LF3 · ≈1945 spoken words · ≈13 min at 150 wpm

LF#3 · Full script · "How AI Books My Calls While I Sleep (5-Step Setup)"

Engine: BOOK (speed to lead + AI setter). Draft only, nothing posted. Package (titles, thumbnails, description, chapters, pinned comment): PACKAGE.md.

Runtime ≈ 12:00 spoken (1,800 words at 150 wpm), ≈ 13 min finished with screen holds. Setup A, landscape, award wall, same chair as the VSL. Film in 9 chunks (one per # beat), 2 s pause between chunks.

Rules baked in: evergreen (no dates, no weekdays), no prices, no phone tiers, one CTA (the free session, mid-roll + end), proof only from brand-assets/proof-library/PROOF-INDEX.md (clients by first name, "Individual result. Not typical." said aloud), "Nikhil Sai" or "Nik", "you" never "you guys", no "live" claim for the session (say "my team in the chat"). Humanizer pass done.

1. Hook (yt-script hookscore.py, 5 written, top 2 kept)

Score	Hook	Formula
⭐ 52	A lead filled out your form at 2 a.m. Nobody answered. By 9 a.m. they're on a call with your competitor. That's a setter problem, and I fixed it with an AI.	The Stake
41	Every lead you answer after five minutes is a lead you're paying for twice. Here's the AI setter that answers in sixty seconds and books the call.	The Reveal

Winner uses the cold-open scene the viewer recognises (the 2 a.m. form fill) and names the cost before the fix. Same scale LF2's winner scored (44), so the opening below puts the stakes in the first line and the proof right after.

2. First 30 seconds, word for word

[SCREEN: black. Phone lock screen. One notification: "New lead · 2:07 AM · Form: Strategy Call"]

A lead filled out your form at two in the morning. Nobody answered. By nine, they're on a call with the business that replied first. That's a setter problem, and I fixed it with an AI.

[SCREEN: cut to Nik, award wall behind. Lower third: "Nikhil Sai · AI implementation"] (PROOF: C01)

I'm Nikhil Sai. I've trained over a hundred thousand business owners to implement AI live, and the one thing I set up in every single business first is this. An AI that answers every lead in under a minute, qualifies them, and books the call on the calendar. Mine's called Cassie.

[SCREEN: GoHighLevel conversation view, Cassie's reply at 2:08 AM, lead name blurred]

By the end of this video you'll have the five-step build, the exact prompt I give her, and the one number to track so you know it's working. Step four is where most people's AI setter falls apart, so stay for that one.

(≈ 0:00–0:38, 150 words)

3. Full script with retention beats

Legend: [SCREEN: …] editor insert · [LOOP OPEN] / [LOOP CLOSE] open question planted and paid · [PATTERN BREAK] change of rhythm, prop or framing · [PROOF] indexed proof moment · [CTA].

0:00 · Cold open + hook + promise + plan (0:38)

(the 150 words above, verbatim)

0:38 · Why speed is the whole game (≈ 1:10)

[SCREEN: a clock graphic. "Lead arrives" at 0:00. Markers at 1 min, 5 min, 30 min, next day]

Understand this. A lead is at their hottest in the first few minutes after they fill out the form. They're on your page. They're thinking about the problem. Their phone is in their hand. Every minute you wait, that attention drains, and by the next afternoon they've forgotten which form they filled out.

[SCREEN: the old-way timeline: "Form 2:07 AM → setter sees it 10:30 AM → first call 2:15 PM → voicemail"]

That's the old way. The setter starts at ten. Sees the lead. Calls at two. Voicemail. Leaves a note, "tried calling", moves on. You paid for that lead. You paid the setter. And you got a voicemail.

[SCREEN: the new timeline: "Form 2:07 AM → text 2:08 AM → reply 2:11 AM → booked 2:14 AM"] [PATTERN BREAK: fast cuts]

Here's what Cassie did with the same lead. Text at two-oh-eight. They replied at two-eleven, because they were still awake and still on their phone. Booked at two-fourteen. I was asleep. She was closing.

[SCREEN: Morgan's testimonial card, "Individual result. Not typical." footer] [PROOF: C39 Morgan] [consent pending]

Morgan set up this exact piece and his appointments got automated, in his words. Individual result, not typical. The point is the lead didn't wait for a human to wake up.

Now, there are five steps to build this. Step one is the one everyone skips, and it's the reason their AI sounds like a robot. [LOOP OPEN: why AI setters sound like robots]

(≈ 190 words)

1:48 · Step 1: Write her job description before you write a prompt (≈ 1:30)

[SCREEN: card "STEP 1 · THE JOB DESCRIPTION"]

You'd never hire a setter without telling them what the job is. Same with the AI. Before any prompt, I write a one-page job description, and it answers four questions.

[SCREEN: four lines appear one at a time]

Who is she talking to? Not "leads". A coach who runs a group program and wants to add high-ticket clients, who just filled out the strategy call form. That specific.

What does a good lead look like? Three qualifying questions. For us it's: what do you sell, what are you doing per month, and what's the one thing you want fixed first. If you can't write your three questions, you don't have a setter problem yet. You have an offer problem.

What's the one job? Book the call. Not sell. Not explain pricing. Not answer twenty questions. Book the call. Every message points there.

What does she never do? Never quote a price. Never promise a result. Never argue. If the lead gets rude, she hands off to a human.

[SCREEN: the actual job-description doc on screen, scrolling]

This is the actual document. The reason most AI setters sound like a robot is that the owner skipped this and typed "you are a helpful assistant" into a chatbot. The AI didn't know the job, so it did a generic one. [LOOP CLOSE]

Step two is the prompt, and I'll give you the whole thing.

(≈ 240 words)

3:18 · Step 2: The prompt (≈ 1:40)

[SCREEN: card "STEP 2 · THE PROMPT". Then the prompt text, full screen, scrolling slowly as Nik reads the headers]

Here's the prompt structure. It's in the description, you can copy it. Five blocks.

