Grok Bot Bengaluru · 19 Nov 2026
Grok Bot 101 & Pro Tips
From someone who runs work and life on Grok Bot
Mohammed Sanjeed
Grok Bot Ambassador Lead, South Asia
SpaceXSI Community

Grok Bot Bengaluru
About me
- Mohammed Sanjeed. AI consultant.
- Building with LLMs since 2023.
- IIT Madras, then engineer, PM, AI consultant.
- Worked with YC-backed startups to enterprises on agents, evals, GTM.
- Bot ambassador, regional lead, South Asia.
- About 45 Grok Bots run my work and life.
-
sanjeedm.com
X @sanjeed_i
LinkedIn

Grok Bot Bengaluru
Agenda
- Grok Bot 101
- Leveraging agents: the fundamentals
- Real use cases
- Pro tips
- Community demos, then AMA

Grok Bot 101
How we got here
ChatbotsAnswer questions. You do the work.
CopilotsHelp with a task. You drive every step.
A botAutomates a job, start to finish.
A team of botsRuns a whole function. You set direction and review.

Grok Bot 101
Grok Bot: an agent with its own computer
- It works on its own cloud computer: browser, terminal, files.
- It picks the model for each task, so you don't have to.
- You decide what needs your approval and what runs on its own.

Grok Bot 101
What every bot has
Memory
Its own. Your preferences, its rules, a log of its work.
Skills
A task that worked, saved to reuse.
Routines
Runs on a schedule, or when something happens.
Apps
Gmail, Drive, Slack and more. Or your own MCP.
A computer
Browser, terminal, files. Shared by all your bots.

Grok Bot 101
How Grok Bot is safe
Its own cloud computer
Each person gets their own VM. Only your bots share it.
No access by default
It reaches only the accounts you sign it into. It never sees your saved passwords.
Approval gates
Auto Review checks risky actions. Add Ask first rules for anything that must wait for you.
Admin control
For teams: org-wide on and off, network and connector policy, team rules.

Leveraging agents
How to make a good bot
Works with any agent. Here's how each step looks in Grok Bot.
1. Onboard
Context about you.
2. Connect
Hook up your apps.
3. Get it right
Do it together until it's good.
4. Routines
Runs while you sleep.
5. Grow a team
One bot, one job.
Step 1
Onboard it like a new hire
- Start with what's public: your site, LinkedIn, X.
- Then the private stuff, with permission: email, calendar, docs.
- Use voice. Talk for 10 minutes, like onboarding a new hire.
- Ask it to interview you, one question at a time.
- Before it starts, ask it to restate the job in its own words.
Step 2
Connect your apps
Official app
>
CLI + skill
>
MCP
>
API + skill
>
Browser
- Logins and checkouts stay in the browser. Passwords and cards go in Vault.
Step 3
Get it right before you automate it
- Do the task together a few times. That's your eval.
- Show it real examples, including the ones that went wrong.
- Teach it your judgment too: which source to trust, when to wait, when to ask you.
- Happy with it? "Save this as a skill."
Step 4
Put it on a loop
On a schedule. "Every weekday at 8 AM, plan my day." Hourly at most, or it burns usage.
When something happens. An email, a Slack message, a GitHub issue, a webhook.
Nightly review. A routine checks another bot's work and fixes its skills.

Step 5
Grow a team
Talk mostly to your primary bot. Split a job out when it has its own sources, schedule, or approvals.
Jarvischief of staff, my primary bot
Shoppingorders, prices
Groceriesweekly Zepto run
Traveltrips, rides
Researcherlook it up
Musicpractice, covers
Milaa side project
Studiocontent
dr eggbotbuilds my bots
+ 35 moreone job each
Real use cases
Some real life examples
How I got these
One voice note to my workshop bot
It messaged my other bots and asked each one for its routines, how it helps me, and the coolest things it has done.

Real use cases
1 / 11
Jarvis
My chief of staff. I talk to Jarvis; it gets the right bot on the job.
Routines
- 8:00 AMX bookmarks: each one routed to a bot
- 8:40 AMPlan-day: checks Dayflow and my sleep, drafts my day
- 11:00 PMReflect-day: what got done, what stalled
Highlight
Reads Dayflow, the timeline of what I actually did on my Mac, and plans my day around it.

Real use cases
2 / 11
Inbox Zero
Reads my email so my day doesn't start in it.
Why it's useful
- Sorts my inbox every weekday morning: do today, drafts, FYI.
- Replies come drafted. Nothing sends without my click.
- Today's list goes to Jarvis for my day plan.

Real use cases
3 / 11
WhatsApp
Reads every unread chat and hands me only what needs me, replies drafted.
Why it's useful
- It can't send anything.
- Gives me only what needs me: to-dos and replies I owe.
- Hands the action list to Jarvis.
Highlight
A few hundred unread messages become a short to-do list, replies drafted in each person's tone, even Manglish.

