Why Enterprise AI Adoption Is Slower Than You Think — Aaron Levie (Box) + Harrison Chase

Aaron Levie, CEO of Box, sits down with Harrison Chase, co-founder and CEO of LangChain, to dig into why AI agents are thriving in software development but still struggling to break through in the rest of the enterprise. Aaron breaks down the structural reasons the knowledge work gap exists, how Box is building its agent harness, and why being headless might actually drive more usage than it kills.

Chapters:
0:00 Why Aaron tweets so much (and never uses AI to do it)
1:58 Box's core mission: bridging tech breakthroughs to the enterprise
3:28 Coding agents vs. knowledge work agents — why the gap exists
5:25 Every reason coding agents work, and why knowledge work inverts them
7:16 The dad-with-malware story and what it reveals about non-technical users
7:40 The permissions problem agents inherit from humans
10:44 Domain-specific agents and internal "AI forward-deployed engineers"
12:05 How Box is building its agent harness around enterprise content
15:04 MCP, headless vs. in-product, and the Salesforce moment
17:50 The Jevons paradox of headless software
19:11 When being disintermediated is actually good for your business
20:31 Deep Agents and why everything should be built on a coding harness
21:14 What if every knowledge worker had unlimited engineers?
22:44 What Box actually had to teach its agent about file systems
24:17 Single-model vs. multi-model harnesses: the token cost tradeoff
28:39 Token budgets, enterprise EPS pressure, and what comes next
30:25 Closing thoughts

Resources:
→ Deep Agents: https://www.langchain.com/deep-agents
→ LangGraph: https://www.langchain.com/langgraph
→ LangSmith: https://www.langchain.com/langsmith
→ LangChain Academy: https://academy.langchain.com