Block one, identity. "You are Cassie, the booking assistant for Nikhil Sai's team. You text like a real person on a phone. Short messages. One question at a time."

Block two, the goal. "Your only job is to book a strategy call on the calendar link. You do not sell, you do not quote prices, you do not explain the program."

Block three, the three questions, in order, one per message. And this part matters: "Ask the next question only after they answer the previous one. Never send two questions in one message."

Block four, the rules. "Reply within sixty seconds. If they ask about price, say the call covers that and offer two times. If they say they're not interested, thank them and stop. If they're rude or confused, say a team member will reach out, and tag the conversation for a human."

Block five, the voice. Three example exchanges, written the way I'd text. This is the block that makes her sound like us instead of like a bank.

[SCREEN: side by side, "generic bot" message vs Cassie's message to the same lead] [PATTERN BREAK: comparison]

Look at the difference. Left is the bot without block five: "Thank you for your inquiry! We would love to assist you." Right is Cassie: "Hey Sam, saw you grabbed the form. Quick one, what do you sell?" Same AI. One has a job description and examples. One doesn't.

Hear me out, the prompt isn't the hard part. Step three is where it has to live.

(≈ 250 words)

4:58 · Step 3: Wire it into the CRM (≈ 1:40)

[SCREEN: card "STEP 3 · THE WIRING". GoHighLevel workflow builder, screen recording]

I run this inside GoHighLevel because that's where the calendar, the pipeline and the texting already live. The wiring is one workflow, four pieces.

[SCREEN: workflow steps highlighted as he names them]

Trigger: form submitted. Any form, any ad, any landing page.

Action one: wait sixty seconds. Not instant. Instant feels like a bot. A minute feels like someone saw it.

Action two: the AI conversation step, with the prompt from step two loaded in, and the calendar link attached so she can offer real times.

Action three: when a booking lands, move the lead to "Booked" in the pipeline and send the confirmation. When she tags a conversation for a human, notify the team in a channel.

[SCREEN: the pipeline view, a lead moving from "New" to "Booked"]

That's it. If you use a different CRM the shape is identical. Trigger, short wait, AI conversation, calendar, pipeline update.

[SCREEN: Nav's testimonial card, "Individual result. Not typical."] [PROOF: C39 Nav] [consent pending]

Nav closed a twelve-and-a-half-thousand-dollar deal off a hundred and twenty dollars of ad spend, in his words. Individual result, not typical. That lead came in through a form and got answered fast. The wiring is what makes "fast" happen every time, including the ones that come in at two in the morning.

Now, step four. This is where I see most AI setters break, and it's never the technology.

(≈ 250 words)

6:38 · Step 4: The handoff rules (≈ 1:50)

[SCREEN: card "STEP 4 · WHEN SHE STOPS"]

The number one way an AI setter loses you deals is by not knowing when to stop. She keeps texting a lead who's already annoyed. Or she keeps asking qualifying questions to someone who already said "just send me a time". So I give her four stop rules, and they're all in the prompt.

[SCREEN: four rules, one at a time]

One. If the lead asks for a time, offer two times immediately. Skip whatever questions are left.

Two. If the lead asks anything about price, contracts or guarantees, she says the call covers that, offers times, and tags the conversation. A human reads it within the hour.

Three. If the lead says no, or stop, or not interested, she sends one message: "No problem, thanks for letting me know." Then she stops. No "are you sure". No "one last thing".

Four. If she can't understand the reply, after one clarifying question she hands off. Humans handle confused. AI handles clear.

[SCREEN: a real handoff example, lead asks "how much is it", Cassie replies, tag appears, team notification]

Here's one from last week. Lead asks how much. Cassie: "Good question, that depends on what we'd set up for you, which is what the call is for. I've got two slots open, which one works?" Tag goes to the team. The lead picked a slot. And if they hadn't, a human would have stepped in with context instead of starting from zero.

[SCREEN: Nik, straight to camera] [PATTERN BREAK: slower, flat]

My ex was a cold-calling team. We don't talk about it. The reason Cassie works isn't that she's smarter than a human setter. It's that she never forgets the rules, and she's awake at two.

[CTA, soft] If you want to see this whole build on screen, the prompt, the workflow, the handoff rules, there's a free session linked in the description. I'll come back to it at the end. Step five first, because without it you won't know if any of this is working.

(≈ 300 words)

8:28 · Step 5: The one number (≈ 1:20)

[SCREEN: card "STEP 5 · THE NUMBER". A simple two-column table: "Before" / "After"]

One number. Leads to booked calls. Out of every hundred leads that come in, how many land on the calendar?

Most owners I meet don't know theirs. They know how many leads they bought and how many sales they made, and everything in the middle is a feeling. So before you turn Cassie on, pull last month's leads and count how many booked. Write it down. Then turn her on and count again next month.

[SCREEN: our own dashboard frame, date range visible, numbers blurred except the two percentages] (PROOF: C34 framing; use our own numbers only, "Our own business. Not typical.")

In our own business, this is the number that moved first when we put AI on the leads, and it's the number that pays for everything else. Our own business, not typical.

And if the number doesn't move, you know exactly where to look: the job description, the prompt, the wiring, or the handoff rules. It's one of the four. It's never "AI doesn't work".

(≈ 190 words)

9:48 · What this replaces, and what it doesn't (≈ 1:00)

[SCREEN: two columns, "Cassie does" / "Humans do"]

Let me be straight about what this replaces. Cassie does first reply, qualification, booking, reminders, and the three a.m. shift. She does it in sixty seconds, every time, for every lead.

What she doesn't do is close. The sales call is still a human. Objections, pricing, the actual decision, that's you or your closer. And when a lead is confused or upset, that's a human too.

So the team stays. The part of the job that nobody does well at scale, the fast first reply, goes to something that never sleeps. Your setter, if you keep one, moves up to the calls that need a person.

