i created a Grok Bot workflow and earned (full guide)

Cover image

i'm going to show you one cash lane on Grok Bot: demand-first info products.

not "ask AI for a PDF."

the bot listens where people are stuck, scores the pain, builds a landing + checkout before you write anything, then you only ship the product if there's real intent.

one bot per job. approvals on money and publish. math below is illustrative so you can size the machine - not a promise.

what's inside

the loop in plain words
stack (Grok Bot, Whop, landing, tracker)
four lanes on the roster + what stays human
standing rules + copy-paste prompts
listen → score → pre-sell → write → distribute → fulfill
daily/weekly cadence
7-day sprint
failure modes + quality bar
example unit math
paste pack to start today

the mistake i keep seeing

people treat the agent like a notes app.

prompt → paste → 40-page PDF → nobody buys.

usually one of three things broke:

no demand proof (topic felt smart, nobody was stuck)
no distribution (asset with zero pipes)
slop (generic advice, no constraints, no artifacts)

this workflow is built to kill those early.

the loop

hard rule: if the landing doesn't show intent (clicks, checkout starts, "is this out yet?"), you do not write the product.

product type: one narrow info product (guide, pack, templates, checklist + scripts).

payment: one-time checkout (Whop or equivalent).

traffic: communities + X where the pain already shows up.

stack

required:

Grok Bot
Whop for checkout (CLI helps if you want the bot creating products/links)
a landing you can deploy fast (Framer, Carrd, custom, or bot-built static)
file delivery on Whop (PDF, zip, whatever you can support)
one tracker (sheet or tiny dashboard): candidates, status, spend, revenue

optional:

Apify-style scrapes for public data (cap the spend at the provider)
email capture later - v1 can go straight to checkout

print/mail vendors are for other cash lanes. you don't need them for info products.

keep burn small on purpose:

hard caps on tools
no ads until organic winners exist
one product book - don't mix spend across experiments

example sizing only:

one bot, one job

don't build a god-agent that also refunds people and tweets as you.

listener - daily pulls stuck posts. saves exact wording. no cleanup.

scorer / board - running board; promotes a topic to candidate only when it repeats and scores.

landing builder - pre-sell page + checkout link. no fake testimonials. no invented proof.

writer - only after intent. narrow product from the board + your constraints.

you still do:

pick lanes you can actually teach
approve candidates
post and reply from your accounts (auto-post gets banned)
refunds, price changes, payouts
final quality gate

one agent, one outcome. handoffs are fine. role titles that mean nothing are not.

standing law (paste once)

You run ONE business: demand-first info products.

Hard rules:
- Public sources only unless I give credentials.
- Never invent testimonials, revenue, user counts, or screenshots.
- Never refund, change prices, move money, or raise spend caps.
- If a spend cap is hit, stop and tell me.
- Social posting/replies from my accounts = draft only. I publish.
- Exact buyer language is sacred. Do not "professionalize" pain quotes.
- A topic becomes a candidate only if it repeats across the week AND
  I could plausibly teach it. If I can't, say so and skip.
- Build landing + checkout BEFORE writing the product.
- If landing intent is weak, kill the topic. Do not write anyway
  to justify sunk cost.
- Keep this book's files, products, and metrics separate from any other project.
- For the first 3 ships, stop before publish and show me everything.

when rules conflict, approval wins.

step 1 - pick a lane you can sell without cosplay

good:

a job people already pay to fix
recurring confusion (tools, workflows, distribution, shipping agents)
buyers who care about time and money

bad:

pure hype with no steps
medical / legal / finance promises you won't stand behind
topics where you have zero judgment

one sentence:

> > I help [who] get [outcome] without [common waste], using [concrete artifacts].

if you can't fill that, stop. the bot can't invent authority for you.

step 2 - demand surfaces

pick 3-7 places. depth beats tourism.

types:

X search / lists around a tool or job-to-be-done
forums where people post failures
discords where beginners get stuck (listen-only is fine at first)

for each surface write down:

name / url
what "stuck" looks like
posting norms so you don't spam and die

step 3 - daily listen

Every day, scan the last 24h in these surfaces: [LIST].

Pull posts/comments where someone is stuck on something SPECIFIC
and losing money or time.

For each item capture:
- surface + link
- date
- exact quote (no rewriting)
- what they already tried (if said)
- whether they mention paying for something that failed
- rough urgency (low/med/high) with evidence from wording

Output a table. No strategy essay.
Drop jokes, news, and vague motivation posts

step 4 - weekly scoreboard

Update the running board from this week's captures.

Group near-identical pains into topics.

A topic becomes CANDIDATE only if:
1) it appeared multiple days this week (not one viral rant)
2) wording shows money/time loss
3) a digital info product can solve a first win in <7 days of work
4) I can teach it (flag if doubtful)

Score candidates 0-10 on:
- repeat frequency
- urgency
- willingness to pay signals
- narrowness (can we title it in one line?)
- competition spam level (high spam = need sharper angle)

For top 3 candidates give:
- topic name
- buyer language headline options (3)
- who it's for / not for
- product shape (checklist, templates, short guide, pack)
- suggested price band (one-time)
- why it might still fail

kill when:

one loud post, no repeats
entertainment argument only
needs credentials you don't have
only works as ongoing private consulting
you wouldn't put your name on it

step 5 - landing before the product

For candidate #[N], build a landing page. Product body does NOT exist yet.

Requirements:
- H1 = problem in their words
- 5-8 bullets of what's inside (specific artifacts, not vibes)
- who it's for
- who it's NOT for
- plain-English what happens after purchase
- checkout section with THIS link: [WHOP_CHECKOUT_URL]
- zero testimonials
- zero "join thousands"
- zero fake logos
- mobile-first, fast

Deploy and give me the live URL.
Also give me 10 distribution openers that point to this URL
without sounding like a bot.

intent you actually track:

checkout starts
"take my money" replies
email asks
repeated clicks from in-market threads

no intent after a real distribution push → kill.

cheap lesson. expensive lesson is writing 12k words first.

step 6 - paid shell on Whop

manual or bot + CLI:

name = outcome + audience constraint
one-time plan
checkout link
delivery empty until write is done
internal SKU tied to candidate id

Install/auth Whop CLI if needed.
Create one product + one-time plan for [NAME] at $[PRICE].
Create checkout link.
Return ids + url as bullets.
Do not touch other products.

step 7 - write like an operator

only after intent.

Write the v1 product for [CANDIDATE].

Constraints:
- promise ONLY what the landing promised
- length: short enough to finish in one sitting (aim lean)
- structure every section as: goal → steps → template/example → failure mode
- include copy-paste prompts/checklists where useful
- no generic "mindset" chapters
- no invented case studies
- mark assumptions clearly

Deliver:
1) final manuscript outline
2) full draft
3) 1-page quickstart
4) filename + version

human gate: read it once out loud. cut anything you couldn't defend on a call.

step 8 - distribution (this is the money part)

most AI product posts skip this. don't.

step 9 - when someone pays

When a payment hits:
- deliver file immediately
- send a short confirmation: what they got, how to start in 10 minutes, how to reply for support
- log: date, SKU, source pipe if known, issues
- if request is outside scope, answer with boundary + what v2 could include
- never promise refunds unless I say so

honest "not for you" on the landing keeps support sane.

step 10 - weekly read-out

Weekly review for the info-product book:

- candidates created / killed / shipped
- landing unique visitors (if available)
- checkout starts / sales / refunds
- top source pipe
- $ spend (tools) vs $ gross
- one thing to change next week
- which product gets another distribution week vs sunsetting

double down on winners. more distribution beats more PDFs.