Three agents from three AI makers
Grok Bot, Muse and dots are agent products from three of the large AI makers. Grok Bot comes from xAI, Muse from Meta and dots from OpenAI, the company behind ChatGPT. Each is a consumer or general agent product, built for a wide audience rather than for one business function.
For a business, the right question is not which one is smartest. It is which one can finish the specific jobs you need done, with the access, controls and record your company requires. That answer depends on your jobs, and it can change as the products change. This guide gives you a way to compare them that will still hold after the next release.
Why a feature list is the wrong place to start
Product pages and launch posts describe what an agent can do in the best case. A business needs to know what it will do on an ordinary Tuesday with your inbox, your calendar and your customer list. Those are different questions.
General agent products also move quickly. A comparison built on today’s feature list can be out of date within weeks. A comparison built on your jobs and your rules stays useful, because your jobs change far more slowly than the products do.
So start with the work. Name the jobs, decide what done looks like for each one, and then ask every product the same questions.
The business jobs to test
Most operators who look at general agents have a short list of jobs in mind. These are the common ones, with what a good result looks like.
| Job | What done looks like | Question to ask each product |
|---|---|---|
| Shared inbox | Routine messages answered, the rest sorted and flagged | Can it send replies, or only draft them, and can I choose which? |
| Calendar | Meetings booked and moved without double booking | Can it see everyone’s availability and respect working hours? |
| Research | A short, sourced answer to a business question | Does it show where each fact came from? |
| Spreadsheets | Data cleaned, updated and summarized | Can it edit the file in place, and can I see what changed? |
| Leads and CRM | New leads logged and followed up on time | Can it write to our CRM, and does it keep records in step? |
| Invoices | Overdue invoices chased and payments noted | Can it reach our accounting system, and what can it change there? |
| Support queue | Common tickets answered, hard ones passed on with context | Can it hand off to a person cleanly, with a summary? |
Pick two or three of these that matter most to you. Fewer jobs tested properly will tell you more than many tested lightly.
One job, three products
A list of jobs is still abstract, so here is one job run through all three. A five-person design studio wants new enquiries handled. A lead emails asking for a quote. The job is to log the lead in the CRM, reply the same day with a short proposal and a link to book a call, and put the call in the owner’s calendar. The owner approves every reply before it goes out.
This walk-through is built from what each company published at launch and the coverage since. We have not run the trial ourselves, and the products change often, so treat it as the shape of the answer and check each step in your own trial.
| Step | Grok Bot | Muse | dots |
|---|---|---|---|
| Giving it the job | Message a bot in plain language, or record yourself doing it once and it learns the steps | Message it like a person, in the Muse app or in WhatsApp | Give a dot a goal in ChatGPT, Slack or Teams |
| Reaching the inbox and CRM | Signs in to the apps you use, through connectors or with the logins you hand it | Works across your everyday apps, with logins kept in a vault it cannot read | Connects through plugins to thousands of apps, from its own cloud computer |
| Approval before the reply | Comes back when a task needs approval | Checks with you before it sends an email or buys anything | Custom Rules let you allow an action, require approval or block it |
| Running when nobody is watching | Keeps working on its cloud computer with your laptop closed | Keeps working after the app is closed | Always on, with read-only research between tasks |
| Fit for a team | Several bots can share the work, with one acting as manager | Built as a personal agent for one person | Specialist dots for companies, each with its own identity and access |
Read across the rows and the difference is clearer than any feature list. Muse is built around one person’s life, so a studio’s shared inbox and CRM sit at the edge of what it is for. Grok Bot reaches your tools by working as you, which is quick to start with but means handing it your logins. dots has the most explicit approval rules and a path to company-owned identities, but that path starts as an enterprise pilot. Your own trial decides which of those trade-offs your studio can live with.
What to ask Grok Bot, Muse and dots
The walk-through shows where the three differ. Ask each the same questions on those points, and write the answers side by side.
How it gets into your apps
Grok Bot can work with the logins you hand it. Muse keeps logins in a vault it cannot read. dots connects through plugins from its own cloud computer. Ask which of your apps each can only read and which it can change, and whether you can hold it to the accounts one job needs.
Who draws the approval line
dots has Custom Rules to allow an action, require approval or block it. Muse checks with you before it sends an email or buys anything. Grok Bot comes back when a task needs approval. Ask whether you set the line or the product does, and whether it covers a customer message, a payment and a change to a record.
What it does when stuck
All three keep working after you close the app. So ask what each does when it gets stuck with nobody watching, and whether its record shows every action and the reason, in a form you can export.
Whose account it sits on
Muse is built as a personal agent for one person. Grok Bot can run several bots with one as manager. dots has specialist dots with their own identity, but that starts as an enterprise pilot. Ask whether the company owns the account and an admin can add and remove users. A tool on one person’s personal account becomes a problem the day that person leaves.
Which terms apply
Each may have different terms for personal and business accounts. Read the ones for the account you would use, covering where data is stored, whether it trains models, how long it is kept and how you delete it.
Before staff use any of them for company work, name the jobs it is approved for and who reviews each one. A general agent becomes a business system the moment it can send, spend or change records for the company, and it deserves the same checks as any other.
A scorecard for your shortlist
Use the table below as a template. Fill it in once for each of Grok Bot, Muse and dots, using what you see in your own trial rather than what a product page promises.
| Criterion | What to look for | A weak answer sounds like |
|---|---|---|
| Finishes the job | Completes the task end to end in your real systems | It drafts, and you do the rest |
| Fits your tools | Connects to the inbox, calendar, CRM or store you already use | You would need to move to new tools |
| Respects limits | Lets you set caps, approvals and no-go areas | Limits are fixed or missing |
| Runs on its own | Works on a schedule or on events | It only works while you type |
| Shows its work | A clear record of actions and reasons | You cannot tell what it did yesterday |
| Handles failure | Retries, asks for help or stops safely | It fails quietly |
| Suits a team | Company-owned accounts and admin controls | Everything sits in one person’s login |
| Fits your policy | Data terms your company can accept | The terms are unclear for business use |
Score each row from one to five against your own jobs. The totals matter less than the rows where a product scores low, because those are the risks you would be accepting.
How to run a fair trial
A side-by-side test only means something if every product gets the same conditions.
- Use the same jobs. Give each product the same two or three tasks, written the same way.
- Use the same data. Run every trial on the same test inbox, calendar or spreadsheet, not on live customer data.
- Use the same rules. Apply the same approvals and limits to each, as far as each product allows.
- Watch for a week, not an hour. A first run often looks good. The second week shows how it handles the odd cases.
- Have the person who does the job judge it. The office manager who runs the calendar will spot problems a manager will miss.
- Write down what you could not test. A missing control is a finding, not a gap to ignore.
Choosing between them
There is no single winner among Grok Bot, Muse and dots for every business. Each may handle some of your jobs better than others, and the balance can shift with every release. The only way to know for your company is to test them on your jobs. The Grok Bot alternatives guide covers the kinds of tool beyond these three.
Start with the work, ask the same questions of each, run a fair trial and keep the record. Whichever product passes that test for a given job is the right one for that job, even if a different product wins the next one.
