OpenAI Dots moved always-on agents out of slide decks and into ChatGPT. On September 29, 2026, at DevDay in San Francisco, OpenAI said Dots are "remarkably capable, always-on agents built to handle everything," powered by GPT-6 Astra, with their own cloud computers and links to over 4,000 apps.

I'm Nitin Kumar, Lead Project Coordinator at Debugged Pro. I work with software and IT agencies in India and the USA. This guide turns the launch into a briefing your sales and delivery teams can share.

OpenAI Dots always-on AI agents interface shown in TechCrunch coverage — OpenAI Dots interface from the DevDay launch. Image credit: OpenAI / TechCrunch

What OpenAI Dots Are and Why Agencies Should Care

A regular ChatGPT thread waits for the next prompt. A dot keeps working when the conversation stops. That is the product shift agencies need to explain in plain language.

According to OpenAI's Introducing dots post, each dot has its own cloud computer and browser. Users can open that computer anytime to inspect work. Dots learn from feedback, pursue goals around the clock, and can connect through ChatGPT, Slack, or Teams.

TechCrunch's DevDay report notes that Dots are meant to operate with less constant oversight than Codex or a normal ChatGPT session. OpenAI also said teams of dots working together are a longer-term plan.

The agency framing that lands with buyers

Clients do not buy "GPT-6 Astra." They buy fewer dropped handoffs. They buy a teammate that drafts the invoice, scopes the bug fix, or refreshes the proposal while people sleep.

Lead with jobs, not model names. Then show the control layer: Custom Rules, auto-review, and human approval for consequential steps.

OpenAI Dots Features Agencies Must Brief Accurately

Get the facts right before the demo. Clients will ask about computers, channels, memory, and cost in the first call.

Own cloud computer: Separate from the user's laptop unless they connect it on purpose.
Plugin reach: More than 4,000 apps through OpenAI's plugin ecosystem.
Multi-channel context: ChatGPT desktop, web, and mobile, plus Slack and Teams. Texting is coming soon.
Proactive research: When idle, a dot looks for ways to help using connected apps in read-only mode.
Custom Rules and auto-review: Decide what runs alone, what needs approval, and what stays with the human.
Optional local access: Laptop connection is off by default through the ChatGPT desktop app.

OpenAI is clear that Dots can still make mistakes. Agencies should treat "review consequential work" as a delivery requirement, not a disclaimer footer.

OpenAI Dots developer workflow turning customer feedback into pull requests — OpenAI Dots feedback-to-fix example from the official announcement. Image credit: OpenAI

Launch Examples That Map to Client Delivery Work

OpenAI's announcement walks through concrete roles. Use them as workshop prompts, not marketing poetry.

Developer feedback to pull requests

A developer dot watches customer feedback for recurring requests, scopes small fixes, builds and tests them, and returns complete pull requests with videos of the change. The human keeps building the next major feature while reviewing those PRs.

Agency takeaway: scope a "small fix factory" pilot with clear acceptance tests, video proof, and a human merge gate. Do not promise unsupervised production deploys on day one.

Product launch revisions

When product scope changes, a launch-focused dot revises positioning, assets, and docs for review. That is familiar agency work. The difference is continuous drafts between meetings.

Research, sales, and content loops

OpenAI also describes research reruns when new data arrives, enterprise sales proof-of-concept support, and content packaging from interview transcripts into clips, show notes, and social drafts.

An early tester example shared by OpenAI and covered in press: developer Dan McAteer's dot noticed a forgotten freelance invoice, drafted it from the email thread, and sent the PDF after he approved it. That is the kind of narrow, high-value job clients understand immediately.

Plans, Pricing Windows, and Regional Limits

Agencies selling into India and the USA need the access matrix before they pitch.

Pro: Rolling out in eligible markets. At launch, Pro access excludes the European Economic Area, Switzerland, and the UK.
Business Premium: Available across supported ChatGPT regions.
Enterprise, Edu, Healthcare: Beta, off by default, enabled by a workspace admin.
Cost: First dot included at no extra cost on Pro and Business Premium.
Usage: Conversations with a dot do not count toward ChatGPT usage limits. Deeper work draws on a plan allowance. OpenAI describes extended limits for the first month after launch; later per-plan terms were not fully published at announcement time.

India-based delivery teams supporting USA clients should confirm which workspace plan the client actually holds. A Pro seat in an excluded region will not match a Business Premium pilot plan.

Setup is desktop-first for creating a dot. After setup, messaging from mobile ChatGPT is supported, but creation on mobile was not available at launch per detailed launch explainers.

OpenAI Dots product launch revision workflow for marketing teams — OpenAI Dots launch-materials example from the official announcement. Image credit: OpenAI

Specialist OpenAI Dots and Microsoft Agent 365

Personal dots work for one person. Specialist dots take on a defined responsibility for the organization, with their own identity, credentials, and system access.

OpenAI said early internal testing covered procurement, invoice processing, email marketing, customer support, and commercial contracting. Rollout starts with focused enterprise pilots where OpenAI engineers help define duties, tools, and approval paths. There is no self-serve specialist marketplace at launch.

OpenAI is also working with Microsoft so specialist dots can sit under Agent 365 governance and security controls. For USA enterprise buyers, that sentence often matters more than the avatar design. Identity and access management is where many agent pilots stall.