(≈ 150 words)

10:48 · Recap + the one move this week (≈ 0:45)

[SCREEN: the five steps as a list, one highlighted at a time]

That's the five steps. Job description. Prompt. Wiring. Handoff rules. The number.

Here's your one move. Don't build all five this week. Write the job description. Four questions: who she talks to, what a good lead looks like, the one job, and what she never does. One page. If you can write that page, the rest is an afternoon. If you can't, you found your real problem, and it's upstream of AI.

(≈ 110 words)

11:33 · CTA (≈ 0:35)

[SCREEN: Nik, then the session link on screen as a lower third] [CTA]

If you want to see Cassie built on screen, there's a free session linked in the description. I go through the prompt line by line, the GoHighLevel workflow, the handoff rules, and my team is in the chat answering questions. You leave with your own job description written and the prompt filled in for your business.

It's free. Click the link under this video, pick a time, and bring the last ten leads that came in. That's the only prep.

I'm Nikhil Sai. Answer the lead before your competitor does.

(≈ 100 words)

4. Proof and screen-recording list (capture before the edit)

#	What	Source	Notes
P1	Phone lock screen, one form notification at 2:07 AM	staged on Nik's phone, demo sub-account	no real lead name
P2	GoHighLevel conversation: Cassie reply at 2:08, lead reply, booking at 2:14	demo sub-account, names blurred	must match the timeline said aloud
P3	Old-way timeline graphic (form → setter → voicemail)	editor build	text on screen, no stock footage
P4	The job-description doc, scrolling	`ai-acquisition/` Cassie job description (export a clean copy)	no client names
P5	The five-block prompt, full screen	PACKAGE.md §6 (same text as the description)	editor types it or screen-records the doc
P6	Generic bot vs Cassie side by side	editor build from two real messages
P7	GHL workflow builder: trigger, wait, AI step, pipeline update	demo sub-account screen recording	~40 s continuous
P8	Pipeline: lead moves New → Booked	demo sub-account
P9	Handoff example: "how much is it" → Cassie reply → tag → team notification	demo sub-account, staged
P10	Our own dashboard, leads → booked, date range visible	our own GHL reporting, numbers blurred except the two percentages	caption "Our own business. Not typical."
P11	Morgan testimonial card	`brand-assets/proof/testimonial-morgan-reed-appointments.png`	"Individual result. Not typical." footer, [consent pending]
P12	Nav testimonial card	`brand-assets/proof/testimonial-nav-ganesh-12500-on-120.png`	same footer, [consent pending]
P13	Award wall cutaway, 2 s	`brand-assets/nik/nik-award-plaque-wall-01.png` or live B-roll	C01

5. Shorts cut points (3, 9:16, captions burned in)

#	In → out	First line on screen	Why it stands alone
S1	1:10 → 1:48 (old-way timeline vs Cassie timeline)	"Text at 2:08. Booked at 2:14. I was asleep."	the whole argument in 38 s, ends on "She was closing"
S2	6:38 → 7:40 (the four stop rules)	"The #1 way an AI setter loses you deals"	list format, each rule is a caption card
S3	10:48 → 11:33 (the one move: write the job description)	"Don't build the AI this week. Write this page."	clean instruction, ends on "upstream of AI"

Each Short ends with "Full build on my channel" (no link in the Short, link in the comment).

LF#3 package · "How AI Books My Calls While I Sleep (5-Step Setup)"

Script: SCRIPT.md. Engine: BOOK. Nothing uploaded or scheduled. Rules: PLAYBOOK §3 (clear beats curious), no prices, proof only from PROOF-INDEX (C01, C34, C39 Morgan/Nav [consent pending]). Humanizer pass done.

1. Titles (yt-package title.py, 10 written, top 5 kept)

#	Title	Chars	Score
⭐ 1	How AI Books My Calls While I Sleep (5-Step Setup)	50	94
2	Speed to Lead: Reply in 60 Seconds or Lose the Deal	51	94
3	AI Appointment Setter: The 5-Step Build I Use	45	94
4	Stop Hiring Setters. Build This Instead (5 Steps)	49	94
5	How I Book Sales Calls at 2 AM With an AI Setter	48	94

Pick #1: the promise is the outcome (booked calls, no effort) plus a number. #3 carries the search term "AI appointment setter" and is the A/B challenger. Judge on session opt-ins per 1,000 views, not CTR alone.

2. Thumbnail briefs (1280×720, ≤3 words, real assets only). Pick: A.

A · "SHE BOOKS CALLS" (pick)

Text: SHE BOOKS CALLS, white on black, CALLS in gold. Small tag bottom left: 2:14 AM.
Photo: brand-assets/nik/nik-photo-014.jpg (tier A, holding the plaque), cropped to the left 45%, flat "watch this" look. If the shoot day gives a frame of Nik pointing at the phone, use that instead.
Right side: a phone mockup with the GoHighLevel conversation (P2), Cassie's reply bubble visible, lead name blurred, "Booked" chip in green.
Why: the title says "AI books my calls"; the thumbnail says who (she) and shows the receipt.

B · "60 SECONDS"

Text: 60 SECONDS huge, or they're gone small under it.
Photo: nik/nik-talking-head-pink-blazer-01.jpg, crop the burned caption, skeptical look.
Right: a clock graphic at 0:60 with a red slice, a form icon fading out.
Why: pairs with title #2 (speed to lead), promises the rule.

C · "SETTER vs AI"

Text: SETTER vs AI, SETTER with a red strike.
Photo: nik/nik-award-plaque-wall-01.png upper-left crop.
Right: two timelines stacked (voicemail at 2:15 PM vs booked at 2:14 AM).
Why: pairs with title #4. Strongest for comment bait, weakest for clarity.

3. Description (paste-ready; fill {{placeholders}} at upload)

A lead fills out your form at 2 a.m. Nobody answers. By 9 a.m. they're talking to whoever replied first. Here's the 5-step AI setter I install first in every business: the job description, the prompt (below), the GoHighLevel wiring, the handoff rules, and the one number that tells you it's working.

