Box's Aaron Levie: On Reinventing Yourself in the AI Age and Enterprise Diffusion

Watch on YouTube ↗  |  September 15, 2026 at 12:00  |  1:05:22  |  Sequoia Capital
Speakers
Aaron Levie — CEO of Box

Summary

Aaron Levie, co-founder and CEO of Box, discusses how AI is reshaping enterprise software and why the application layer should capture significant value by bridging models to real workflows. He explains Box's AI strategy, including its agentic harness for unstructured data and its model-agnostic routing approach. He also covers open-weight adoption, systems of record, vertical AI agents, and why enterprise AI diffusion will lag coding. He predicts 90% of enterprise tokens in five years will be for background work no human user initiated.

  • Aaron Levie argues the AI application layer is the key enterprise opportunity, bridging models to workflows.
  • Box is betting on AI with an agent harness tuned to its file system, permissions, and search.
  • Levie expects model labs' token subsidies to be temporary and favors model-agnostic routing.
  • Open-weight model adoption should rise as mature workloads are peeled off for cost or performance.
  • Systems of record must build superior agents and expose APIs/MCP to capture agentic value.
  • Vertical AI agents should win as work shifts to background dashboards, queues, and task lists.
  • Enterprise AI diffusion will take longer than expected due to data access, legacy systems, and change management.
  • Levie predicts 90% of enterprise tokens in five years will be user-uninitiated background work.
Ideas
Aaron Levie CEO of Box 2:04
Value accrues to enterprise AI application layer.
In the enterprise, value is not just in the frontier model; it is in the application layer that bridges raw model capability to actual workflows, data systems, permissions, change management, and domain-specific processes in banks, law firms, health care, and government. Labs face a conflict selling tokens and their subsidies are temporary, while application companies can route across models for cost and accuracy and build durable domain expertise, creating a large multi-layer opportunity.
AI-SECTOR HIGH
Aaron Levie CEO of Box 3:47
Box AI pivot leverages proprietary content layer.
Box is betting the company on AI and sits on hundreds of billions of files of unstructured enterprise data. It has built an agentic harness tightly tuned to its own file system, permission structures, and search, which it says beats giving raw Claude or ChatGPT APIs on accuracy and latency. Box is deploying agents for contract review, data extraction, long-running workflows, and Box Labs research, making its proprietary content layer a differentiated AI application play.
BOX HIGH
Aaron Levie CEO of Box 39:57
Systems of record benefit from agentic APIs.
Systems of record become more valuable, not less, in an agentic world if they do two things: build an agent that is provably much better than off-the-shelf agents at their own product, and go headless by exposing deterministic APIs or MCP so external agents can use their data and workflows. Their depth of purpose, data, security, and workflow ownership creates white-space use cases; Salesforce is an example where MCP access through Claude or ChatGPT has increased usage by an order of magnitude. Customers will reward this value creation.
Systems of record software CRM HIGH
Aaron Levie CEO of Box 45:12
Vertical AI wins as agents go background.
As agents shift from chat to background and asynchronous work, the interface becomes dashboards, queues, and task lists rather than a universal chatbot. Horizontal chat products will become heavy and less delightful, while vertical AI players that understand a specific process can manifest the right buttons, tabs, and workflow names. This favors vertical AI software across legal, security, coding, and other fields; he predicts 90% of enterprise tokens in five years will be for work no human user initiated.
Vertical AI software HIGH
Up Next
6:20
Look to market weakness opportunistically, not as the start of a big problem: Chris Verrone
CNBC · Sep 15
39:22
Even if frontier models are delayed, reasoning AI grows... Why GPU and memory demand must be viewed separately | Kim Jang-yeol, Head of Research Center at Unistori Asset Management [Focus Today's Stocks]
3PRO TV (삼프로TV) · Sep 15
4:44:11
[26.09.15 Afternoon Full Broadcast] KOSPI Held Back by the Triple Threat of Interest Rates, Oil Prices, and AI… Will Funds Avoiding Samsung and SK Hynix Flow to KOSDAQ?
3PRO TV (삼프로TV) · Sep 15
24:14
[Real Market Conditions] Passing the last 'steep hill'... Trading volume has dried up / The stock market will flow 'like this' going forward | Director Park Hyun-sang
815 Money Talk (815머니톡) · Sep 15
48:37
Trump Calls Nvidia’s Jensen Huang, Says AI Dangers Are a ‘Hoax’
Bloomberg Markets · Sep 15

This Sequoia Capital video, published September 15, 2026, features Aaron Levie discussing AI-SECTOR, BOX, Systems of record software, CRM, Vertical AI software. 4 trade ideas extracted by AI with direction and confidence scoring.

Speakers: Aaron Levie  · Tickers: AI-SECTOR, BOX, Systems of record software, CRM, Vertical AI software