Keep pulling the thread on Aaron Levie.
Box is betting its business model on the premise that AI agents will increase the creation and use of files, creating demand for a platform optimized for agent-based file interaction.
It is not feasible to replicate complex enterprise systems like SAP through simple, prompt-based "vibe coding."
The paradigm of AI agents using code to interact with tools is being advanced by products like Claude Code and unannounced projects from OpenAI and Perplexity.
A growth marketer at Anthropic used Claude Code to automate the work of what would have previously required 5 to 10 employees in siloed roles.
Box is developing an AI agent that can dynamically choose between using a pre-existing skill, an existing Box tool via API, or writing new code on the fly to solve a user's problem.
Box has officially launched a command-line interface (CLI) for its platform.
By providing the Box CLI as a tool to a large language model like Anthropic's Claude, users can interact with their entire Box system using natural language.
The widespread adoption of AI capabilities within large enterprises will take longer than many in Silicon Valley anticipate due to complex security, integration, and control challenges.
The domain knowledge of complex enterprise software like SAP is not just in the data layer but is deeply embedded in the UI, middle-tier logic, and established user workflows, making it difficult to replace via API access alone.
The shift to agent-driven, API-first business models could lead to decreased revenue for some SaaS companies like Workday, while creating significant revenue growth for others.
In the future, AI agents, rather than analyst firms like Gartner, will increasingly be the source of recommendations for which enterprise software a company should use.
Aaron Levie predicts that companies like Workday will develop methods to influence or "buy" AI agent software recommendations, corrupting a purely meritocratic selection process.