- OpenAI Dots workflow automation lets an always-on agent keep working between conversations, powered by GPT-6 Astra.
- Each dot runs on its own dedicated cloud computer with a browser for multi-step tasks.
- Connect apps like Slack and email (4,000+ supported connections) so the dot can act across services.
- Use scheduled tasks for recurring checks, reminders, and proactive updates.
- Autonomy settings and workspace controls define exactly what a dot may do without approval.
What Is OpenAI Dots Workflow Automation?
OpenAI Dots workflow automation is the practice of assigning persistent, goal-oriented work to an always-on AI agent inside ChatGPT instead of issuing one-off prompts. Announced at DevDay on September 29, 2026, Dots are driven by GPT-6 Astra, a flagship model built for complex reasoning, computer use, and research tasks. Unlike a normal chat session that ends when you close the tab, a dot keeps its goal, its files, and its connected tools across work sessions.
Every dot operates with three pillars that make automation possible:
| Pillar | What It Does | Why It Matters for Automation |
|---|---|---|
| Dedicated cloud computer | Gives each dot its own remote environment with a browser | Long tasks keep running without your local device |
| Connected apps | Links tools like Slack, email, and workspace apps (4,000+ connections) | Workflows can span multiple services |
| Communication channels | Delivers updates via ChatGPT, SMS, email, and Slack | Results arrive without you re-prompting |
A regular ChatGPT conversation answers a question. A dot pursues an ongoing goal, schedules its own follow-ups, and reports back proactively through your chosen channels.
The strongest automation candidates are workflows where context must persist: ongoing research, recurring monitoring, calendar review, and cross-app follow-ups that would be tedious to restart in every chat.
Step-by-Step: Building Your First Automated Workflow
Follow this sequence to turn a manual, repeated task into a self-running workflow. The setup flow runs from ChatGPT on desktop web or the supported desktop experience.
Create and Name Your Dot
Open the Dots experience in ChatGPT and start a new dot. Choose a short, descriptive name tied to its responsibility, such as "Weekly Research Digest" or "Inbox Triage."
Define an Ongoing Goal
Write a clear outcome the dot should continually work toward, not a single prompt. A focused goal keeps the dot's direction stable across multiple actions and work sessions.
Set Working Instructions and Autonomy
Describe what the dot should prioritize, what boundaries it must follow, and how independently it may act. Decide when it should proceed on its own versus ask for approval.
Connect Apps and Attach Context
Link the services the workflow needs, such as Slack or email, using only the permissions you grant. Attach files and reference material so the dot has full project context.
Assign a Scheduled Task
Tell the dot what should happen and when. Tasks can run at a specific future time or repeat on a recurring schedule for checks, summaries, and follow-ups.
Review Activity and Refine
Open the activity dashboard to inspect completed actions, active tasks, and scheduled work. Adjust instructions and autonomy based on what you see.
Always start with narrow autonomy and a limited set of connected apps. Expand permissions only after the dot has demonstrated reliable behavior on repeated runs.
Scheduled Tasks and Recurring Work Patterns
Scheduled tasks are the engine of OpenAI Dots workflow automation. They transform a dot from a conversational helper into an ongoing assistant that executes work between your visits.
| Task Type | Trigger | Example Use Case | Output Style |
|---|---|---|---|
| One-time future task | Specific date and time | Prepare a briefing before a meeting | Delivered report at set time |
| Recurring check | Repeating schedule | Monitor a topic and flag changes | Digest or alert on update |
| Recurring summary | Daily or weekly cadence | Summarize calendar and inbox each morning | Structured summary via channel |
| Follow-up reminder | After prior work completes | Chase an unanswered email thread | Nudge with context included |
The task management view shows active, scheduled, and completed work in one place, so you can audit exactly what the dot did and when. Proactive updates surface results through your configured channels, meaning useful information arrives without starting a new conversation.
Recurring Monitoring
- Watch changing information
- Surface only meaningful updates
- Runs on its own schedule
Proactive Updates
- Results pushed to you
- Arrives via Slack, email, or ChatGPT
- No manual re-prompting
Task Audit Trail
- Full activity history
- Active vs completed tasks
- Reviewable at any time
Pair every recurring task with a clear output format in the dot's instructions. A summary that arrives in a consistent structure is far easier to act on than free-form responses.
Cloud Computer and Cross-App Workflows
Two capabilities separate dot automation from simple scheduled reminders: the dedicated cloud computer and connected-app workflows. Together they let a dot execute browser-based, multi-step work that continues independently of your machine.
The cloud computer gives each dot its own remote environment with a browser. Cloud-based work keeps running while you move on to other activities, and you can open the computer view at any time to inspect its state or step in when manual input is useful. Importantly, cloud execution is separate from optional local-computer access, so you can distinguish remote agent work from anything involving your own device.
Cross-app workflows build on this foundation. With connected tools granted through user or workspace permissions, a dot can gather information in one service, act in another, and report through a third:
| Workflow Stage | Typical Tool | What the Dot Does |
|---|---|---|
| Gather | Browser on cloud computer, connected apps | Collects and organizes information |
| Process | Files and project context | Synthesizes findings, drafts outputs |
| Act | Email, Slack, workspace apps | Sends messages or updates services |
| Report | Communication channels | Delivers results and flags exceptions |
Reserve cloud computer tasks for work that genuinely requires browser interaction or many sequential steps. Pure information tasks often run fine with connected apps alone and are easier to audit.
Permissions, Controls, and Safety
Automation is only useful when it is controllable. Dots use configurable permissions covering connected services, computer capabilities, autonomous behavior, and organization-level availability.
| Control | Managed By | What It Governs |
|---|---|---|
| Connected app permissions | User and workspace | Which apps and services the dot may access |
| Cloud computer access | Dot capability | Whether browser-based, multi-step tasks are allowed |
| Local computer access | Optional user capability | Any interaction with your own device, kept separate from cloud work |
| Autonomy settings | User configuration | How independently the dot continues work between interactions |
| Workspace availability | Administrator | Whether members can use Dots at all in managed environments |
| Enterprise controls | Enterprise admin | Workspace-level restrictions and governance |
OpenAI's official safety documentation includes a dedicated Dots section, reflecting that these controls are part of the product's core design rather than an afterthought.
Availability is plan-dependent and rollout-stage dependent. Workspace administrators control whether members can use Dots and which connected capabilities are permitted, so check your workspace settings before designing workflows.
Use this checklist to keep every automated workflow safe and reviewable:
Workflow Safety Checklist:
- Define a single focused goal per dot
- Grant only the app connections the workflow needs
- Set autonomy to require approval for sensitive actions
- Review the activity dashboard on a regular cadence
- Keep cloud and local computer access separated
FAQ
Q: What is OpenAI Dots workflow automation best used for?
It works best for ongoing work where context must persist: recurring research, calendar and inbox reviews, scheduled monitoring, reminders, and multi-step workflows that span connected apps like Slack and email. One-off questions are better handled in a normal ChatGPT conversation.
Q: Do I need to keep my computer on for a dot to keep working?
No. Each dot runs on its own dedicated cloud computer with a browser, so long-running tasks continue independently of your local device. You can open the computer view anytime to inspect progress or intervene.
Q: How do scheduled tasks deliver results?
After a scheduled task finishes, the dot surfaces results proactively through your configured communication channels, such as ChatGPT, email, or Slack, without requiring a new prompt from you.
Q: Who controls whether Dots are available in my organization?
Availability depends on your ChatGPT plan, region, and rollout stage. In business and enterprise workspaces, administrators manage whether members can use Dots, which connected tools are allowed, and what governance applies.