For agencies already advising on agent control planes, pair this section with our OpenClaw enterprise AI agent control plane guide. Control planes and managed specialist agents solve different layers of the same client fear: "Who can this bot act as, and who can stop it?"

How OpenAI Dots Compare with Muse, Codex, and Custom Builds

Clients will ask how Dots fit next to tools they already heard about.

Versus Meta Muse: Muse launched earlier for small business workflows. Dots compete in the always-on personal agent category with ChatGPT distribution. See our Meta Muse for small business agency checklist when a client wants a side-by-side brief.
Versus Codex cloud environments: Codex remains a coding and task harness. A dot can start or manage Codex work, but usage there still counts toward normal limits. Use our Codex cloud environments agency guide for shared setup and security questions.
Versus building on the API: Dots have no public API of their own. If a client needs exact cost accounting, custom tools, and reproducible evals, keep building agents on the model APIs. Use Dots for recurring work that never justified a full custom build.

Need help scoping that split for a client roadmap? Start with our software development consulting page.

What Agencies in India and the USA Should Do This Week

Turn the news into a shared checklist for offshore delivery and onshore client success.

For USA agencies selling mid-market and enterprise

Expect questions about data use, Slack posting identity, laptop access, and liability when a dot takes a wrong action. Bring admin settings, Auto-review examples, and kill-switch steps to the first workshop.

Enterprise buyers will also ask whether specialist dots belong under existing Microsoft security tooling. Mention the Agent 365 direction, and be honest that specialist rollout is pilot-based today.

For India agencies owning build and integration

Indian teams often own channel wiring, test scripts, and runbooks. Write acceptance tests for: read-only proactive research, scheduled jobs, approval prompts, password-change refusal, and resume-after-pause behavior.

Also document which work stays in ChatGPT versus Codex or ChatGPT Work, because usage accounting differs. That prevents surprise invoices after the first-month window.

Agency Checklist: Brief Clients on OpenAI Dots

Use this list in discovery and sprint planning.

Name the jobs, not the avatar. Invoice follow-ups, PR scaffolding, proposal refresh, content packaging.
Confirm plan and region. Pro exclusions and Enterprise admin flags change the pilot timeline.
Map Slack and Teams early. Context across channels is a product claim clients will test.
Write Custom Rules before scale. Decide auto, ask, pre-approved, and handoff per action class.
Keep laptop access off until needed. Cloud computer first; local access only with a written reason.
Separate personal dots from specialist pilots. Different identity, credentials, and success metrics.
Require human review for consequential work. Shipping, payments, legal language, and customer emails.
Track usage after month one. Conversations and deeper work are not the same bucket.
Compare with Muse and custom API agents. Pick managed Dots only when the job fits.
Publish a kill switch runbook. Pause, reset, revoke app access, and admin disable paths.

How Debugged Pro Helps Agencies Act on This Signal

We help agencies turn agent launches into scoped delivery plans. That includes channel design, permission matrices, specialist-pilot criteria, QA scripts, and client-facing briefings that sales and engineering can both use.

If a client asks whether to buy OpenAI Dots, wait for Muse, or keep building custom agents, we start with the recurring jobs nobody owns today. That is where always-on agents earn trust: clear goals, tight rules, and review gates that match real risk.

More delivery notes live on the Debugged Pro blog.

Key Takeaways

OpenAI Dots launched September 29, 2026 at DevDay as always-on ChatGPT agents powered by GPT-6 Astra.
Each dot has its own cloud computer and browser and can connect to over 4,000 apps through plugins.
Pro and Business Premium users in eligible markets get a first dot at no extra cost; Enterprise access is admin-enabled beta.
Proactive research is read-only; Custom Rules and auto-review govern independent actions; sensitive tasks stay with humans.
Specialist dots for org roles are entering focused pilots, with Microsoft Agent 365 governance in progress.
Agencies in India and the USA should brief jobs, permissions, channels, and usage windows before promising autonomy.

FAQ: OpenAI Dots for Agency Teams

What are OpenAI Dots?
Always-on agents inside ChatGPT announced at DevDay on September 29, 2026. They run on GPT-6 Astra, use a dedicated cloud computer and browser, and can reach over 4,000 apps through plugins.

Which plans get access?
Pro and Business Premium in eligible markets, with Pro excluding the EEA, Switzerland, and the UK at launch. Enterprise, Edu, and Healthcare need an admin to enable the beta. The first dot is included at no extra cost on Pro and Business Premium.

What are specialist dots?
Organization-owned agents with their own identity and credentials for defined roles such as procurement or customer support. They start in focused enterprise pilots, with Microsoft Agent 365 integration underway.

How do permissions work?
Custom Rules and auto-review decide what a dot can do alone. Proactive research stays read-only. Password changes and similar sensitive tasks always stay with the user.

What should agencies do next?
Map jobs and channels, write rules before scale, separate personal and specialist paths, and keep human review on consequential work. Confirm plan, region, and post-month-one usage terms with the client.

About the Author

Nitin Kumar is Lead Project Coordinator at Debugged Pro. He helps software and IT agencies plan practical AI delivery for clients in India and the USA.