Run ads to the offer that already sells without them
The offer already has receipts. The page never had a VSL. Write the story from real wins, put a video on every page, buy the click, keep the numbers in one desk.
Fit Offer people already buy (referrals, organic, GHL receipts). Will put a VSL on the page. Will approve ads.
Unfit No offer. Wants a startup idea or to clone someone else's product. Text-only checkout. Wants you to rip a SaaS one to one. Scaling spend at 10% net without looking. Furniture / dropship idea hunting.
The map
Does
If the offer already sells without ads, write the proof, put a VSL on a saucy page, buy the click, and keep GHL as the desk. Not a cold-vs-warm fork. Not a new product.
IF ROAS target is lower than you need → raise the ceiling
IF net < the number you named → do not raise budget
IF a number has no GHL/ad source → write Unknown, do not scale
IF CRM is not GHL → use GHL
IF they want to clone someone else's product → kill
One opened-up account. Illustration, not a promise. ~$25–30K/mo. ~$387/day ads. ~32 opt-ins. ~25 qualified. 15% of those self-checkout. 85% abandon. Sales hits the 85%. Industry opt-in ~15%. Their days 23%, 15%, 22%. Last 30 days ~13%. Current net sat 10–15%. $100K at 10% is a job. Goal shape they named: $100K at 30% net net. Use your GHL numbers, or write Unknown.
The journey
Start here
Proof card
your receipts
Live01
Write the proof card from YOUR receipts
Last 90 days of sales. How they found you. Did ads ever run? If you are picking which of your offers to push: proven, still skippable with paid. Spoken filters when people browse public revenue (inspiration only): $1K–$10K, then $10K–$20K, restated $0–$20K because those operators ground organic for years. A public verified-revenue directory is an optional later look for category inspiration. It is not a product to copy. Template: skills/paid-proof/proof-card.md.
Bot fills the card from GHL / invoicesYou the facts only you know
Receipts
GHL, invoices, or Stripe. If unverified, write Unknown. Do not scale on a vibe.
Kill
No receipts → stop or pick a different offer you already sell. Startup idea, furniture / dropship, or someone else's product one to one → stop.
Repeated work
not 10 SaaS ideas
Research
02
Look for repeated work. Do not ask for ten ideas.
The worst first prompt is “Give me 10 profitable AI SaaS ideas.” You get a polished list with almost no proof anyone cares. Point the research at the work their buyers already do — or at which of their offers to push. If they want a startup idea, stop. Template: skills/paid-proof/research-prompt.md.
Bot runs the research promptYou name the offer it is allowed to look at
Find
People already monitor several sources by hand, copy into docs or a sheet, ask “what changed?”, then decide. They could pay $30–$100/mo to stop.
Return
user · repeated pain · current workaround · frequency · public evidence · existing alternatives · smallest useful product · realistic monthly price
Rank
pain × frequency × willingness to pay × simplicity of MVP.
Worst first prompt on the left. Repeated work on the right. Eight fields. Rank the product of four.
03
The smallest product answers three questions
The useful answer is not another list of links. It is: What changed? Why does it matter? What should I do next? First version does not need teams, ten dashboards, or 40 settings. It needs to beat their current workaround. The VSL and the page say those three lines.
Bot writes the three answers from the proof cardYou nothing
GHL scoreboard
100 inbound
GHL
04
GHL is the sales machine
Inbound comes in. Sales need pipelines and dispositions. They named Go High Level. Never another CRM. Of 100 inbound: how many got touched, how many conversations lasted at least 30 seconds, how many closed, how many did not. You need someone system / ops / tech oriented, or the File does this on paper first. Dirty data = do not scale. Clean GHL plus ad metadata before any forecast. Field map: skills/paid-proof/ghl-field-map.md.
Bot writes the scoreboard from GHLYou GHL login if it is not connected
Of 100
Touched. Conversations ≥30 seconds. Closed. Not. If a line has no source, write Unknown.
Liaison
Sales people do not know the tech. Tech people do not know sales. Someone has to speak both, or you run the File on paper until they do.
Math
05
Write the funnel from GHL. Label the opened-up account.
The numbers below are one opened-up account. Illustration, not a promise. Replace every line with your GHL and ad sources, or write Unknown.
Bot copies your numbers, never inventsYou open GHL and the ad account
Opened-up
~$25–30K/mo. ~$387/day ads. ~32 opt-ins. ~25 qualified. 15% of qualified self-checkout. 85% abandon. Sales hits the 85%.
Opt-in
Industry ~15%. Their days: 23%, 15%, 22%. Last 30 days ~13%. Nothing crazy. Levers: raise opt-in a little, raise self-checkout, then sales on abandon.
Net
Goal shape they named: $100K/mo at 30% net net. Current net sat 10–15%. Quiet part: $100K at 10% is a job. Prefer $50K/mo and net $25K over $100K at 10%. Name net before you raise budget.
ONE OPENED-UP ACCOUNT. Not a promise. Use your GHL numbers or write Unknown.