I'm Nikhil Sai. ClickFunnels gave me their $100M Two Comma Club C award, I spoke at the HighLevel Summit, and I've trained 100,000+ business owners to implement AI live. In our own business we collected $29.4k in cash on $422 of ad spend (our own business, not typical).

Your move this week: write the one-page job description (who she talks to, what a good lead looks like, the one job, what she never does).

THE PROMPT (copy it):
Block 1, identity: "You are {{NAME}}, the booking assistant for {{BUSINESS}}. You text like a real person on a phone. Short messages. One question at a time."
Block 2, goal: "Your only job is to book a {{CALL_NAME}} on {{CALENDAR_LINK}}. You do not sell, quote prices, or explain the program."
Block 3, questions, one per message, in order: 1) {{Q1}} 2) {{Q2}} 3) {{Q3}}. "Ask the next question only after they answer the previous one."
Block 4, rules: "Reply within 60 seconds. If they ask about price, say the call covers that and offer two times. If they say no or stop, thank them once and stop. If they're rude or confused, say a team member will reach out and tag the conversation for a human. If they ask for a time, offer two times immediately."
Block 5, voice: three example exchanges written the way you text.

⏱ Chapters
{{CHAPTERS}}

👉 Watch me build this on screen in the free session: {{SESSION_URL}}
(My team is in the chat. Bring your last 10 leads.)

📺 Next video: {{NEXT_VIDEO_URL}}
🎬 95% of owners use AI wrong (LF#2): {{LF2_URL}}

Results shown are from our own businesses and real clients. They are not typical and not a promise of your results.

#AIAppointmentSetter #SpeedToLead #GoHighLevel #HighTicketSales #NikhilSai

4. Chapters (from the script; re-time to the final cut)

0:00 The 2 a.m. lead nobody answered
0:38 Why speed is the whole game
1:48 Step 1: Write her job description first
3:18 Step 2: The prompt (5 blocks)
4:58 Step 3: Wire it into the CRM
6:38 Step 4: The handoff rules (where setters break)
8:28 Step 5: The one number
9:48 What this replaces, and what it doesn't
10:48 Your one move this week
11:33 See it built on screen

5. Pinned comment

Your move this week: write the one-page job description for your AI setter. Who she talks to, what a good lead looks like, the one job (book the call), and what she never does. Reply "DONE" when it's written and tell me your three qualifying questions. I'll tell you if they're tight enough.

Full prompt is in the description. See it built on screen here: {{SESSION_URL}}

6. One CTA

Free evergreen session ("Copy My AI Acquisition System", Lobbies evergreen room), soft mention at 7:35, full ask at 11:33. Fill {{SESSION_URL}} with the GHL registration page once the room's eg_watch_url is final.

7. Open items

P2/P9 screen recordings need the demo sub-account staged with Cassie live (Karan).
Thumbnail A needs the phone mockup; Claude builds the Paper layout once Nik picks A/B/C.
Morgan and Nav cards: written consent not on file (CONSENT-TRACKER) → keep "[consent pending]" until cleared, else swap for our own-business proof only.

LF4 · ≈1598 spoken words · ≈11 min at 150 wpm

LF#4 · Full script · "Turn Dead Leads Into Booked Calls in 5 Days (AI)"

Engine: ATTRACT → BOOK (database reactivation). Draft only, nothing posted. Package: PACKAGE.md.

Runtime ≈ 12:10 spoken (1,830 words at 150 wpm), ≈ 13 min finished. Setup A, landscape, award wall. Film in 9 chunks (one per # beat), 2 s pause between chunks.

Rules baked in: evergreen, no prices, no phone tiers, one CTA (free session, mid-roll + end), proof only from PROOF-INDEX (first names, "Individual result. Not typical." aloud), "Nikhil Sai"/"Nik", "you" never "you guys", no "live" claim for the session. No invented reply or booking rates: every number said aloud is either our own (tagged) or a hypothetical the viewer fills in. Humanizer pass done.

1. Hook (hookscore.py, 5 written, top 2 kept)

Score	Hook	Formula
⭐ 43	Stop buying leads for a minute. The list you already have can fill your calendar, and the campaign takes five days to run.	The Mistake
42	Every month you spend on new ads while a thousand leads sit in your CRM, you're paying to find people you already found.	The Stake

Winner: the viewer's own mistake named in the first four words, and a time box. The runner-up becomes the second line of the cold open.

2. First 30 seconds, word for word

[SCREEN: black. A CRM contact list scrolling fast, thousands of rows, names blurred. Counter in the corner ticks up: 2,000 … 2,400 … 3,100]

Stop buying leads for a minute. The list you already have can fill your calendar, and the campaign takes five days to run.

Every month you spend on new ads while a thousand old leads sit in your CRM, you're paying to find people you already found.

[SCREEN: cut to Nik, award wall. Lower third: "Nikhil Sai · AI implementation"] (PROOF: C01)

I'm Nikhil Sai. I've trained over a hundred thousand business owners to implement AI live, and the fastest booked calls I've ever seen come from a list the owner forgot they had. In this video you get the five-day reactivation campaign, the five messages, the AI that handles the replies, and the two numbers that tell you if it worked. Day three is the one that books, so stay for that.

(≈ 0:00–0:40, 150 words)

3. Full script with retention beats

0:00 · Cold open + hook + promise + plan (0:40)

(the 150 words above, verbatim)

0:40 · The list you're sitting on (≈ 1:20)

[SCREEN: three buckets, filling one at a time: "Said not now" · "Booked, didn't show" · "Bought the small thing, never upgraded"]

Here's who's in your CRM right now. People who filled out a form and said "not now". People who booked a call and didn't show. People who bought your five-dollar workshop or your book and never heard from you again.

