AI-Native Workspace

AI-native workspace is Moxt’s product category in “AGI 来了？我用了一周，头皮发麻“｜对谈张昊然：Moxt 联合创始人. The idea is that a work environment should be designed around how agents read, write, remember, act, and coordinate, rather than treating AI as an add-on inside a workspace built mainly for human editing and navigation.

In Moxt’s version, the workspace looks somewhat like a document and file tree for humans, but its deeper design is agent-readable. Markdown, CSV, JSON, and HTML become practical replacements for Doc, Excel, and PPT because they are easy for models to inspect, modify, compose, and transform into Generated Work Interfaces. A file-system-like structure also gives agents a familiar substrate for Context Engineering and Agentic Workflow.

The source distinguishes an AI-native workspace from older collaboration tools with AI features. A traditional product can add agents, summaries, or automations, but Moxt’s claim is that if most code, documents, analyses, and plans are AI-produced, the underlying work substrate should change too. That puts Organizational Context, AI Coworkers, and dynamic output surfaces at the center of the product.

Key Claims

• The workspace itself is part of the agent harness because it controls what context exists, how it is structured, and what agents can modify.
• AI-readable artifacts can matter as much as model quality when agents need to work across documents, data, project plans, code, and meetings.
• Human interfaces may become review, feedback, judgment, and ritual surfaces while agents handle more drafting, synthesis, analysis, and generated visualization.
• The migration problem is real: the product must create enough new experience value to exceed the switching cost from tools such as Notion, Feishu, DingTalk, Jira, and Slack.
• An AI-native workspace increases the importance of values and permission design because agents live close to sensitive organizational context.

Connections

• Moxt and Zhang Haoran — product and founder source for the concept.
• Organizational Context — shared work substrate the workspace is meant to preserve.
• AI Coworkers and Momo — agent roles enabled by the workspace.
• Generated Work Interfaces — dynamic UI pattern made possible by agent-readable context.
• Context Engineering, Agentic Workflow, and Agent Harness — existing agent concepts the workspace operationalizes.
• Agent-Facing Interfaces, Headless Software, and On-Demand Apps — adjacent interface and software-product patterns.
• AI Organization Design and Human-Agent Collaboration — organization and human-role implications of moving work into an AI-native space.