke up and they were all Tony Stark. It is unbelievable.” (19:58, Casado)

• Major shift from writing to reviewing code:

  “They’re [small teams] fundamentally engineering in a different way... you’re really in the business of doing code review, not code writing.” (15:22, Levie)

• Early adopters are highly forgiving, reminiscent of the early Internet era:

  “The early Internet people didn’t complain that the Internet was slow.” (17:07, Sinofsky)

• However, measuring true productivity is difficult due to dazzling new tools and “shadow productivity”:

  “People are so enthusiastic about using AI, but it really hasn’t impacted their output… there’s almost shadow productivity.” (19:17, Casado)

• Bottom-up, personal adoption of AI tools is succeeding where top-down, centralized projects often fail.

3. The Changing Nature of Work and Expertise

[27:44 – 36:28]

• The panel compares the current shift to the post-spreadsheet era—jobs, workflows, and decision quality are fundamentally changing:

  “Pre-spreadsheet, post-spreadsheet is a perfect example… your quality of decision was really bad [before].” (28:52, Sinofsky)

• AI enhances, but does not replace, domain expertise:

  “The biggest gains of AI go to people who have some degree of expertise in an area... experts just get more powerful in this world.” (31:28, Levie)

• The rise of “prosumers” and new TAM (total addressable market):

  “AI has created a third category... product utility gain to me that is worth $20 a month.” (34:43, Levie)

• Tools lower the bar for enthusiasts to enter a field (AI as bootstrapping for new careers), but professionals and experts continue to drive value.

4. Startup Dynamics and the AI Platform Shift

[37:11 – 43:21]

• AI resets the playing field: younger founders are again able to build huge companies, reminiscent of earlier tech booms.
• Previous lulls were due to “box checking”—mature SaaS and consumer platforms had solved most old problems. AI opens all-new territory:

  “You have a complete reset of the landscape… none of the disadvantages of a big company.” (39:23, Levie)

• The only real incumbent advantage now is distribution, but scale is rapidly democratized by AI tools and viral adoption.

5. Incumbents vs. New Challengers: Who Survives the Shift?

[40:58 – 57:48]

• The panel reflects on platform shifts (Internet, mobile, cloud) and the often-overestimated advantage of incumbents:

  “The advantages to incumbents are wildly overestimated… historically, even if you do make the transition, you really didn’t.” (42:05, Sinofsky)

• Incumbents rarely self-disrupt; “skunk works” or startup-style teams drive real breakthroughs.
• Modern disruption is less about overtaking incumbents than about massively expanding markets:

  “Microsoft can be a $4 trillion company and you can have all these new categories emerge… it all works together as one sort of ecosystem.” (42:40, Levie)

• Casado notes incumbents struggle most when user behaviors shift:

  “Incumbents are very bad when new user behaviors and buying behavior show up… AI is definitely a new user behavior and a new buying behavior.” (43:21, Casado)

6. Software Eating Services: The New Frontier

[47:58 – 51:27]

• For the first time, AI allows software to subsume entire professional services sectors—not just replace other software:

  “This opening up of non-software TAM for software… you’re packaging up intelligence for a particular domain and workflow.” (48:27, Levie)

• The “incumbents” are now services businesses; often, these pros become the key users of the very AI tools that might disrupt them.

7. Pervasive AI Adoption & Consumerization Cycle

[51:50 – 54:55]

• AI user adoption is rapidly advancing, blending from consumers to prosumers to enterprise workers.
• Consumer use cases, including Levie’s teacher sister, are now dominant:

  “My sister, not in tech at all… she was like, ‘yeah, I was asking chat this question’… ChatGPT—that’s what normal people call this.” (53:09, Levie)

• Consumer adoption precedes and pressures enterprise to upgrade:

  “We now have the conditions laid for the next phase… consumer adoption now goes first and then it gets basically pulled into the enterprise.” (53:53, Levie)

• Ubiquity of smartphones means distribution is no longer a bottleneck.

8. What Comes Next: Incumbents, Disruptors, and Leadership

[54:55 – End]

• Future: a mix of big incumbents and new disruptors—expect a “boring” equilibrium like SaaS/cloud, but also totally new categories.
• The most important competitive edge will be thought leadership and the ability to set the agenda:

  “Incumbents become bigger but nobody wakes up in the morning wondering what they’re up to.” (57:24, Sinofsky)

• Legacy firms have rare comebacks, but AI could let laggards regain ground, e.g., Oracle’s recent resurgence.

Notable Quotes & Memorable Moments

• “About 30% of our code right now is coming from AI.” — Aaron Levie ([13:39])
• “The more senior small teams that use AI are superhuman. It’s like they woke up and they were all Tony Stark.” — Martin Casado ([19:58])
• “This AI wave is so personal… most people in the company are using ChatGPT… much harder to measure.” — Casado ([24:22])
• “Experts just get more powerful in this world… Be really good at a particular field, and then AI is merely a turbocharger of your capability.” — Levie ([31:28])
• “AI has created a third category… product utility gain to me that is worth $20 a month.” — Levie ([34:43])
• “There was a period… we were actually in kind of a bit of a lull as an industry… Now we have that era in AI. And that is why I’m so unbelievably pumped up.” — Levie ([37:47])
• “The only real incumbent advantage now is distribution… but now everybody has all of the ingredients.” — Sinofsky ([54:00])
• “Incumbents are very bad when new user behaviors and buying behavior show up… AI is definitely a new user behavior and a new buying behavior.” — Casado ([43:21])
• “Microsoft can be a $4 trillion company and you can have all these new categories emerge…” — Levie ([42:40])
• “This opening up of non-software TAM for software… you’re packaging up intelligence for a particular domain and workflow.” — Levie ([48:27])

Timestamps for Key Segments

• [01:28 – 11:54] H1B visa policy, market pricing, and implications for startups vs. incumbents
• [12:23 – 27:44] AI productivity, code generation at Box, measurement challenges, and new workflows
• [27:44 – 36:28] Analogies to past tool adoption, AI complementing expertise, rise of prosumers
• [37:11 – 43:21] Startup opportunities amid AI platform shift, velocity, advantages and disadvantages
• [43:21 – 49:00] Incumbent challenges, market expansion, and real world business building with AI
• [51:50 – 54:55] AI adoption curve, consumerization, and velocity of enterprise change
• [54:55 – End] Long-term competitive outlook and strategies for incumbents and disruptors alike

Tone and Style

The discussion is lively, honest, and at times irreverent—layered with war stories, analogies to tech history, and an optimistic yet realistic view of the AI revolution. The group balances hard data (like Box’s AI code stats) with anecdotal, on-the-ground observations.

Bottom Line

• AI’s platform shift is turbocharging both individual and startup productivity, pushing software well beyond tech into professional services and the broader economy.
• Incumbents remain strong, but their advantages are narrowed—distribution is less of a moat, and user behavior is swiftly evolving.
• The winners of this cycle—whether in software, services, or new AI-native verticals—will be those who can best harness expertise, adapt new tools, and respond to bottom-up, user-driven adoption.

“It will just fundamentally change people’s daily patterns. We now have the conditions laid for the next phase.” — Aaron Levie ([53:59])

For those who haven’t heard the episode, this summary captures the technical nuance, strategic debate, and forward-thinking attitude that defined one of a16z’s liveliest recent conversations.