None of them are dead. They're ignored. They raised their hand once, which is the hardest part, and then your follow-up stopped after three emails because that's where the sequence ended.

[SCREEN: a typical "nurture sequence" in GHL: 3 emails, then nothing. The last step says "End"] [PATTERN BREAK: zoom on "End"]

Look at the end of most sequences. Three emails and "End". The lead didn't end. The automation did.

[SCREEN: our own CRM count, number visible, names blurred] (our own business; say the real count on shoot day, no estimate)

In our own business, when we pulled the list, it was bigger than I expected. People from years back. The reason it works is simple. A lead who already knows your name costs nothing to reach, and a reply from them is worth the same booked call as a brand-new lead you paid for.

Now, the reason most reactivation campaigns fail is day one. They open with the pitch. [LOOP OPEN: what to send on day one instead]

(≈ 200 words)

2:00 · Day 1: The permission message (≈ 1:30)

[SCREEN: card "DAY 1 · PERMISSION". The actual text message, phone mockup]

Day one is one text. Not an email. A text, from a real-looking number, to everyone on the list who gave you a phone number. And the text doesn't sell anything. It asks a question.

[SCREEN: the message, full screen]

"Hey Sam, it's Nik from Funneling Media. You grabbed our AI workshop a while back. Quick one: are you still running the coaching business?"

That's the whole message. Here's why each word is there.

[SCREEN: words highlighted as he names them]

Their name, so it isn't a blast. My name and the business, so they place me. The thing they did, so they remember raising their hand. And one yes-or-no question about them, not about me. People answer questions about themselves. Nobody answers "are you ready to take your business to the next level".

[SCREEN: a reply coming in: "Yeah still going, why?"]

The reply you want is anything. "Yeah." "No, I sold it." "Who is this?" Any reply reopens the conversation, and from that second the AI takes over. That's day two. But first, the people who don't reply, because that's most of them. [LOOP CLOSE]

(≈ 240 words)

3:30 · Day 2: The AI handles every reply (≈ 1:50)

[SCREEN: card "DAY 2 · THE REPLY DESK". GHL conversation view, several replies stacking up, names blurred]

Day two is when the replies land, and this is where owners used to give up. Forty replies in a morning, each one needs a human answer, and the human is on sales calls. So the replies sat for a day and went cold again.

Now my girl Cassie answers them. She's an AI, she lives in the CRM, and she has one job on this campaign: find out if the problem is still there, and if it is, book a call.

[SCREEN: three example threads, side by side]

"Yeah still going." Cassie: "Nice. What's the one thing you'd fix first right now, leads, booked calls, or clients not sticking?" They answer. She offers two times.

"No, I sold it." Cassie: "Congrats. What are you doing now?" If they're starting something new, they're a lead again. If not, she thanks them and tags them "closed, not a fit".

"Who is this?" Cassie: "Nik's team at Funneling Media. You grabbed our AI workshop a while back. Still running the business?" Back to the question.

[SCREEN: the stop rules card from LF#3, shortened: no price, no selling, hand off when confused]

Same rules as every AI setter I build. She never quotes a price, she never argues, and the second a reply is confused or angry, a human takes it. The point is every reply gets answered in a minute instead of a day.

[SCREEN: Aaron's testimonial card, "Individual result. Not typical."] [PROOF: C39 Aaron] [consent pending]

Aaron tenxed his leads with us, in his words. Individual result, not typical. A chunk of that was this: the leads he already had, finally answered.

Day three is the message that books. [LOOP OPEN]

(≈ 290 words)

5:20 · Day 3: The reason-to-talk message (≈ 1:50)

[SCREEN: card "DAY 3 · THE REASON". Phone mockup]

Day three goes to everyone who didn't reply on day one. Different message. This one gives them a reason to talk to you now, and the reason has to be about them.

[SCREEN: the message, full screen]

"Sam, one more. We're doing a free session where I map out which of the five places money leaks out of a coaching business is costing you the most. Want me to send the link, or not the right time?"

[SCREEN: words highlighted]

Still one question. Still their choice. And the thing on offer is useful on its own, whether or not they ever buy. That's the rule for this message: give them something they'd want even if they never spend a dollar. A map, an audit, a checklist. If you don't have one, make one before you run this campaign.

[SCREEN: Cassie thread: "send it" → link → "booked" chip]

When they say "send it", Cassie sends the link and offers two times for the call. When they say "not the right time", she says "no problem, I'll check in down the line" and tags them for ninety days out. That tag is your next campaign, already built.

Day three is where most of the bookings come from in our own business, because by now they've seen your name twice in three days, and the second message has a reason attached. Our own business, not typical. [LOOP CLOSE]

[CTA, soft] That session I just described is real, by the way. It's linked in the description, and I'll come back to it at the end.

(≈ 280 words)

7:10 · Day 4 and Day 5: The breakup and the email (≈ 1:30)

[SCREEN: card "DAY 4 · THE BREAKUP"]

Day four is the breakup text, to everyone still silent. "Sam, last one from me. If the timing's off, no stress, I'll stop here. If you want the leak map, just reply 'map'."

[SCREEN: a "map" reply, Cassie sending the link]

The breakup gets replies because it's the first message that gives them an easy out. People reply to "I'll stop" more than to "one more thing". Anyone who replies "map" goes straight into day three's flow.

[SCREEN: card "DAY 5 · THE EMAIL". An email on screen, short, plain text, no images]

Day five is one email, to the whole list including the ones with no phone number. Same reason-to-talk as day three, written as a plain email from me, no banner, no design, one link.

[SCREEN: the pipeline at the end of day five: "Replied" · "Booked" · "Not now (90 days)" · "Closed"]

After five days, every contact on that list is in one of four places. Booked. Replied and in a conversation. Not now, with a date to check back. Or closed. Nobody's floating in "End" anymore.

