Box CEO Aaron Levie Points Out that Agents Will Become the Largest Users of Software, Headless APIs Are Inevitable
Box CEO Aaron Levie pointed out in a lengthy article that as AI Agents become the largest users of software, all software must provide a headless mode—Agents will primarily interact through APIs rather than UI.
Based on Box's practices and industry discussions, he proposed three major directions for future software business models:
1. User Seats will remain, but must include Agent API quotas.
Seats must come with a certain amount of API usage, allowing Agents to seamlessly operate data and tools on behalf of users (supporting ChatGPT, Codex, Claude, Gemini, Cursor, Copilot, etc.). Software that does not support this model will be directly eliminated.
2. Agents can have independent "Seats."
This applies to long-running, stateful Agents that require independent permissions (e.g., Openclaw-style). Pricing cannot simply replicate human Seats, as the usage scenarios differ greatly between companies using 1 or 1000 Agents, requiring flexible design.
3. Excess usage adopts a Consumption model.
Agent behavior exceeding Seat quotas will be billed based on actual usage or priced based on "results/tasks" rather than per API call. This will become the mainstream model when Agents work autonomously on a large scale.
Levie emphasized that the growth scenarios for Agents are nearly limitless, and Headless + hybrid pricing (Seat + Consumption) will be an inevitable trend in the software industry.
Source: Public Information
ABAB AI Insight
Aaron Levie has long promoted the transformation of enterprise content management towards cloud collaboration since the founding of Box. This lengthy article continues the path from traditional SaaS to Agent-ready infrastructure. Earlier, Box quickly integrated multiple AI Agent interfaces and opened APIs.
From a capital perspective, Box combines traditional Seat revenue with new Agent consumption revenue, locking in users through embedded API quotas while capturing the explosive usage brought by Agents. The motivation is to shift the revenue structure from fixed licenses to a more flexible hybrid model, fully benefiting from the deep calls of Agents on enterprise data.
Platforms like Salesforce, Notion, and Airtable are currently undergoing Agent integration and pricing adjustments. The software industry is in the early acceleration phase of transitioning from human user dominance to Agent dominance, focusing on Headless API design, permission control, and hybrid business model exploration.
Essentially, this represents a reconstruction of the industrial chain under technological substitution: AI Agents replace human UI operations through Headless APIs, with the mechanism of Seat + Consumption hybrid pricing shifting software from "selling to humans" to "simultaneously selling to humans and Agents," structurally transferring the pricing power and usage scale of enterprise software from fixed licenses to nearly unlimited Agent consumption, accelerating the entire SaaS ecosystem's evolution towards Agentic infrastructure.