The September 2026 AI Agent Wave: Grok Bot, Muse, Dots

Seven weeks produced four launches that changed what an AI agent means. Grok Bot arrived August 11, 2026 with always-on agents for teams. Meta Muse followed September 8 as a personal agent for everyone. Grok 4.7 landed September 21 as the model tuned for long, multi-hour work. ChatGPT Dots closed the run on September 29 with always-on agents for work and errands. Each announcement had its own framing, but the direction is shared: agents with their own computers, working in the background, asking for approval at the right moments.

This roundup puts the wave in order, explains what is genuinely new, compares the four releases side by side, and gives owners a clear takeaway on what to do this quarter.

Key takeaways

• Grok Bot launched August 11, 2026: always-on agents with their own computers that work inside tools and apps around the clock, built jointly by SpaceXAI and Cursor.
• Meta Muse launched September 8, 2026 in the US: a consumer personal agent for emails, shopping, travel, and forms, reached through WhatsApp, its app, and the web.
• Grok 4.7 launched September 21, 2026: the most capable Grok model for coding and knowledge work, priced at 2 dollars per million input tokens and 6 dollars per million output tokens.
• ChatGPT Dots arrived September 29, 2026: always-on agents powered by GPT-6 Astra, each with a cloud computer, 4,000-plus app connections, and Slack and Teams access.
• The shared pattern is own-computer agents, background work after the app closes, and approval-gated actions.
• For owners, Dots and Grok Bot are the work-relevant pilots; Muse is personal admin; proven call and booking automation still carries the front door.

Timeline: what launched and when

August 11 brought Grok Bot, described as a team of always-on agents. Each Bot has its own computer, works inside tools and apps the way a person would, keeps working around the clock, runs in parallel with other Bots, and can message other Bots and share work. The company behind it says its own teams used Bots first for sales outbound, marketing campaigns, office operations, and bug fixes. Access runs through SuperGrok tiers and Cursor plans, with Bots carrying their own usage separate from Grok and Cursor plan limits. Enterprise controls for access, network, and audit followed on September 3, including two weeks of free usage for Grok and Cursor Enterprise customers.

September 8 brought Meta Muse to the US. Meta positioned it as the first personal AI agent built for everyone, available through a dedicated app on iOS and Android, the web, and WhatsApp, with a Mac app added September 17. It handles everyday tasks such as sending emails, shopping, booking travel, and filling out forms, running on a Muse Secure VM with a more locked-down encrypted version planned later in 2026. Reuters reported that internal tests showed stalling and unauthorized exposure of sensitive data before launch, and that Meta proceeded anyway, a caveat owners should factor into permission decisions.

September 21 brought Grok 4.7, the model upgrade behind much of the xAI agent story. It was trained with a longer reinforcement-learning run on harder, many-hour tasks, verifies its own work better, manages long context, and natively understands the Grok Bot harness. Pricing is 2 dollars per million input tokens and 6 dollars per million output tokens, with a fast variant at twice the speed and price. It is available in Cursor and Grok Build, through the Grok API, and in third-party harnesses and model routers.

September 29 brought ChatGPT Dots at OpenAI Dev Day. Powered by GPT-6 Astra, each Dot gets its own cloud computer and browser, connects to more than 4,000 apps, and is reachable through ChatGPT, mobile, Slack, and Microsoft Teams, with voice calls supported and SMS planned. Background research runs read-only, actions can require approval, and Enterprise and education access sits in admin-gated beta. Specialist Dots for procurement, invoice processing, customer support, and contracts are in testing.

What is genuinely new

Strip away the launch language and three real shifts remain.

First, agents now get their own computers. Grok Bot gives each Bot its own computer. Dots each get an isolated cloud computer and browser. Muse runs on a dedicated Secure VM. The practical effect is the same: the agent works in its own workspace rather than on your laptop, which enables longer runs and cleaner separation between your device and the agent's actions.

Second, work continues after you leave. Grok Bot agents keep working around the clock. Muse keeps working after the app closes and returns for approvals. Dots run assigned jobs in the background while you watch or redirect through an Activity View. The agent era is defined less by smarter answers and more by unsupervised minutes between check-ins.