(≈ 220 words)

8:40 · The two numbers (≈ 1:10)

[SCREEN: card "THE TWO NUMBERS". Two blanks: "Reply rate ___%" · "Booked ___"]

Two numbers. Reply rate: out of everyone you texted, how many said anything. And booked: how many calls landed on the calendar.

I'm not going to tell you what yours should be, because I've seen the same campaign get very different numbers depending on how old the list is and how good the day-three reason is. What I will tell you is how to read them.

[SCREEN: a two-by-two: low reply / high reply × low booked / high booked]

Low reply rate, the day-one message is wrong or the list is wrong. Fix the question. High reply, low booked, Cassie's conversation is wrong or the day-three reason isn't good enough. Fix the offer.

[SCREEN: our own numbers from the last run, date range visible, "Our own business. Not typical."] (PROOF: fill with the real figures on shoot day, else cut this line)

Those are ours from the last run. Our own business, not typical. Track yours, and the campaign gets better every time you run it, because the "not now" tags become the next list.

(≈ 180 words)

9:50 · The three ways this goes wrong (≈ 1:10)

[SCREEN: three cards, one at a time]

Three ways I've seen this blow up.

One. Texting people who never gave you a number for that. If they opted in for emails only, they get the day-five email, not the texts. Keep the list clean.

Two. Sending all five messages as a blast with no AI on the replies. Forty people reply, nobody answers them, and now they're colder than before you started. The reply desk is the campaign. The messages are just the doorbell.

Three. Opening with the pitch. "We have a new program, want to hear about it?" No. Day one is a question about them. The pitch never shows up at all, because the call is where that happens.

(≈ 160 words)

11:00 · Recap + the one move this week (≈ 0:40)

[SCREEN: the five days as a strip: Permission · Reply desk · Reason · Breakup · Email]

That's the five days. Permission text. AI reply desk. Reason-to-talk text. Breakup text. One email.

Your one move this week. Pull the list. Every contact in your CRM with a phone number who hasn't heard from you in ninety days. Count them. That number is how many booked calls you've been leaving in a database, and you'll know it in ten minutes.

(≈ 100 words)

11:40 · CTA (≈ 0:30)

[SCREEN: Nik, session link lower third] [CTA]

If you want the five messages word for word and Cassie's reply prompt for this campaign, they're in the free session linked in the description. I run through all five days on screen, and my team is in the chat answering questions.

It's free. Click the link, pick a time, and bring your list count. That's the only prep.

I'm Nikhil Sai. The cheapest lead you'll ever book is the one you already paid for.

(≈ 90 words)

4. Proof and screen-recording list

#	What	Source	Notes
P1	CRM contact list scrolling, counter ticking	demo sub-account or our own with names blurred
P2	A nurture workflow ending at "End" after 3 emails	demo sub-account	zoom on "End"
P3	Our own CRM contact count	our own GHL, names blurred	say the real count aloud on shoot day
P4–P7	Day 1, 3, 4 texts and day 5 email, phone/email mockups	copy from PACKAGE.md §6
P8	Three Cassie reply threads side by side	demo sub-account, staged	names fake
P9	Stop rules card (short)	reuse LF#3 P9 card
P10	Pipeline with four end-states	demo sub-account
P11	Our own reply rate and booked from the last run, date range visible	our own GHL reporting	"Our own business. Not typical."; if not run yet, cut the line at 9:35
P12	Aaron testimonial card	`brand-assets/proof/testimonial-aaron-fragnito-10x-leads.png`	footer + [consent pending]
P13	Award wall cutaway	`brand-assets/nik/nik-award-plaque-wall-01.png`	C01

5. Shorts cut points

#	In → out	First line on screen	Why
S1	2:00 → 2:55 (the day-one permission text, word by word)	"Day 1 is one text. It doesn't sell anything."	a complete, copyable script in 55 s
S2	9:50 → 11:00 (three ways this goes wrong)	"3 ways a reactivation campaign blows up"	list, ends on "the call is where that happens"
S3	0:40 → 1:25 (three buckets + "the automation ended, the lead didn't")	"Your dead leads aren't dead. They're ignored."	the thesis, ends on the "End" zoom

LF#4 package · "Turn Dead Leads Into Booked Calls in 5 Days (AI)"

Script: SCRIPT.md. Engine: ATTRACT → BOOK (reactivation). Nothing uploaded. Proof: C01, C39 Aaron [consent pending], own-business numbers only if run before the shoot. Humanizer pass done.

1. Titles (title.py, 10 written, top 5)

#	Title	Chars	Score
⭐ 1	Turn Dead Leads Into Booked Calls in 5 Days (AI)	48	98
2	Why Your CRM Is Your Cheapest Lead Source (5 Steps)	51	98
3	How to Revive 1,000 Dead Leads With AI (Step by Step)	53	94
4	Stop Buying Leads. Reactivate the 2,000 You Already Have.	57	94
5	The 5-Day Reactivation Campaign Every Coach Should Run	54	94

Pick #1: outcome + time box + the tool, nothing vague. #3 is the search challenger ("revive dead leads"). "Database reactivation" as a phrase scored 76 (61 chars, cut on desktop) so it lives in the description and tags, not the title.

2. Thumbnail briefs (1280×720, ≤3 words). Pick: A.

A · "DEAD → BOOKED" (pick)

Text: DEAD → BOOKED, DEAD in grey with a strike, BOOKED in gold. Tag: 5 DAYS.
Photo: brand-assets/nik/nik-photo-014.jpg left 45%, or a shoot-day frame of Nik holding up a phone.
Right: a CRM list (blurred, grey) with three rows lit up green and a calendar chip "Booked".
Why: shows the before and after the title promises.

B · "YOU ALREADY PAID"

Text: YOU ALREADY PAID (3 words), small for these leads under it.
Photo: nik/nik-talking-head-pink-blazer-01.jpg, caption cropped, deadpan.
Right: a stack of contact cards with dust/grey tone, one being pulled out.
Why: pairs with title #4 and the "cheapest lead" closing line.

