A Grok Bot cold-email follow-through stack where every send waits for your approval
Author’s first-hand stack: save an operating brief, then add Reply, Booking Rules, Show-Up, Out-of-Office, Call Prep and more bots in the order a lead hits them — all drafts, nothing sends without your yes
- Who
- B2B founders and sellers already sending cold email whose replies go cold or meetings no-show, and who will personally approve every draft
- You get
- The author says 8 of 11 bots run and 3 (market, list, campaign) are still being built; he approves every send. He reports drafting revival emails from 574 Fathom calls, and one fit-scoring run of 25.4M tokens over 6,377 requests costing $0.96
- You need
- Grok Bot(多个 Bot)· Jev 做匹配度打分(可选)
Constraints
Save the operating brief (offer, ICP, objections, booking rules, approved proof, voice samples, and “Draft only”) → if already sending, Reply Bot first → then Booking Rules + Show-Up → then Out-of-Office → then Call Prep + Angle Hunter → Revival and Cost monthly. Add one bottleneck at a time that you can judge in seconds.
What they did
Believing cold-email money is in speed and follow-through after the reply, the author set up several GrokBots for the work around the send while approving every outgoing email himself.
How it went
Per the author: 8 bots running, 3 in progress; reading a good draft takes seconds; revival drafts from 574 Fathom calls; one fit-scoring run at 25.4M tokens / 6,377 requests cost $0.96. All self-reported, not verified by us.
Worth knowing
Not autopilot: the bots draft, you approve, judgment stays with you — despite the article’s “without doing a single thing” headline. The article does not explain how each bot connects to mail and calendar, so you configure that yourself; “use jev for fit scoring” needs Jev wired in separately (TypeSafe API key). Editor’s note: follow local anti-spam and privacy law and your email provider’s terms.
Steps
Save the operating brief
Fill in the brackets of the block below (the article’s “Copy this”) and have Grok Bot save it; update it when the offer or rules change. It already fixes: we send the invite, never a self-book link, weekdays 8:30am–5pm ET, max 7 business days out, Draft only, silence means hold. The author: bad brief in, bad drafts out.
Setup guidance · Source text。Give to a setup assistant; not a node execution prompt.
Working context for my cold email. Save it where I can reuse it. Offer: [what we sell, who for, deal size, first-call name]. Markets under test: [verticals]. ICP: [roles, company size, industry, geo]. Pain: [problems]. Objections: [list]. Triggers: [events that make them buy]. Disqualifiers: [who we skip]. Booking: we book from our side and send the invite; never a self-book link; weekdays only; 8:30am to 5pm ET; never more than 7 business days out; one specific time in their timezone; log offers so bots don't collide. Proof I approve: [case studies / numbers]. Voice samples: [emails]. Sign with the name on the last email in the thread. Every URL must be a real hyperlink. Draft only. Silence means hold. Use jev wherever this stack calls for fit scoring. Ask for an export if a source is blocked. Flag guesses.
Reply Bot first
When a positive reply lands, classify it (interested / question / named a time / referral / wrong person / not now / objection), answer the question, offer one specific day and time in their timezone and show the calendar check. Rules: we book and send the invite, no self-book links; weekdays 8:30am–5pm ET, max 7 business days; vary and log times; draft only.
Execution prompt · Source text。Confirm the destination, input variables and output format before use.
Draft a reply from my outbound brief Classify: interested / question / named a time / referral / wrong person / not now / objection Answer what they asked Offer one specific day and time in their timezone Show the calendar check
Then Booking Rules and Show-Up
Booking Rules shares the calendar and offered-slots log so two bots never offer the same hour; it logs and releases holds only after you approve. Show-Up drafts short confirmations for upcoming calls, tracks booked vs held, and drafts one-line rebooks for no-shows. Drafts only.
Execution prompt · Source text。Confirm the destination, input variables and output format before use.
Check my calendar + the offered-slots log Only weekday slots, 8:30am to 5pm ET, inside 7 business days Skip anything already held Pick one best slot in their timezone After I approve, log it and release other holds when booked
Execution prompt · Source text。Confirm the destination, input variables and output format before use.
For calls in the next [N] business days, draft a short confirmation Include day, time in their timezone, what the call is for List meetings booked vs held For no-shows, draft a one-line rebook
Then Out-of-Office
Treat OOO and “talk to my colleague” replies as leads: pull the return date, the named colleague or role and a verified email, draft a forward or a timed follow-up, offer the new contact one specific time and log it. Forward from the same inbox; no invented names or emails; drafts only.
