OpenAI chose the last day of September to transform ChatGPT from a chat interface into a shared office. On September 29, the company announced ChatGPT Space — a Notion-style collaborative workspace where teammates, ChatGPT, and autonomous agent personas called “Dots” can work on pages, files, and other assets in one place. It is rolling out to Pro, Business, and Enterprise subscribers starting October 1, with web and desktop first and mobile to follow. Replaces Library for those tiers.
The angle is not the workspace itself — shared document editors are old news. The angle is that this is where autonomous agents will actually work alongside humans. Dots launched yesterday with cloud computers, 4,000+ app integrations, and a $500-per-month Pro 500 tier. Space is the surface where those agents operate — shared pages, team chats, and collaborative contexts. OpenAI is not just selling agents; it is building the workplace where agent-human collaboration happens.
What ChatGPT Space Actually Does
Space functions as a companion word processor designed for human and agent collaboration. Pages — the primary workspace unit — support real-time co-editing between humans and Dots agents. Team chat lets agents be tagged into conversations as active participants, not passive tools. File integration allows agents to read, annotate, and act on shared documents.
Starting October 1, Word access is enabled by default in ChatGPT Business workspaces. Slides functionality has been announced but is not yet live — OpenAI says it will roll out in the coming weeks. The combined package turns ChatGPT into a direct competitor to Microsoft Copilot workspaces, Google Workspace AI, and Slack AI channels.
The competitive positioning is deliberate. Recon Analytics estimates that roughly 70% of enterprise workers prefer ChatGPT when free to choose their AI tool — and OpenAI now counts approximately 15 million paid seats with roughly 160% year-over-year growth. Even marginal conversion of that base to Space users would create a distribution advantage that no workspace competitor currently holds.
The Dots Integration Is the Real Moat
The workspace is the container; the agents are the product. Each Dot runs on GPT-6 Astra, has its own cloud computer, and can operate across 4,000+ connected applications. In Space, Dots can be tagged into team chats, assigned tasks on shared pages, and granted access to workspace files — all without counting toward ChatGPT usage limits.
This creates a structural difference from Microsoft’s approach. Copilot operates as an assistant attached to existing Office documents. Google’s Gemini Workspace integration operates similarly — the AI augments the human’s work within their existing tools. OpenAI’s model is different: the agent is a co-worker, not an assistant. It has its own workspace identity, its own compute resources, and its own persistent context.
The distinction matters for enterprise adoption. A KPMG Q3 2026 survey found that 62% of large enterprise executives are building or deploying agents. But a Cisco survey from March 2026 found only 5% have reached production, with 60% citing security as the primary barrier. Space attempts to solve that gap by providing a governed workspace where agents operate within a shared, auditable context rather than independently across disparate systems.
The Platform Play Completes
With Space, OpenAI’s platform strategy now has three legs. The first is commercial monetization through the $500-per-month Pro 500 tier and autonomous agents. The second is government lock-in through the OneGov agreement — $0 licensing for all federal, state, local, and tribal organizations, effective today. The third is the shared workplace surface where both commercial and government users will actually deploy agents.
The OpenClaw Enterprise open-source control plane and Meta’s announced-but-not-launched enterprise platform represent alternative approaches — OpenClaw from the governance layer, Meta from the consumer surface. But neither offers a shared workspace with integrated autonomous agents at scale. OpenAI is combining the workspace, the agents, and the distribution in a single product.
The harness pattern that the infrastructure layer has been converging on — identity, events, sandboxes, governance — is now embedded in the consumer-facing product. Space is the user-facing layer of the same architecture that OpenClaw implements at the infrastructure level.
The Open Questions
Three uncertainties will determine whether Space becomes a genuine workspace platform or a feature extension. First, enterprise security architecture: OpenAI has not disclosed audit trails, agent identity management, or data residency specifics for Space — the same gaps that enterprise buyers have flagged across every agent platform. Second, pricing after the Dots promotional period: the first month is free, but OpenAI has not said what multi-agent workspaces will cost at scale. Third, the governance layer: Anthropic faces a supply chain risk classification while OpenAI launches autonomous agents in a shared workspace — the regulatory asymmetry is unresolved.
The agentic AI workspace category exists now. Whether it becomes the dominant model for enterprise computing depends on whether OpenAI can deliver governance that matches the ambition.
Note: Recon Analytics’ 70% preference figure is directional, sourced via secondary reporting. OpenAI’s 1.2 billion weekly user base and 15 million paid seats are self-reported and have not been independently audited. KPMG survey figures cover building, developing, or deploying — not just active production. Cisco survey is vendor-commissioned.