C · "5 TEXTS"

Text: 5 TEXTS huge, 5 DAYS small.
Photo: award wall crop.
Right: five phone bubbles stacked, day labels 1–5, day 3 highlighted.
Why: list promise, strongest for saves.

3. Description (paste-ready)

Stop buying leads for a minute. Every CRM has hundreds of people who raised their hand once and never heard back. Here's the 5-day AI reactivation campaign I run on that list: one permission text, an AI reply desk (Cassie), a reason-to-talk text, a breakup text, and one plain email. Plus the two numbers that tell you what to fix.

I'm Nikhil Sai. ClickFunnels gave me their $100M Two Comma Club C award, I spoke at the HighLevel Summit, and I've trained 100,000+ business owners to implement AI live. In our own business we collected $29.4k in cash on $422 of ad spend (our own business, not typical).

Your move this week: pull every contact with a phone number who hasn't heard from you in 90 days. Count them.

THE 5 DAYS (copy, swap the brackets):
Day 1 text: "Hey [first name], it's [you] from [business]. You grabbed our [thing they did] a while back. Quick one: are you still running the [their business]?"
Day 2: AI answers every reply within a minute. One question about their biggest leak, then two call times. No prices, no selling, hand off when confused.
Day 3 text (non-repliers): "[First name], one more. We're doing a free session where I map out which of the five places money leaks out of a [business type] is costing you the most. Want me to send the link, or not the right time?"
Day 4 text (still silent): "[First name], last one from me. If the timing's off, no stress, I'll stop here. If you want the [useful thing], just reply '[keyword]'."
Day 5 email (whole list, plain text, one link): same reason as day 3, written as a note from you.

⏱ Chapters
{{CHAPTERS}}

👉 The five messages word for word + Cassie's reply prompt, built on screen in the free session: {{SESSION_URL}}

📺 Next video: {{NEXT_VIDEO_URL}}
🎬 The AI setter build (LF#3): {{LF3_URL}}

Results shown are from our own businesses and real clients. They are not typical and not a promise of your results.

#DatabaseReactivation #LeadFollowUp #GoHighLevel #AIforBusiness #NikhilSai

4. Chapters

0:00 The list you're paying to ignore
0:40 Who's actually in your CRM
2:00 Day 1: The permission text
3:30 Day 2: The AI reply desk
5:20 Day 3: The reason-to-talk text
7:10 Day 4 + 5: The breakup and the email
8:40 The two numbers
9:50 Three ways this blows up
11:00 Your one move this week
11:40 Get the five messages

5. Pinned comment

Your move this week: pull every contact with a phone number who hasn't heard from you in 90 days and count them. Reply with the number. I'll tell you which of the five days matters most for a list that size.

All five messages are in the description. See the whole campaign built on screen: {{SESSION_URL}}

6. One CTA

Free evergreen session, soft mention at 7:00, full ask at 11:40. {{SESSION_URL}} = GHL registration page for the evergreen room.

7. Open items

Run the campaign on our own list before the shoot so P3 and P11 are real numbers; otherwise cut the two "our own" lines (marked in the script).
Aaron card: consent pending.
Thumbnail A layout in Paper once Nik picks.

LF5 · ≈1661 spoken words · ≈11 min at 150 wpm

LF#5 · Full script · "Why Your Sales Calls Don't Close (AI Reviewed 50 of Them)"

Engine: CLOSE (AI call review → close rate). Draft only, nothing posted. Package: PACKAGE.md.

Runtime ≈ 12:20 spoken (1,850 words at 150 wpm), ≈ 13 min finished. Setup A, landscape, award wall. Film in 9 chunks, 2 s pause between chunks.

Rules baked in: evergreen, no prices, no phone tiers, one CTA (free session, mid-roll + end), proof only from PROOF-INDEX, first names, "Individual result. Not typical." aloud, "Nikhil Sai"/"Nik", "you" never "you guys". The "50 calls" in the title is our own review; confirm the real count from the Fathom library on shoot day and say that number (title adjusts to match). No invented close-rate figures. Humanizer pass done.

1. Hook (hookscore.py, 5 written, top 2 kept)

Score	Hook	Formula
⭐ 58	If you can't say why your last five lost deals were lost, you're guessing with your closer's salary. Here's the AI review that ends the guessing.	The Reveal
41	Your closer says the leads are bad. The AI call review says the calls are bad. Here's the prompt that settles it in twenty minutes.	The Reveal

Winner is the highest-scoring hook across all four videos (names the cost: the closer's salary). Runner-up becomes the second beat of the cold open.

2. First 30 seconds, word for word

[SCREEN: black. A pipeline board, five cards sliding from "Proposal sent" to "Lost". No reason field filled in on any of them]

If you can't say why your last five lost deals were lost, you're guessing with your closer's salary.

[SCREEN: cut to Nik, award wall. Lower third: "Nikhil Sai · AI implementation"] (PROOF: C01)

Your closer says the leads are bad. The leads say the call was pushy. And nobody's listened to the recording, because who has time to listen to fifty calls.

I'm Nikhil Sai. I've trained over a hundred thousand business owners to implement AI live, and I ran an AI over every sales call in our own library. It found the same seven moments where deals die, over and over, and none of them were the price. In this video you get the seven, the exact prompt I run on every call, and the twenty-minute weekly review that replaces the guessing. Number four is the one that surprised me, so stay for that.

(≈ 0:00–0:42, 160 words)

3. Full script with retention beats

0:00 · Cold open + hook + promise + plan (0:42)

(the 160 words above, verbatim)

0:42 · Why nobody reviews calls (≈ 1:00)

[SCREEN: a Fathom call library, dozens of recordings, names blurred. Scroll]

Every sales call you take is recorded now. Zoom, Fathom, whatever you use. And almost nobody watches them back, because a call is forty-five minutes and you've got fifty of them. So the only feedback your closer gets is the number at the end of the month, and that number doesn't tell anyone what to change.