Real use cases
4 / 11
Mila
A PM bot for a client product. Checks usage, app logs and AI agent logs, and launches cloud agents to fix what breaks.
Routines
- 8:00 AMUsage and log check: cloud agents fix breaks
- 8:30 AMConcise update to send the client
- Weekdays 7 PMReviews open PRs
- On alertEvery new Slack alert

Real use cases
5 / 11
SpaceXSI Community
Runs logistics for meetups like tonight, so I can spend my time on people.
Skills
- Brand assetsLogos, fonts and colors for every event
- Inbox sweepEmail and WhatsApp, one list
- Food splitVeg and non-veg counts for the venue
Highlight
Ran tonight's event inbox: about 30 guest emails turned into 12 reply drafts before lunch.

Real use cases
6 / 11
Credit Cards
Knows every card's caps and fine print, so I never think about which one to swipe.
Why it's useful
- A weekly spend recap flags when I'm close to a cap.
- Watches offers so they don't expire unused.
Highlight
Reads every statement line by line, and caught bonus points the bank missed.

Real use cases
7 / 11
Groceries
Compares live prices across delivery apps and drafts the cheapest cart. Never orders without my yes.
Why it's useful
- A Sunday fridge check puts what's low on the list.
- Logs prices, so it knows a real deal.
Highlight
Compares four delivery apps at my address, with coupons and card offers, and drafts the cheapest cart.

Real use cases
8 / 11
Finances Sanjeed
Keeps a ledger of my statements and shows me where the money goes.
Routines
- 10:00 PMDaily check-in: I add notes to the day's transactions
- Mon, Thu 11 AMSpend analysis
- 1st, 9 AMMonthly statement ledger

Real use cases
9 / 11
Rotten Poteto
Picks what I watch, read and play. Then sets a stop time, with a bet on it.
Why it's useful
- Picks from my tastes, saved in my vault.
- Watches game sales and pings me only on a discount.
Highlight
One queue across books, shows and games, and it never suggests something I've already seen.

Real use cases
10 / 11
Researcher
Only for deep research.
Tools
- Web searchThe open web, with sources
- Reddit, YouTube, TikTokWhat real users say
- X searchLive takes from people building
- last30days skillOnly what people said in the last 30 days
- Research trackerA Page with every open question
- Obsidian vaultWhat I already know

Real use cases
11 / 11
Phone
Makes AI phone calls for me: checks stock at a store, books a reservation.
Why it's useful
- Built-in calls are US only, so mine runs on Vapi.
- I approve the script before every call.
- Other bots hand it calls when they need to talk to a business.
Highlight
One approved script, six store calls for stock, and a summary of each call.
Pro tips
Pro tips from me and the team
Pro tip 1
Import your memory first
- Bring over what your other AIs already know about you.
- Your primary bot starts on day one already knowing you.
Pro tip 2
Add your work account too
- Add your work account, then switch between personal and work.
Pro tip 3
Let it use your own computer
- It can open files and run tasks on your Mac. Auto Review still checks first.
- The default is Ask every time. Start there.
- Route its traffic through your computer. Handy for sites that block cloud servers.
Pro tip 4
Start with the goal, not the process
- Filling a form or sending an email is a step, not the goal.
- Name the end result: a refund, a lower bill, a build task done.
- Think big: save money, find clients, run a whole workflow end to end.
Pro tip 5
Set the guardrails you need
- Custom Rules: "Ask first" or "Allow automatically", per action.
- All your bots share one computer, sign-ins included.
Pro tip 6
Turn clicks into scripts
- Same lookup in the browser every day? "Read the site's network calls. Write me a script."
- Ask: "Which parts of this could be scripts?"
- Keep them in a private GitHub repo. Cheaper than the bot clicking.
Pro tip 7
Push back. Ask what it checked.
- Ask for sources, and for what it didn't verify.
- When it gets something wrong, ask it to fix its own skill.
Pro tip 8
Bring your bot to meetings
- It gathers its own context, so it gets smarter.
- Ask it questions mid-meeting, or have it analyse something and get back to you.
- Can't make it? It attends and takes notes for you.
- On Team, Enterprise and Ultra plans.
Pro tip 9
Reply to one part of a message
- Select any part of its reply and press Cmd L to add it to chat.
- Then reply to exactly that line. Not needed, but it helps.
Pro tip 10
Use Pages
- Want to share something? Ask it to make a Page.
- Jarvis updates my daily dashboard every morning. I keep it open all day.
- Shopping keeps a Page of active items: what it researched, what's next.

Pro tip 11
Don't start blank
I highly recommend dr eggbot. There are many more.

Grok Bot Bengaluru
Links worth saving

Grok Bot Bengaluru
Follow for Grok Bot and AI tips

Grok Bot Bengaluru
Come to the next one
Follow the Bangalore calendar on Luma to hear about future events.