How I use Grok Bot for AI Workflows

Short answer: I use Grok Bot two ways. On stage, a group of role bots shows a team of AI agents working together live. For my own work, I started with a team of bots and cut it to one bot that writes one daily briefing. One agent with one clear output beat a crowd.

Grok Bot is xAI's app for persistent AI teammates, launched in August 2026. Bots run on a cloud computer they share, can sit in a group chat together and hand work to each other. Learn more about Grok Bot. My setup post from October explains the rules I set first; this post is what happened next.

On stage: a team of bots you can watch

In a talk I put several role bots in one group chat and give the group one task. The audience watches them hand work to each other.
Example 1, for factory owners: a finance manager, a purchasing manager, a quality manager and a sales manager.
Example 2, for a sales playbook: a sales director, a customer, a new hire and a reviewer.
Why it works on stage: people understand AI agents faster when they watch them disagree and pass work along, than when they hear a definition.
The full four-bot meeting with each role's brief is in my setup post.

At my desk: I started with a crowd

15 October: I set up a support office of bots: one coach bot that talked to me, and several desks for email, calendar, news, workshop prep, content and event photos.
Each bot was useful on its own. Together they made more messages for me to read, not fewer.
The coach sent me questions with buttons. I did not answer one of them.
17 October: I trimmed the team. 22 October: I paused every bot except one.
My own words: there was so much noise. For my own desk, I am not a fan of multi-agents.

The one bot: one briefing a day

It works on the cloud computer while my laptop is closed. It reads my mail, my calendar and my files.
Every morning it writes one document. A short evening update follows.
The document always has the same four sections, in this order:
1. Needs me: at most three items, each with a recommendation and options A, B or C.
2. For my assistant: draft text, reminders and follow-ups. My main AI assistant on my Mac picks these up.
3. FYI.
4. Closed since yesterday.
Every morning it also looks for money items: bills, renewals and unpaid invoices.
It messages my phone only if a deadline is less than 24 hours away or money is at risk, and at most once a day.

What the bot is not allowed to do

It never sends anything.
It never writes to my calendar.
It never deletes or moves mail.
It never pays.
Its briefing is a lead, not a fact. My assistant checks the source before acting on any item.

Did it work?

The one-bot trial ran for a week and ended on 29 October. I kept it.
Every deadline reached me in time.
Not perfect yet: the target was one briefing a day, and I still get a morning and an evening document. I am tightening that next.

The lesson: one agent, one output

Many agents are good for showing how agents work together. For your own day, start with one bot, one document and one fixed format.
Add a second bot only when the first one is reliable and boring.
How to copy it, four steps:
1. Pick the one document you want each morning.
2. Fix its sections and their order.
3. Write the list of things the bot must never do.
4. Decide who checks the bot's report before anything happens.
For the decisions that need debate, I use a different setup: a crew of AI assistants that only speaks when I open it. That is in my companion post.

FAQ

Are multi-agent setups a bad idea?

No. They are a good way to show and test how agents work together. For a daily routine, one bot with one output is easier to trust and easier to read.

What should the one bot write?

One document with fixed sections, starting with the few things that need you, each with a recommendation.

Does the bot send emails for me?

No. It only reads and writes its briefing. Sending stays with me.

Sam Wong (黃力桐) is a Hong Kong AI adoption trainer who has trained 11,000+ people.