How I use Grok Bot to run my inbox and calendar
Most mornings, the first thing I read is a short message from one of my Grok bots with the day’s agenda, the packages due today and a summary of the emails that need me. My inbox is already sorted by then, and what’s left is the stuff that needs a decision.
That’s what the bots do for me day to day. A handful of small ones handle the sorting, tracking and scheduling, and I spend that time on the parts that need judgment. By my estimate, the email side alone gives me back 30 to 60 minutes a day. That’s a gut feel. I haven’t measured it.
A note on the name: xAI calls the product Grok Bot, and each agent is a Bot. I run several, so I call them my Grok bots. They work on a cloud computer xAI runs, so they keep going after I close the app.
My setup, briefly
Grok Bot comes with the X Subscription. I pay for Premium, $30 a month, which bundles Grok Bot, Cursor and X Premium. Each bot has one job:
- Inbox triage. It files orders, newsletters, shopping and updates into their own folders, and flags what looks like it’s from a real person.
- Morning digest. The agenda, packages and important emails, as a chat message in the Grok app. My wife gets her own digest of the day’s events.
- Calendar. We have six kids, so we have a lot of calendars. I ask for changes in plain words, and when school sends home a document, the bot pulls out the dates for each kid’s grade. This is the one I’d miss most.
- Packages. It pulls tracking codes out of my email and tells me what’s late, delivered or postponed.
- Travel. It planned a family trip to Boston and Salem and put the hotel and car confirmations on the calendar. It hasn’t booked anything.
- Data brokers. It found my information on 8 data broker sites and filed removal requests. I’m off 4 of them so far. It runs every quarter to look for new ones, and it added my phone numbers to the Do Not Call Registry.
The details are in two case studies: email triage and the smaller admin bots.
What I’d want you to know before trying it
Three of these come from xAI’s docs on approvals and security, and the last is my own rule.
- All your bots share one computer. That includes files, browser sessions and saved logins. The docs say it plainly: “Do not use separate Bots as a security boundary.” If your package bot is signed into your email, assume your travel bot can get there too. I treat mine as one assistant with several jobs, because underneath that’s what they are.
- Approvals time out on unattended runs. Before a sensitive action, a bot asks you to allow it once, always allow it or deny it. If a scheduled run asks and nobody answers in about 10 minutes, the request expires and the action doesn’t run. That’s the safer way to fail, but a 6 a.m. job that needs your OK won’t happen while you’re asleep.
- “Always allow” means it stops asking. It’s convenient, and it’s the setting I’m most careful with. You can add “Ask first” rules for things you always want to see, and those override it.
- Read-only where I can, and it never sends email for me. It sorts, flags and summarizes. I write and send my own replies. I’ll probably loosen that someday, but not yet.
The alternatives: Muse and dots
Meta and OpenAI both shipped something similar within about seven weeks of Grok Bot. I haven’t used either one, so this comes from their docs and announcements.
Meta’s Muse is the cheapest way to try the idea. Meta’s announcement has the details.
- What it does: It runs on phones, the web and inside WhatsApp, and connects to Gmail, your calendar and Meta’s apps. It’s aimed at errands like shopping, booking and forms.
- Controls: On paper they look careful. Many connectors can be read-only, and by default it asks before sending email or buying. A separate agent has to approve its internet traffic.
- Price: There’s a free tier with a usage meter (you still add a payment card to start), plus $20 and $100 tiers.
- The catch: It’s US only and adults only. Meta says Muse data isn’t shared with its ad systems. Given Meta’s privacy record, I’d read those terms closely before connecting my email.
OpenAI’s dots live inside ChatGPT. OpenAI’s dots docs cover the details.
- What it does: Each dot gets its own cloud computer and browser. It reaches more than 4,000 apps through plugins, plus Slack, Teams and an optional link to one of your computers.
- Controls: The most detailed of the three. You set a rule for each type of action, from acting on its own to handing it back to you, and an automatic review step can’t be turned off.
- Price: This is the catch. It comes with ChatGPT Pro, which starts at $100 a month.
- Limits: It isn’t available in the EEA, the UK or Switzerland, and setup is desktop-only. A dot also doesn’t get its own email address the way a Grok Bot can.
| Grok Bot (xAI) | Muse (Meta) | dots (OpenAI) | |
|---|---|---|---|
| Price to start | $30/mo with X Subscription Premium (not the $8 Plus tier) | Free with a usage meter (card required) | ChatGPT Pro, $100/mo and up |
| Connects to | Connectors, its browser for the rest, its own email address | Gmail, calendar, Meta apps, WhatsApp | 4,000+ app plugins, Slack, Teams, your computer |
| Controls | Allow once, always or deny; unattended approvals expire in ~10 minutes | Read-only connectors, asks before sending or buying | A rule per action type, auto-review always on |
| Best for | Business admin | Personal errands, cheapest trial | Teams already on ChatGPT, Slack or Teams |
Who should try which (my opinion)
I’ve only used one of the three, so weigh this accordingly.
- Muse if you want to see what this feels like without spending much. The free tier should tell you whether you like handing off errands at all.
- Grok Bot if your chores are admin: sorting email, tracking orders, keeping a calendar straight. I think it fits that best, though I’m biased since it’s the one I use.
- dots if your business already runs on ChatGPT, Slack or Teams, because it lives where your team already works. At that price, I’d want a clear payoff before recommending it for one person’s chores.
What this means for a small business
I think the pattern matters more than the brand. Picture an office manager with a shared inbox, where quote requests come in mixed with vendor invoices, promos and shipping notices. One bot that flags new quote requests and another that tracks supplier shipments would take a real chunk of that off their plate. Each job is small enough that the person who used to do it can check the bot’s work. It’s the same shared-inbox problem I walk through in the email triage case study. If the pile is mostly customers asking the same questions, a FAQ and booking chatbot on your website can answer them before they reach the inbox.
The cautions matter more once customer data is involved. I’d start read-only, keep a person on anything that goes out under your name, and give the bots only the logins those jobs need, since they all share one computer.
I’m still working out how far I’d trust these with the sending side. Until then, the better question is which chore to hand off first.