Ceiling06
Set the ROAS ceiling you actually want
Whatever number you give the media buyer is the ceiling they hit. They said 1.5, then immediately changed to 2. If you need 2, do not say 1.5. Name net before you raise budget. If net is under the number you named, do not raise budget.
Bot writes the ceiling and the net on the logYou name both numbers
Story script
root cause, not symptoms
Story
07
Write the story from a real day
Everybody is a copywriter if they know the beats. Write from a real painful event, then do the same from a client win. Brain is old software. Stories beat features. Template: skills/paid-proof/story-script.md.
Bot drafts both scriptsYou the Tuesday only you were at
Beats
1 What happened (a specific day). 2 What really happened. 3 Pain after. 4 Common solutions. 5 Why they failed (symptoms). 6 Pivotal moment. 7 Why it worked (root cause). 8 Life after — I can X again.
Shop
House example: truck died on a Tuesday, season died, they almost closed. Rental, the part, more Facebook — symptoms. A summer customer paid a deposit before the quote. Root cause: strangers never heard the Tuesday. Do not use a medical plot as the house copy.
Cheat
Pain-community success posts (joint pain, acne, arthritis, entrepreneurship). Everyday people. Not travel groups. That post is a finished VSL. Template the structure. Do not paste medical claims.
Same spine for the page, the VSL, and the ad. Root cause, not symptoms.
VSL
every landing page
Video
08
Put a VSL on every landing page
Text-only self-checkout on a small service is wrong. Even an AI VSL that autoplays converts better. They do not know why. Infomercials still work — that is the lesson, not a product to sell. 60 seconds is enough to start. Test 10+ minutes, even an hour, because trust is hours of content. Organic is powerful because people already watched tens or hundreds of hours. Paid has to put that trust in the video. Brief: skills/paid-proof/vsl-brief.md.
Bot writes the brief from the story scriptYou record, or approve the AI cut
Page
09
One saucy page. Not a brochure.
One offer. Video on the page. GHL form. Abandoned-cart SMS in GHL (email-only abandon is the miss they just started fixing). Do not invent a twelve-section wireframe. Do not send traffic to text-only checkout.
Bot writes the page from the briefYou login / 2FA / pay for the host
Video. GHL form. They stay in GHL. Not a brochure.
Meta Ad Library
steal the structure
Ads
10
Steal a working ad structure. Swap your offer.
Do not invent from a blank canvas. Meta Ad Library is the house tool. Screenshot a working in-category ad. Prompt AI with constraints, not “I’m hungry, cook me something.” Study this image. Keep the composition. Swap MY offer. Never generate an ad image from scratch. Template: skills/paid-proof/ad-swap.md.
Bot drafts the swap from the screenshotYou pick the reference ad and approve
$50
11
Run a $50 test in draft. Then a human approves.
3 ad angles. 3 hooks. 3 copy variations. One Meta sales campaign. Use the live checkout / GHL form. Keep everything in draft. Show creatives, audience, budget, and campaign ID before launch. RESEARCH → CREATE → DRAFT → APPROVE → LAUNCH. Spend still needs approval. Pause a losing group only with their OK.
Bot prepares the draft testYou approve, or do not
Illustration
Outcome angle “Know what changed before Monday’s meeting” CAC $21. Tech angle “AI-powered competitive intelligence” CAC $63. Learning: sell the outcome, not the technology. Not a promise.
Loop
TEST → RESULT → LEARNING → NEXT TEST. Save the learning. The next test does not start from a blank prompt.
Illustration from one $50 test write-up. Sell the outcome. Draft, then approve.
Lock12
Ask where the number came from. Open GHL.
Do not assume the bot hooked GHL. After every write: where did you get that number? Walk the chain. Open GHL and verify. Hallucination is real. Revenue data needs 100% accuracy. Do not copy-paste numbers between people and AIs. That is the broken desk this File kills. If a number has no GHL or ad source, write Unknown. Do not scale on it.
Bot shows the source chainYou click it in GHL
Checkout
13
Localhost is not a business
Most build tutorials stop when the page loads. The chain is: product → price → checkout → customer → revenue → ads → data. Every write ends with a read. Verify product, price, billing, and checkout before you share the URL. A command finishing is not proof the business state is correct. GHL stays the CRM. Whop is checkout / membership when the offer is digital. Do not replace GHL.
Bot writes, then reads backYou login / 2FA / pay
Every write ends with a read. product ✓ price ✓ billing ✓ checkout ✓ before the URL goes out.
14
The first 10 customers are a learning problem
Not scale. Do not automate yet. Find people publicly talking about the pain. Generate the deliverable before asking them to buy. Learn: do they read it, return, pay, keep paying? Once that happens the business starts generating its own context. Then ask a better question: what is the biggest constraint this week?
Bot finds the public pain and drafts the deliverableYou send, or do not
15
Sale → balance → ads → sale
The live checkout feeds the next test. Prepare a small acquisition test. Draft. Human approves. After enough data, find the winning angle and save the learning. That is the loop. Not “run ads.” Pause a bad ad group only with their OK.
Bot prepares the next test from last week’s learningYou approve spend
Roadmap
16
The job changes at every stage
This is not a new-product hunt. It is how the desk behaves once the offer already sells. If they are still at “give me an idea,” they are unfit.
