OpenAI Dots vs Meta Muse for Small Business: The Always-On Agent Era Arrives

For three years, the dominant AI interface was a chat box: you ask, it answers, it waits. On September 29, 2026, two of the largest AI companies shipped products built around a different idea: agents that keep working when you are not talking to them.

OpenAI launched Dots at DevDay 2026 in San Francisco. The same day, Meta expanded its Muse agent to small businesses. Here is what each actually does, how they differ, and what builders should take from the design choices.

OpenAI Dots: Agents With Their Own Computer

According to OpenAI's launch coverage:

• Model: Dots run on GPT-6 Astra.
• Runtime: each Dot gets its own cloud computer and browser, isolated unless the user connects it to their accounts.
• Integrations: connects to more than 4,000 apps through plugins.
• Channels: reachable in ChatGPT (web, desktop, mobile), Slack, and Microsoft Teams, with text messaging announced as coming.
• Always-on behavior: Dots can work across multiple projects, send progress updates, ask questions mid-task, and do "proactive research" in the background. That proactive mode is restricted to read-only access to connected apps.
• Visibility: users can watch the Dot's computer while it works instead of waiting for a final answer.

Controls and Safety

The launch leans heavily on permissioning, which is the right instinct:

• Built-in rules decide what a Dot can do autonomously and what needs approval.
• Custom Rules let users "allow, block or require approval for specific actions."
• Auto-review applies to actions that affect accounts, and sensitive tasks such as password changes require user approval.

Availability and Pricing

• One Dot is included at no extra cost for ChatGPT Pro and Business Premium users in eligible markets.
• For Pro, reports list eligible markets as excluding the European Economic Area, Switzerland, and the UK at launch.
• Enterprise, Edu, and Healthcare workspaces get a beta when an admin enables it, with enterprise pilots for specialist Dots.
• OpenAI said Dots conversations do not count toward ChatGPT usage limits, and described options to scale a Dot's speed or monthly work volume. Coverage noted no published per-Dot price for extra capacity yet.

Meta Muse for Small Business: Context Over Autonomy

Meta's Muse agent launched earlier in September in the US and Canada. On September 29, Meta expanded it with Muse for Small Business.

• Goal: help owners run operations and customer communication using context about what the business sells, its brand voice, and its customer inquiries.
• Integrations: Shopify, Dropbox, Slack, Asana, Box, Canva, Figma, Granola, HighLevel, Intuit QuickBooks, Klaviyo, Lovable, Notion, Stripe, and Zoom, plus Instagram professional analytics, Facebook Pages, and Meta ad accounts.
• Tasks cited in coverage: customer communication, cash flow and inventory processes, and website creation.
• Pricing: a free tier with usage limits, with reported paid tiers at $20 and $100 per month.

Meta's framing, quoted by TechCrunch: small business owners "told us they're short on hours, not ideas."

Side by Side

	OpenAI Dots	Meta Muse for Small Business
Launch	Sept 29, 2026 (DevDay)	Sept 29, 2026 (expansion of Muse)
Core idea	General always-on agent with its own computer	Business-context agent inside existing tools
Execution environment	Dedicated cloud computer and browser	Integrations with business software and Meta properties
Integrations	4,000+ apps via plugins	Shopify, Stripe, QuickBooks, Slack, Notion and more
Background work	Proactive research, read-only	Operations help grounded in business data
Controls	Allow / block / require-approval rules, auto-review	Account-level integrations and usage tiers
Pricing	One Dot included in Pro and Business Premium	Free tier; reported $20 and $100 monthly tiers

Architecture Lessons for Teams Building Their Own Agents

Both launches validate patterns I have been recommending to clients building internal agents.

1. Give agents a sandbox, not your laptop

Dots run in isolated cloud computers. That is the correct default: the blast radius of a confused or manipulated agent is limited to its own environment until a human connects more.

For your build: run browser and code-execution agents in ephemeral containers or VMs, with network egress controls and no standing credentials.

2. Separate "observe" from "act"

OpenAI restricts background proactive research to read-only access. Reading is where agents deliver most value with least risk; writing is where incidents happen.

For your build: issue read-scoped tokens by default, and require a separate, approval-gated flow to obtain write scopes.

3. Make permissions user-legible

"Allow, block, or require approval" is a permission model a non-engineer can understand. That matters because the person approving actions is usually a business user.

For your build: express policies as a small set of human-readable rules per tool and action, backed by enforcement at an API gateway, not by prompt instructions.

4. Context is the moat for vertical agents

Meta's bet is that an agent knowing your catalog, customers, and brand voice beats a smarter generic agent. For most businesses, that is correct: retrieval quality and integration depth decide usefulness more than raw model capability.

5. Always-on means always-attacked

An agent that reads email and web pages continuously is continuously exposed to prompt injection. Treat every piece of content the agent reads as untrusted, keep authorization outside the model, and log every action with enough detail to reconstruct what happened.

Key Takeaways

• OpenAI Dots (GPT-6 Astra) are always-on agents with their own cloud computer and browser, 4,000+ app integrations, and allow/block/approval controls; one Dot is included with Pro and Business Premium.
• Meta Muse for Small Business grounds an agent in business context through integrations like Shopify, Stripe, and QuickBooks, with a free tier and reported $20 and $100 paid tiers.
• The shared design lessons: sandboxed execution, read-first permissions, user-legible approval rules, and deep context.

Sources

• The Next Web: OpenAI launches Dots, always-on AI agents with their own cloud computers
• 9to5Google: OpenAI launches Dots, new always-on agents
• Traictory: OpenAI's dots and the pricing ladder
• TechCrunch: Meta is expanding its AI agent Muse to small businesses
• CNBC: Meta launches Muse for Small Business
• Axios: Meta takes fresh aim at businesses with Muse tools