TITLE: OpenAI Dots: Always-On AI Agents for Business Teams DATE: 2026-09-30 COMPANY: OpenAI TOPIC: Agent Systems SUMMARY: OpenAI launched Dots at DevDay 2026 on 29 September, introducing always-on AI agents that run 24/7 on dedicated cloud computers and connect to over 4,000 business applications. Alongside Dots, OpenAI unveiled ChatGPT Space, a persistent team workspace where humans and agents collaborate on documents and projects in a single environment. The announcements mark a structural shift from AI as a query tool to AI as a continuous, task-completing team member. WHAT CHANGED: OpenAI launched Dots at its DevDay 2026 conference in San Francisco on 29 September 2026. Each dot is an always-on AI agent running on the GPT-6 Astra model, assigned its own cloud computer and browser. Dots work toward the goals you give them continuously, including when you are not at your desk, pausing only when they need human input or reach an action defined as requiring approval. You can interact with a dot through ChatGPT, on a voice call, or through Slack and Teams integrations, with SMS support coming in the following weeks. ChatGPT Space was announced alongside Dots. It replaces ChatGPT's existing Library for Pro, Business, and Enterprise users and creates a persistent shared environment where teams, ChatGPT, and dots work together on the same files and documents. Space is positioned to sit alongside tools like Notion, Slack, and Google Drive, with AI agents embedded as collaborators rather than as an external layer. A new document type called Pages supports human and agent co-authoring, including writing, research, charts, and image generation. Collaborative slides with PowerPoint and Google Slides export are scheduled for release within the following few weeks. OpenAI also confirmed that its enterprise business more than doubled in the third quarter of 2026, with overall revenue growing approximately 70 percent quarter to date to reach an annualised rate of nearly $70 billion. On the data privacy side, OpenAI confirmed the general availability of Zero Data Retention with Private Safety Processing, which allows enterprises to use frontier models with no customer prompts or responses retained after processing. Enterprise content is not used to train models unless customers explicitly opt in. Private Inference, a confidential-computing architecture designed to make AI inference verifiable and privacy-preserving, was previewed as a coming release for later in the year. WHY IT MATTERS: Dots can complete long-horizon tasks, such as researching a market, drafting a proposal, and formatting it for presentation, as a single continuous job rather than a series of prompted exchanges The 4,000-plus app integrations make it practical to connect dots to existing business software without custom development ChatGPT Space reduces the context-loss problem: agents inherit shared files and conversation history, so they do not need to be re-briefed for each session Zero Data Retention removes the primary governance objection for businesses that have been reluctant to use ChatGPT with client or financial data Built-in guardrails specify which actions a dot can take independently and which require human confirmation, addressing the control concerns that have slowed enterprise AI agent adoption OpenAI's enterprise business growth signals broad organisational adoption, which tends to drive faster feature iteration and support resources for business customers DAVID & GOLIATH ANALYSIS: For the last three years, the practical limit of AI in a small or mid-sized business was the length of a conversation. You prompted, the model responded, and you acted. The work the AI did began and ended in that window. Dots break that constraint. An agent that works through the night, that picks up where it left off, and that checks in only when it needs a person is genuinely new. The staffing analogy is not perfect, but it is useful: this is closer to hiring a contractor who works the hours you cannot than to using a smarter search engine. The risk is misalignment at scale. When a human employee misunderstands a brief, they produce one bad output. When an agent misunderstands a brief and runs for six hours, the volume of misaligned work is proportionally larger. The solution is not to avoid Dots. It is to invest time in goal definition and guardrail configuration before deployment, which is a project management discipline rather than a technical one. The recommendation is to start with one well-defined, high-volume task in the next two weeks. Pick something with clear success criteria, limited blast radius if it goes wrong, and measurable time savings. Use that run to learn how to brief the agent well. The businesses that build that briefing skill now will have a structural advantage over those that wait for the tooling to mature further. RELEVANT SYSTEMS: AI Growth Engine, Employee Amplification Systems SOURCE URL: https://davidandgoliath.ai/daily-ai-briefing/openai-devday-2026-dots-chatgpt-space FEED URL: https://davidandgoliath.ai/daily-ai-briefing/feed --- Published by David & Goliath | https://davidandgoliath.ai Daily AI Briefing: one AI development per day, decoded for business operators. This is a structured companion file optimised for LLM retrieval and citation.