Bot names the current stage from the logYou nothing
$0
Validate by hand. Talk. Find the value moment.
$500
Fix retention. Why they stay. Why they cancel.
$1K
Start paid tests. One audience. One channel. Bounded spend.
$2.5K
Save learnings, segments, rules. Stop starting from zero.
$5K
Automate analysis. Not the spend.
$10K
Stop asking “how do we grow?” Find the current constraint, prepare one experiment, verify the result.
The operating job changes. At $10K it is one constraint, one experiment, one verify.
17
Later: one desk. Grok Bot + terminal. Whop CLI if digital.
The broken roadmap: find an idea → landing page → payments → ad manager → analytics → paste the numbers back into AI. Fix: Grok Bot on your computer. You only do login, 2FA, or pay. GHL stays the CRM for inbound, dispositions, abandon SMS, sales. Whop is checkout / membership when the offer is digital. Do not replace GHL with Whop. Client never types API keys into chat.
Bot runs the desk after the File already works by handYou login, 2FA, or pay
Install
curl -fsSL https://whop.com/install.sh | sh then whop login (browser or API key — you do that, not chat). Then whop skills add. Then whop --llms.
Write
whop products create → product_id → whop plans create → plan_id → whop checkout-configurations create → live URL. Every write ends with a read.
An LLC flow exists on the same surface. They do not need an LLC to test an offer.
Monday
Compare this week vs last: revenue, new members, churn, ad spend, stuck ads. Agent runs stats + ads list. You do not paste dashboards. Catch refund / dispute spikes. Pause a bad ad group only with your OK. Template: skills/paid-proof/monday-brief.md.
Stop tab-hopping. Grok Bot holds the reasoning. GHL holds the CRM. Whop holds digital checkout when you need it.
Done
The File is closed
The proof card is filled from their receipts. A VSL is on the page. GHL has the form and abandon SMS. The ROAS ceiling and the net are named. A draft ad is waiting for a human. They can approve spend without another meeting.
Free · Operator SOP
Run the paid-proof File
If the offer already sells without ads and they will put a VSL on the page, do this. You only do login, 2FA, pay, or approve ads.
Fit gate
If they have receipts (referrals, organic, GHL) and will approve a VSL and ads, stay.
If there is no offer, or they want a startup idea, furniture / dropship, or someone else's product, stop.
If the page is text-only and they refuse a video, stop.
If they want to scale at 10% net without looking, stop.
Tools in order
Proof card (skills/paid-proof/proof-card.md)
Research prompt (skills/paid-proof/research-prompt.md)
GHL field map (skills/paid-proof/ghl-field-map.md)
No other CRM. No funding File. Agent drafts. Human approves spend.
Numbered run
Fill the proof card from THEIR last 90 days. How they found you. Ads ever? If no receipts, stop.
Bot fills the cardYou facts only you know
If they are picking which offer to push, run the repeated-work research. Do not ask for 10 SaaS ideas. Rank by pain × frequency × willingness × simplicity.
Bot runs the promptYou name the allowed offer
Write the smallest product as three answers: What changed? Why does it matter? What should I do next?
Bot writes the three linesYou nothing
Instrument GHL. Of 100 inbound: touched, ≥30 seconds, closed, not. Dirty data = do not scale.
Bot writes the scoreboardYou GHL login if needed
Write funnel math from GHL. Label any opened-up-account numbers as illustration, not a promise. Name net before budget.
Bot copies sourced numbersYou open GHL and ads
Set the ROAS ceiling you actually want. If you need 2, do not say 1.5.
Bot writes the ceilingYou name it
Write the story script from a real day, then a client win. Root cause. Shop example, not a medical plot.
Bot drafts bothYou the Tuesday
VSL on every page. 60s is enough to start. Autoplay / AI allowed.
Bot writes the briefYou record or approve the cut
Saucy page: one offer, video, GHL form, abandon SMS.
Bot writes the pageYou login / 2FA / pay
Ad swap from Meta Ad Library. $50 test stays in draft until they approve. Every write ends with a read.
Bot draftsYou approve spend
Ask where each number came from. Open GHL and verify the chain.
Bot shows the chainYou click GHL
File log columns
date · shop · offer · receipts · GHL scoreboard · VSL live · ROAS ceiling · net named · ad spend · note
Done rule
Stop when the proof card is filled from their receipts, a VSL is on the page, GHL has the form and abandon SMS, the ROAS ceiling and net are named, and they can approve a draft ad without another meeting.
Do not
Do not clone someone else's product.
Do not invent a GHL snapshot URL.
Do not name a CRM other than GHL.
Do not raise budget when net is under the number they named.
Do not scale on a number with no source.
Do not spend without approval.
Do not start from “give me 10 profitable AI SaaS ideas.”
Free · Run it in your AI
Prompt pack
Free means they use the AI they already have. Paid is the installer that walks it for them.
Ask AI
Paid
Installer builds it
We take the shop, enrich it, they upload, the agent connects what it needs and stands GHL up. They only tap login, 2FA, or pay. The installer runs the house skill.
