- OpenAI Dots are persistent AI agents that keep working around the clock on assigned jobs
- Each Dot gets its own cloud computer and connects to 4,000+ apps like Slack and email
- Availability: Live on ChatGPT Pro and Business Premium; Enterprise in beta
- Beginner rule: Start with one small, low-risk job before expanding responsibility
- Current state: Early reviewers report bugs, so supervise sensitive tasks closely
What Are OpenAI Dots?
Traditional AI assistants complete a single task and stop — book a table, draft an email, done. An OpenAI Dot flips that model entirely. You hand it a responsibility, not a prompt. A Dot can watch incoming bug reports, test fixes, chase a vendor who cancelled, or monitor an inbox while you sleep.
Each Dot runs on its own dedicated computer in the cloud and plugs into more than 4,000 applications you already use, including Slack, your calendar, and email. It learns how you prefer things to be done, and for anything sensitive, it pauses and asks for your approval first.
On stage at DevDay 2026, OpenAI's Holly Lee demonstrated a full product launch run alongside her Dot, "Dotty." When a review meeting moved, Dotty had already shifted her calendar, flagged new feedback, and sent an updated design — without being asked.
| Feature | Traditional Assistant | OpenAI Dot |
|---|---|---|
| Task scope | One task, then stops | Ongoing responsibility |
| Uptime | Only while you chat | Works around the clock |
| App access | Limited per request | 4,000+ integrated apps |
| Learning | None | Learns your preferences |
| Sensitive actions | N/A | Stops and asks you first |
Dots launched with visible rough edges. During the DevDay demo, Dotty stalled mid-presentation, and early-access reviewers have reported that Dots are currently too buggy to recommend for critical workflows unsupervised. Treat your first Dot as an intern, not an autopilot.
Availability, Plans, and Limits
Dots went live following DevDay 2026, but access depends heavily on your plan and region.
| Plan | Dot Access | Notes |
|---|---|---|
| ChatGPT Pro | Available | Adults only; not yet in Europe, UK, or Switzerland |
| Business Premium | Available | Full rollout at launch |
| Enterprise | Beta | Gradual rollout |
| Plus / Free | Not available | Upgrade required |
Your first Dot is included with an eligible plan, and simply chatting with it does not count against your usage limits. Real work the Dot performs runs through Codex, and that usage does draw on your plan's quota.
OpenAI reopened the $200 Pro plan with half the Codex usage it previously included. Existing subscribers keep their old limits for roughly one month before the reduction applies — check your usage page before the cutoff.
A useful companion for Dot owners is ChatGPT Space, a shared page where teammates and their Dots work side by side. You tag a Dot in a comment the way you'd tag a coworker, and it handles the task — for example, turning customer feedback into a bar chart that refreshes hourly. Space is available on Pro, Business, and Enterprise (desktop and web), with shared slides rolling out in the following weeks.
Beginner Setup: Your First Dot in 5 Steps
The single most important beginner principle: start small and verify trust before delegating more.
Confirm Eligibility
Verify you are on ChatGPT Pro or Business Premium, are an adult user, and are located in a supported region (not Europe, UK, or Switzerland yet).
Create Your First Dot
Set up your Dot from your eligible account. Your first Dot is included, and conversation with it does not consume your message limits.
Connect One Tool
Link a single low-stakes channel first — one Slack channel or one inbox. Resist connecting everything on day one.
Assign One Small Job
Give the Dot a bounded responsibility, such as "watch this inbox and summarize overnight messages each morning."
Review and Expand
Check its output daily for a week. Only add responsibilities — bug triage, calendar management, vendor follow-ups — once accuracy is consistent.
Safe Starter Jobs
- Inbox monitoring and daily summaries
- Slack channel bug-report watching
- Drafting responses for your review
Medium-Trust Jobs
- Calendar reshuffling after changes
- Turning feedback into recurring charts
- Drafting fixes and pull requests
Hold Off For Now
- Unreviewed sends to customers
- Financial or legal actions
- Anything irreversible without approval gates
OpenAI engineers reported their Dots fixing dozens of bugs per day autonomously — a dropped bug report in Slack was picked up and fixed without human prompting. That is the ceiling of what's possible; your starting point should be far lower.
Common Beginner Mistakes
Because Dots run continuously, small setup errors compound quickly. These are the pitfalls new users hit most often.
| Mistake | Why It Hurts | Fix |
|---|---|---|
| Granting access to everything at once | Errors spread across many apps | Connect one tool, add slowly |
| No approval gate on sensitive actions | Dot acts before you see it | Require ask-first on anything irreversible |
| Vague job descriptions | Dot guesses your intent | Write jobs like onboarding docs |
| Ignoring Codex usage | Real work burns quota fast | Monitor usage weekly |
| Assuming reliability | Early builds stall and err | Audit outputs daily at first |
Even OpenAI's own on-stage Dot froze mid-demonstration. If the flagship demo can have a slow morning, your first-week Dot certainly can. Build human checkpoints into every workflow.
The Broader DevDay 2026 Ecosystem
Dots work best when paired with the other launches announced alongside them.
| Launch | What It Does | Beginner Relevance |
|---|---|---|
| GPT 6.1 Soul | Near-flagship model at about one-fifth the price | Cheaper engine for Codex work |
| Ultrafast mode | Up to 8x faster in Codex | Pro 500 plan ($500/mo) only |
| Cloud Codex | Coding agent runs in the cloud | Start jobs, close laptop, check from phone |
| Decisions API | Picks from options in ~150ms | For builders of games and live apps |
| Sign-in with ChatGPT | Users' plans pay their AI usage | Big for solo app builders |
Sam Altman framed the moment not as an industrial revolution but as a potential renaissance — "AI should be about giving people more power over their own lives." For beginners, that power starts with one well-scoped job and a week of careful review.
First Week With Your Dot:
- Confirm plan eligibility and region support
- Connect exactly one tool or channel
- Assign one small, low-risk recurring job
- Review the Dot's output every morning
- Check Codex usage before limits change
FAQ
Q: What is an OpenAI Dot in simple terms?
A Dot is a persistent AI agent that takes on an ongoing responsibility instead of a single task. It runs on its own cloud computer, connects to thousands of apps, and keeps working around the clock — pausing to ask you about anything sensitive.
Q: Which plans include OpenAI Dots?
Dots are live on ChatGPT Pro and Business Premium, with Enterprise in beta. Pro access is adults only and not yet available in Europe, the UK, or Switzerland. Your first Dot is included, and chatting with it does not count against your message limits.
Q: What is the best first job for a beginner Dot?
Pick something small, observable, and low-risk — such as monitoring one inbox and delivering a morning summary. Verify accuracy for a week before expanding into bug triage, calendar management, or drafting fixes.
Q: Are Dots reliable enough to run unattended?
Not yet for critical work. The DevDay demo stalled on stage, and early-access reviewers describe the current build as buggy. Treat your Dot as a supervised assistant, with approval gates on anything irreversible.