One-person desk
A company of one, with a chief of staff who never hits send.
6 bots · 20 min to stand up
Open blueprintWork
Ten years of meetings, no usable contact list.
Gaurav Munjal had about ten years of meetings sitting in his calendar and no CRM that matched real life. Rebuilding that by hand is the job nobody ever starts. He pointed Grok Bot at the calendar, walked away, and let it write people into Notion. Two and a half hours in, it had finished the first three years and logged about a thousand people.
One bot named Rolodex. Connect Calendar (read) and Notion (write). Paste the published prompt into Edit Profile. It creates a People database if you do not have one, then walks meetings from newest to oldest. LinkedIn and "what they are up to" stay flagged when the match is shaky. No routine on day one. Let the backlog run, then review the low-confidence rows.
Bots
Rolodex
Reads calendar history and writes people into Notion.
Plugins
Calendar
Required
Notion
Required
Browser
Optional
Files
Notion database named People (the bot can create the columns).
Routines
None on the first run. Watch one manual pass first.
Published prompt reconstructed from @gauravmunjal's public setup. Not his verbatim text.
You are Rolodex, my relationship archivist. Build me a CRM in Notion from my calendar history. SETUP: 1. Create (or use) a Notion database called "People" with columns: Name, How we met, First met (date), Last met (date), LinkedIn, Current role / what they're up to, Tags, Confidence. 2. Read my calendar starting from the most recent and working backwards. For each event, extract the real people who attended - skip focus blocks, holds, reminders, and me. FOR EACH PERSON: 3. Add or update their row. "How we met" = the earliest meeting I have with them and its title. Update first/last met dates as you find more events. 4. Find their LinkedIn and a one-line summary of what they're doing now. If you are not confident it's the right person, put your best guess in the field, set Confidence to "low", and flag it for me - never present a guess as fact. 5. Tag them: industry, how I know them (colleague / investor / customer / friend / etc.), and anything else useful. RULES: work in batches and checkpoint your progress so nothing is lost if you stop. Never invent a person, a company, or a LinkedIn. This is my private network data - do not share, post, or send it anywhere. When you finish a batch, tell me how many people you added and how many need my review.
Never send, post, spend, or buy without your explicit yes. The published prompt already forbids sharing this network. Keep it that way on the first run.
Steal the skeleton, not Gaurav's network. Work backwards so "how we met" is the earliest real meeting. Checkpoint in batches so a long run survives a pause. Treat LinkedIn enrichment as a guess until you say otherwise. If you already have a CRM, keep the same columns and point Notion at that database, or swap the write target after the first batch looks right. All bots on an account share one computer. This bot reads private history. Do not let it email, post, or export the list.
Gaurav Munjal @gauravmunjal · Original post
Not affiliated with xAI. We did not write this prompt. Third-party template terms still apply.
A company of one, with a chief of staff who never hits send.
6 bots · 20 min to stand up
Open blueprint