Muse vs Grok Bot vs Dots: Which Always-On Assistant Should a Business Owner Use?

The Short Answer

Muse, Grok Bot and Dots are always-on AI assistants: each gets its own cloud computer, signs in to your apps, and keeps working after you put the phone down. For a business owner or executive, the choice comes down to one question: will the assistant hold only your own logins, or the company's?

• Muse if it is only your own life: your inbox, your calendar, your errands. Free to start, US and Canada only, and Meta sells no team controls for it.
• Grok Bot if it will touch the business: the CRM, the shared inbox, the team. It is the only one of the three with a team plan on sale today, and for a business it is the best of the three right now.
• Dots if the company already runs on ChatGPT. A dot comes with ChatGPT Pro or a Business Premium seat and is the most polished way to talk to an assistant from anywhere. The company-owned version is still in pilots.

We have used all three. For the personal agent against the work agent, see our two-way comparison of Muse and Grok Bot. This post adds the new entrant, OpenAI's dots, launched at DevDay on September 29, and asks how dots stack up against each of the other two for the person whose name is on the company.

At a Glance

What We Saw: A President, a Car, and a CRM

The deployment that shaped how we think about this is the president of a 50-year-old company, a client of ours, who we connected to the company CRM through Grok Bot. The president now asks "what is going on with my team?" from the car and gets an answer from the CRM, without opening a laptop. That is what an owner actually wants from an always-on assistant. Not a chatbot, but a chief of staff who has read everything since yesterday.

Setting it up was where the real decisions were: what the assistant may read, what it may change, and who answers for it. So we built those answers into our CRM. In our CRM, an always-on agent is its own user: its own login and its own scoped credentials, a role that limits what it can touch, guardrails like approvals before deletes or outbound email, and an audit log that records it as the agent that acted. It never signs in as a person. Each agent also has a named human owner, who is accountable for it and can pause or revoke it.

The same agent role serves Grok Bot, Muse, scheduled Claude routines, ChatGPT agents and dots. The connection wizard ships presets for the first four. Dots are not a preset, but the role is not tied to a vendor: any agent that can call the CRM's tools or API, or sign in with its own browser login, is set up the same way. Every run follows the same standing orders: whoami, then a feed of what changed since its last run, then an inbox of tasks and notes that @mention it, then act within its policy, then save its place. An agent that works through a browser on its own cloud computer gets its own web login.

The presets say a lot about the products. Grok Bot starts with reads and ordinary edits allowed, and deletes and outbound email held for approval. Muse starts stricter, with deletes off entirely, because an agent clicking around its own cloud computer makes mistakes that are hard to see from outside. Pausing an agent takes effect on its next request. Revoking one destroys every credential it holds and cancels anything it had waiting for approval.

The question an owner should ask first is not "what can it do?" It is "how do I take it back?"

What we saw running Dots

Our first impression: a dot is closer to Muse than to Grok Bot. It is one agent that you work with. Grok Bot spins up several Bots at once, which gets you something close to a small team. A dot is a single assistant, not a team.

What it has that the other two do not is everything already inside ChatGPT. The connectors and the knowledge you have set up there come with it, so there is nothing to connect again. It has been helpful from day one. This is a very early look, though. Dots launched on September 29, and we expect how they work to change a lot over the next few weeks.

Our verdict today: for a business, of the three you can actually buy right now, Grok Bot is the best. It is the only one that works like a team, and the only one with team controls on sale.

What an Owner Actually Hands Over

The briefing

The first job for any of these is read-only: a morning summary of the pipeline, overdue invoices, what the team moved yesterday. All three can do it. Dots have the edge on reach here: OpenAI says a dot can be messaged or called in ChatGPT on desktop, web and mobile, and messaged in Slack and Teams, and that its background "proactive research" runs only on tools restricted to read-only. Muse lives in its own app and in WhatsApp, which is where many owners already are. Grok Bot runs on desktop and iOS.

The drafts

The second job is drafts behind approval: follow-ups, status updates, replies. Here the checks matter. Muse runs a separate Sentinel agent on the same machine, isolated at the system level, and nothing reaches the internet unless it approves. Dots use Auto-review, which OpenAI's dots safety post describes as a separate system that checks each planned action, such as an email's recipient and content, against your instructions and Custom Rules. Some steps, such as changing a password or moving money between accounts, always come back to you. Grok Bot comes back when it needs your approval.

We benchmarked that kind of check ourselves across 150 hand-labelled agent actions. A useful gate sent 8% of traffic to a human; a poorly calibrated one caught its errors only by escalating 92%. None of the three vendors publishes that number for its own gate, so judge them in a pilot, not from the launch page.

The company logins

The third job is the one that decides the product: letting the assistant change company systems. Muse is one person's agent. Meta's September 29 launch of Muse for Small Business added what Meta calls "a collection of new skills and connectors inside Muse", including Slack, QuickBooks and Stripe according to TechCrunch, but no admin console and no way for a company to revoke what it was given.

Grok Bot is sold as a team product. According to SpaceXAI's plan page, the $40-a-seat Standard Teams plan adds centralized billing, shared usage analytics and SAML or OIDC single sign-on, and SpaceXAI's Enterprise tier adds SCIM and data residency. Bots can also run in parallel, one on outbound, one on reporting, one on systems, which is how a team uses them.

Dots sit in between. The dot you get today is yours: OpenAI's help center describes it as included in a Pro or Business Premium plan, with a handle of @yourname-dot. What OpenAI previewed next is the part that matters for a company: specialist dots that the company sets up "with its own identity, credentials, and access to the systems it needs", managed through Microsoft's Agent 365. That is a company-owned assistant. It is in focused enterprise pilots, not on sale.

Pick Yours

Answer three questions and the choice mostly makes itself. Change any answer and watch which product drops out.

Where owners get this wrong

The failure is rarely the assistant doing something stupid. It is an owner connecting the company CRM, the shared inbox or the bank portal to an assistant that lives in their own account, because it was free or already included. It works until the day someone else needs to see what it could reach, or turn it off. A personal dot inside a ChatGPT Business workspace makes this easier to get wrong, because it feels like a company tool. Until specialist dots ship, it is not one.

Reuters also reported that Meta shipped Muse despite internal concerns about its handling of sensitive personal data. We have no equivalent reporting on Grok Bot or dots, which does not mean they are safer; it means their testing did not leak. OpenAI's own safety post says it plainly: dots can still make mistakes, so review consequential work.

Do You Need a Partner?

Not to try one. Muse is free, Grok Bot starts at $20 a month, and a dot is already in the plan if you pay for ChatGPT Pro. A partner earns the fee when the assistant touches systems other people rely on: deciding which one gets which logins, writing the policy for what it may do without asking, and giving it its own seat in your CRM instead of your password. That is the work in our agent engagements and our custom CRM builds, and the president in the car is what it looks like when it is done.

Sources

• Introducing dots - OpenAI
• How we build safety, security and privacy into dots - OpenAI
• Getting started with your dot (plans and regions) - OpenAI Help Center
• Premium seats for ChatGPT Business - OpenAI
• ChatGPT pricing - OpenAI
• Introducing Muse: a personal AI agent - Meta Newsroom
• Muse for Small Business - Meta Newsroom
• Meta debuts its Muse AI agent - TechCrunch (pricing)
• Meta is expanding its AI agent Muse to small businesses - TechCrunch
• Grok Bot product page and plans - SpaceXAI
• Grok plan comparison and Enterprise controls - SpaceXAI
• Despite security concerns, Meta launches AI agent Muse - Reuters (via Carrier Management)