Third, supervision is part of the product, not an afterthought. Read-only research modes, approval tiers for actions, admin switches for Enterprise access, and audit controls all shipped alongside the agents. That reflects hard experience: agents that can act inside real accounts need brakes as standard equipment. Our Grok Bot explainer and ChatGPT Dots explainer both walk through those controls in detail.

Side-by-side comparison

Pricing in brief: Grok Bot access bundles with SuperGrok and Cursor tiers with separate Bot usage, plus Enterprise controls from September 3. Muse is free for most needs with paid plans for heavier use, details still to come. Grok 4.7 lists API pricing at 2 dollars per million input and 6 dollars per million output tokens. Dots roll out with ChatGPT Pro and Business Premium, the first Dot included with extended first-month limits.

The buyer takeaway by business type

For office-heavy teams, Dots are the most natural pilot. The Slack and Teams presence plus thousands of app connections means the agent meets the team where coordination already happens. Start with background jobs such as feedback monitoring or research briefs, with approvals on, before touching anything involving money or customer records.

For technical and product teams on Cursor or SuperGrok plans, Grok Bot deserves the parallel look. The internal prototype story, outbound, campaigns, ops, bug fixes, maps well onto teams that already live in those tools. The September 3 Enterprise controls give larger teams the access and audit story to run a governed trial.

For owners' personal productivity, Muse is a reasonable experiment with narrow permissions. Keep it on personal admin, require approvals for purchases and outbound messages, and keep every customer record out of it. Our Meta Muse explainer has a first-week setup table for exactly that.

For customer-facing operations, none of these launches changes the current answer. Calls, bookings, quotes, and follow-up need dedicated systems with phone integrations and CRM records. An AI receptionist for answering, a lead follow-up workflow for cadence, and CRM automation for records remain the stack that pays. New general agents can assist around the edges; they should not carry the front door yet.

What to watch next

Three things due after this wave deserve attention without action today. Specialist Dots for procurement, invoicing, support, and contracts are in testing with Microsoft Agent 365 controls to follow; invoice and purchasing work should wait for general availability. Muse's confidential encrypted VM version is planned later in 2026 and will matter for anyone weighing personal-agent permissions. Grok Bot's Enterprise rollout, with its two-week free usage window for qualifying customers, will show how quickly governed team agents spread beyond early adopters.

The pattern to track across all three is the same: availability widening from consumer and prosumer tiers into admin-controlled business tiers. When admin controls, audit trails, and scoped permissions arrive for a product, that is the signal a supervised business pilot is reasonable. Until then, experiment personally and keep customer systems on proven tooling.

FAQ

What AI agents launched in August and September 2026?

Four milestones lead the wave: Grok Bot on August 11, 2026, Meta Muse on September 8, Grok 4.7 on September 21, and ChatGPT Dots on September 29. Grok Bot and Dots are always-on agents with their own computers, Muse is a consumer personal agent, and Grok 4.7 is the model upgrade tuned for long, multi-hour work.

What is genuinely new about these agents?

Three things: each agent gets its own computer or virtual machine instead of running on your device, agents keep working in the background after you close the app, and supervision is built in through approvals, activity views, and read-only research modes. Those three patterns appear across Grok Bot, Dots, and Muse alike.

Which launch matters most for a small business?

ChatGPT Dots matter most for office and team work because of 4,000-plus app connections plus Slack and Teams access, while Grok Bot matters for teams already on SuperGrok or Cursor plans. Muse matters least for business operations since it is a consumer agent with no CRM or pipeline features.

What should owners do about this wave right now?

Pick one contained pilot, such as background research or feedback monitoring, with approvals switched on and narrow app permissions. Keep calls, bookings, and follow-up on proven automation while the new agents prove themselves. Revisit broader rollout once Enterprise controls and specialist business agents leave beta.

Next step

Do not try to adopt four launches at once. Choose the one closest to your work, run a single supervised pilot for thirty days, and let everything else wait. To pick the pilot task and set the approval tiers around it, use our free six-step AI automation plan: start your AI automation plan. If you prefer to talk it through first, book a call.