← All episodes

The Era of AI Agents | Aaron Levie on The a16z Show

| 13 products mentioned
a16z a16z host
Aaron Levie guest
Watch on YouTube ai agents enterprise software architecture system integration data security legacy systems modernization ai economics compute infrastructure

Aaron Levie discusses how AI agents will fundamentally reshape software architecture and enterprise operations, arguing that the diffusion of AI capability will take longer than Silicon Valley expects due to legacy system constraints and security complexities. The episode explores the tension between enabling agents with broad system access and protecting enterprise data, while suggesting that agents will eventually force companies to rebuild infrastructure around their needs rather than human interfaces. [Claude]

Key takeaways
  • AI agents will require software to be built with agent interfaces as the primary design constraint, not a secondary consideration—meaning APIs, CLIs, and system-wide accessibility matter more than polished user interfaces.
  • The biggest near-term problem is that enterprises will struggle with information security and access control when agents operate autonomously across multiple systems, as agents can be socially engineered far more easily than humans and will inevitably leak sensitive data if given access to it.
  • Legacy enterprise systems like SAP and Workday won't disappear quickly—instead, agents will repeatedly hit walls trying to work with these systems, eventually forcing companies to modernize their infrastructure stack or lose competitive advantage.
  • Most financial models underestimate AI's opportunity by at least an order of magnitude because they assume fixed revenue pies rather than recognizing that new use cases and business models emerge with every major technology shift (similar to how cloud computing created 1000x more demand than server sales).
  • Compute budget management will become as critical as R&D spending decisions within the next 2-3 years, creating tension between startups that burn through capital experimenting and enterprises that freeze due to uncertainty—with winners emerging in the middle ground.
  • The monetization shift from per-seat SaaS to usage-based token pricing will enable new micro-transaction business models where agents can economically access previously behind-paywall data and services that humans wouldn't bother purchasing.

Recommendations (3)

SAP
SAP uses

"I use SAP all day. I work in finance. I have to go and generate all these reports"

a16z · ▶ 11:26

NanoCloud uses

"I spent all weekend implementing my nano cloud bot"

a16z · ▶ 12:55

Claude
Claude uses

"you give Claude the box CLI and you can now interact with your entire box system via natural language"

Aaron Levie · ▶ 16:03

Mentioned (10)

Claude Computer Use
Claude Computer Use "That's the Claude Computer Use phenomenon. That's whatever OpenAI is kind of cooking up with the ..." ▶ 1:55
PowerPoint
PowerPoint "I spent my life sitting next to people on airplanes saying how can I make PowerPoint do X" ▶ 12:07
Word
Word "it hurt physically hurt to watch somebody suffering with bullets and numbering in Word" ▶ 12:13
Excel
Excel "let me just make a two-sided a two axis graph in Excel which like is rocket science like almost n..." ▶ 12:18
Slack
Slack "with regular humans you don't get to look at the Slack channel of the person that that is working..." ▶ 19:39
Workday
Workday "it's been a long-standing tension point with the likes of Workday and SAP and stuff like how much..." ▶ 28:44
Salesforce
Salesforce "Salesforce went through three different massive platform redesigns" ▶ 28:48
Gartner
Gartner "it used to be that you would go to like Gartner to be like, tell me what to do. Tell me what syst..." ▶ 35:04
AWS
AWS "we went through this with with AWS like people learned to do the credit" ▶ 47:09
Stripe
Stripe "We use Stripe for payments"