For years, enterprise generative AI has operated on a strict pay-as-you-go conversational model. You open a chat window, feed it context, wait for a response, and start over when priorities shift. OpenAI’s DevDay 2026 announcements shatter that transactional paradigm with two interconnected releases: Dots, a new class of persistent AI agents powered by the GPT-6 Astra model, and ChatGPT Space, a shared collaborative layer designed to bring those agents directly into team workflows.
Together, these features signal a profound shift. OpenAI is no longer just building chatbots that answer questions; it is constructing an operating environment where AI agents act as persistent coworkers capable of working across software stacks while their human counterparts are offline.
Beyond the Chat Window: What Dots Actually Do
The core limitation of early generative AI tools was their ephemerality. They vanished the moment you closed the browser tab. Dots change that equation by giving persistent agents their own cloud computer and browser access. Announced by OpenAI CEO Sam Altman, these agents run on GPT-6 Astra and are engineered to keep monitoring projects, executing multi-step tasks, and bringing completed work back for approval long after an employee logs off.
Unlike standard chatbots confined to a single dialogue thread, a Dot can handle several background projects simultaneously. In engineering environments, a Dot can watch incoming customer feedback, identify recurring bugs, build and test fixes, and return completed pull requests complete with demonstration videos. In marketing and product teams, agents can ingest audience positioning documents, track product updates, and automatically revise campaign materials. Sales teams can deploy Dots to compare enterprise prospect requirements against account history, draft technical proofs of concept, and keep proposals current as negotiations evolve.
Beyond web and desktop chat interfaces, OpenAI has confirmed that Dots will soon extend across messaging platforms like Slack and Microsoft Teams, and even function over the phone as advanced audio models. Connected through OpenAI’s extensive plugin ecosystem to more than 4,000 applications, Dots can gradually learn an individual’s preferences, operational standards, and working habits over time.
Crucially, this execution happens asynchronously. OpenAI terms one key capability "proactive research," allowing agents to scan connected applications for items requiring attention while users focus elsewhere.
ChatGPT Space and the Death of Private Silos
Individual assistants are helpful, but enterprises live and die by collective workflows. To bridge that gap, OpenAI is introducing ChatGPT Space, a collaborative environment replacing the traditional Library for Pro, Business, and Enterprise tiers.
Space acts as a home for Pages, spreadsheets, presentations, and shared team materials. Instead of keeping AI work locked inside private chats, employees and teams can bring human colleagues, ChatGPT, Codex, and Dots into shared documents. Colleagues can tag a Dot or ChatGPT for assistance, build upon work completed by other agents, and maintain continuous context across projects.
Pages within Space combine text, images, charts, and interactive tools in documents that employees and AI can edit together in real time. Connected tools such as Google Drive and Slack supply additional context, while collaborative slides and spreadsheets round out the suite. A project page can pull action items from various channels while an assigned Dot keeps deadlines and task owners updated as plans shift. Because Dots communicate natively across ChatGPT, Slack, and Microsoft Teams, they follow employees across communication channels without breaking the shared context maintained in Space.
For enterprises, ChatGPT Space turns Dots from individual assistants into participants in shared team workflows. Teams can bring goals, plans, designs, and other materials into one common environment, then ask ChatGPT or a Dot to maintain dashboards, translate a teammate’s design into engineering tasks, flag blockers, or keep project documents synchronized with new information.
Governance, Permissions, and the Guardrails of Autonomy
Allowing an AI agent to operate autonomously when nobody is babysitting introduces significant enterprise risk. An error made by an agent operating authenticated business software can potentially alter records, send sensitive information, or trigger unintended downstream workflows. OpenAI has built guardrails into the foundation of Dots, though questions remain about how they will perform in complex corporate environments.
During proactive research, background tools are strictly read-only: they cannot send messages, modify content, or control user hardware. For active tasks, organizations configure access permissions through ChatGPT's existing app controls, supplemented by Custom Rules that permit, restrict, or require approval for specific actions. An Activity View lets users audit background work and intervene as needed. Furthermore, an automated review system checks consequential actions against user instructions and safety requirements before granting autonomy.
ChatGPT Space adds another governance dimension because the outputs of those agents can become shared organizational artifacts. Workspace administrators’ existing sharing rules continue to apply, and teams can grant view or edit access to Pages and later change or revoke it. Files uploaded directly into a Page inherit that Page’s permissions, while merely linking to an external file does not grant collaborators access to the original—although any material copied or summarized onto the shared Page becomes visible to everyone who can access it.
Navigating the Private-to-Shared Boundary
A critical nuance in team collaboration environments is the boundary between private context and shared deliverables. OpenAI draws a strict line between a user’s private ChatGPT context and shared Space content. Sharing a Page does not expose private conversations or personal memory to coworkers.
However, if ChatGPT uses something from a person’s memory and writes it onto the shared Page, that information becomes visible to collaborators. This makes the shared Page an important control point for employees and administrators to review what information crosses from private AI context into shared organizational work.
Regarding data privacy and model training, OpenAI reiterates that for Business, Enterprise, and Edu customers, workspace content is not used to improve its models by default. Personal-plan users can choose whether Dot conversations and work contribute to model improvement. The company explicitly states it does not train directly on a dot’s proactive research or its private notes to itself.
Pricing, Packaging, and the Unanswered Economic Questions
The commercial structure surrounding Dots and ChatGPT Space introduces both immediate opportunity and strategic uncertainty. The first Dot is included at no additional charge for Pro and Business Premium customers. However, OpenAI has not yet disclosed the numerical allowance for intensive background work, the eventual cost of additional Dots, pricing for higher speed or workload capacity, or commercial tiering for specialist agents. Enterprise, Edu, and Healthcare customers gain access via a beta program when an administrator enables it.
Those missing numbers matter immensely. If persistent agents genuinely take over substantial chunks of administrative, engineering, sales, research, and creative work, organizations will need clear visibility into unit economics.
ChatGPT as an Enterprise Operating Environment
Stepping back from individual features, the broader architecture emerging from DevDay resembles an operating environment. Operating systems historically became platforms where applications run, identities are managed, and files are organized. By combining plugins, event-driven automations, Codex cloud environments, and shared spaces like ChatGPT Space, OpenAI is positioning ChatGPT as a coordinating layer for enterprise software.
For IT leaders, this consolidation brings efficiency alongside familiar architectural trade-offs. Relying more deeply on a single platform for agentic workflows increases dependency on OpenAI's infrastructure, raising valid questions around vendor lock-in, pricing stability, and data governance.
The transition from individual chat assistants to collaborative team agents is underway. Whether enterprises embrace this shift willingly or find themselves retrofitting governance after the fact, Dots and ChatGPT Space have firmly established that the AI coworker is no longer a hypothetical concept—it has a seat at the team table.