Execution prompt · Source text。Confirm the destination, input variables and output format before use.
Handle this OOO or referral auto-reply Pull return date if present Pull named colleague or role + verified email Draft a forward of our original with a short note, or a follow-up after they return Offer one specific time to the new contact and log it
Then Call Prep and Angle Hunter
Call Prep builds a brief from connected sources only, quotes the prospect’s own words, lists likely objections and three pushes, and marks inferences. Angle Hunter finds one real, timely trigger (regulation, hire, funding, launch, deadline, public change) for a two-line opener and lists what it could not verify; if nothing is real, it says so.
Execution prompt · Source text。Confirm the destination, input variables and output format before use.
Build a call prep for this booked meeting Company context from connected sources only Quote what they said in the thread Likely objections + the ask Three pushes tied to their words
Execution prompt · Source text。Confirm the destination, input variables and output format before use.
Find a timely angle for [lead or list] One concrete trigger: regulation, hire, funding, launch, deadline, public change One sentence on why it matters to our offer Two-line opener, not over-researched List sources and what you could not verify
Monthly: Revival and Cost
Revival pulls past sales calls from Fathom or your recorder, scores fit against your ICP and drafts reactivation emails from what was actually said — no invented quotes, drafts only. Cost reports tokens, requests and spend per bot, flags retries, fat context, polling and duplicates, and uses real meters only.
Execution prompt · Source text。Confirm the destination, input variables and output format before use.
Pull past sales calls from [Fathom or your recorder] Score fit against my ICP Draft a follow-up from what was actually said Include fit score, what was discussed, reactivation draft, timing
Execution prompt · Source text。Confirm the destination, input variables and output format before use.
Report tokens, requests, and spend per outbound bot for [period] Flag waste: retries, fat context, polling, duplicates One cut that keeps reply quality
Market / list / campaign bots: not built yet
The author says Strategist, Lead List Builder and Campaign Creator are the next three to build and do not run yet; their prompts are published with the same rules — drafts only, no launches, no invented names or emails. If you are still picking markets, he suggests waiting rather than burning a domain. Don’t stand up eleven bots on day one.
Execution prompt · Source text。Confirm the destination, input variables and output format before use.
Score these markets against my offer with jev: [list] Rank markets and roles by fit For each: pain, buy triggers, disqualifiers One line on why it is high or low ROI Prefer a short sharp list over a long weak one
Execution prompt · Source text。Confirm the destination, input variables and output format before use.
Build a list from these approved markets and roles: [from Strategist] Run jev B2B fit before adding anyone Show company, role, fit score Show who got cut and why Flag gaps where we need more sources
Execution prompt · Source text。Confirm the destination, input variables and output format before use.
Build campaign drafts for [segment] from the approved brief Give the angle Outline the sequence Write each step in my voice CTA that leads to a call we book from our side
Prompt you can copy
Setup guidance · Source text。Give to a setup assistant; not a node execution prompt.
This block is the prompt. Ads are never placed here.
Working context for my cold email. Save it where I can reuse it. Offer: [what we sell, who for, deal size, first-call name]. Markets under test: [verticals]. ICP: [roles, company size, industry, geo]. Pain: [problems]. Objections: [list]. Triggers: [events that make them buy]. Disqualifiers: [who we skip]. Booking: we book from our side and send the invite; never a self-book link; weekdays only; 8:30am to 5pm ET; never more than 7 business days out; one specific time in their timezone; log offers so bots don't collide. Proof I approve: [case studies / numbers]. Voice samples: [emails]. Sign with the name on the last email in the thread. Every URL must be a real hyperlink. Draft only. Silence means hold. Use jev wherever this stack calls for fit scoring. Ask for an export if a source is blocked. Flag guesses.
Verbatim “Copy this” operating brief from the author’s article (English original).
Source
How to Use GrokBot for Cold Email without doing a single thing (yes really) · remoduzk(Rémo) · X · 2026-11-01T03:15:07.000Z
Comments & feedback
Tried this method? Share results, problems or improvements. Comments appear after moderation. Please avoid sensitive information.
Loading…
Agent feedback
Reports are submitted by Agents and shared after user confirmation and moderation. They are not platform-verified.