[SCREEN: the "Lost" column of the pipeline, reason field empty on every card] [PATTERN BREAK]

Look at your lost column. Is there a reason on each card? A real one, with a timestamp? If it says "price" or "timing" on all of them, that's not a reason. That's what the lead said on the way out.

So I gave the recordings to an AI, with one prompt, and asked it to find the moment each deal actually died. Here's what came back. [LOOP OPEN: the seven moments]

(≈ 160 words)

1:42 · The prompt (≈ 1:20)

[SCREEN: card "THE PROMPT". Prompt text full screen, scrolling]

The prompt is in the description. It has four parts.

[SCREEN: four parts highlighted]

Part one, context. "This is a sales call transcript for a high-ticket coaching program. The goal of the call was a decision on the call."

Part two, the scorecard. Seven checks, which I'll go through in a second, each scored yes or no with the timestamp.

Part three, the moment. "Find the single moment where the lead's energy dropped or the decision was lost. Quote the line and give the timestamp."

Part four, the one fix. "In one sentence, what should the closer have said at that moment instead."

[SCREEN: a real output, names blurred: seven yes/no rows, a quoted line with timestamp, a one-line fix]

That's a real output. Seven rows, one quoted line, one fix. It takes about a minute per call, so fifty calls is under an hour, and I only read the fix column.

Now the seven checks. These are the seven moments where our deals died, in the order they happen on the call.

(≈ 200 words)

3:02 · Moments 1 to 3: The first six minutes (≈ 2:00)

[SCREEN: card "1 · NO FRAME". Timestamp badge "0:00–2:00"]

Moment one. No frame. The call starts with "so, tell me about your business" and runs for twenty minutes before anyone says what the call is for. The lead doesn't know a decision is coming, so when it comes, it feels like an ambush.

The check: did the closer say, in the first two minutes, what the call is for and that there'll be a decision at the end? Yes or no.

[SCREEN: card "2 · NO NUMBER". Timestamp "2:00–6:00"]

Moment two. No number. The lead says "I want more clients", the closer says "great", and moves on. Nobody asked how many, what they're doing now, and what the gap is worth. Without the number, the price at the end is compared to nothing.

The check: did the closer get a current number and a target number before minute six? A hundred leads a month, two closing, and they want eight. That's a gap. Now the price has something to stand next to.

[SCREEN: card "3 · THE REAL PROBLEM". Timestamp "6:00–12:00"]

Moment three. The surface problem. The lead says "my ads aren't working". The closer takes that at face value and sells the ads fix. The AI found that in most of those calls the lead said, later, almost in passing, that leads were coming in and nobody followed up. That was the real problem, and the closer pitched the wrong one.

The check: did the closer ask "and what happens to the leads after they come in" or the equivalent? Did the problem change between minute six and minute twelve?

[SCREEN: the three cards side by side, each with a timestamp]

Three moments, all in the first twelve minutes. That's where more of our lost deals died than at the price, and that's the thing nobody expects. [LOOP CLOSE] Number four is the one that surprised me.

(≈ 330 words)

5:02 · Moment 4: The monologue (≈ 1:20)

[SCREEN: card "4 · THE MONOLOGUE". A talk-time bar: closer 80%, lead 20%]

Moment four. The monologue. The AI measured talk time. In the lost calls, the closer talked for most of the pitch section without a single question. Eight minutes, ten minutes, in one case fourteen minutes straight.

Understand this. The lead can't buy while you're talking. They can only buy when they say something, and the longer the monologue, the more their original energy drains out of the call.

[SCREEN: a transcript excerpt, closer's block of text, then the lead's reply: "ok, so what's the investment"]

Here's the tell. After a long monologue, the lead's next line is almost always "so what's the investment". Not because they're ready. Because they want the call to end. The closer hears a buying signal. It's an exit signal.

The check: longest stretch of closer talk time without a question. If it's over three minutes, flag it.

[CTA, soft] I go through the full scorecard on real calls in the free session linked below. I'll come back to it. Three more moments first.

(≈ 220 words)

6:22 · Moments 5 to 7: The close (≈ 2:00)

[SCREEN: card "5 · PRICE WITH NO BRIDGE"]

Moment five. Price with no bridge. The closer goes from the pitch straight to the number. No recap of the gap from moment two. The lead hears a price with nothing next to it.

The check: in the thirty seconds before the price, did the closer restate the lead's own numbers? "You said you're at two a month and you want eight." If yes, the price has a bridge. If no, it's a cliff.

[SCREEN: card "6 · THE FAKE OBJECTION"]

Moment six. The fake objection gets answered. Lead says "I need to think about it". Closer explains the program again, for the third time. The AI found that "think about it" was almost never the real objection. The real one, when the closer bothered to ask, was usually one of three things: a partner, the cash this month, or they didn't believe it would work for them specifically.

The check: when the lead objected, did the closer ask a question, or give a speech? "What specifically do you want to think through?" is a question. The program recap is a speech.

[SCREEN: card "7 · NO NEXT STEP"]

Moment seven. No next step. Call ends with "I'll send you the info, let me know". Nothing on the calendar, no date, no owner. The AI flagged this on nearly every lost call. The deal didn't die on the call. It died in the week after, because nobody had a reason to pick the conversation back up.

The check: did the call end with a time on the calendar? Either the close, or the follow-up with a date and a reason. "Let me know" is a no.

[SCREEN: all seven cards in a row with timestamps: Frame · Number · Real problem · Monologue · Bridge · Objection · Next step]

That's the seven. And the thing to notice is only one of them is about the price, and even that one is about what came before the price.

[SCREEN: Ellis's testimonial card, "Individual result. Not typical."] [PROOF: C39 Ellis] [consent pending]

Ellis sold multiple fifteen-thousand-dollar masterminds using funnels we built, in his words. Individual result, not typical. The calls that closed those had a frame, a number and a next step. The ones that didn't, didn't.

(≈ 380 words)

8:22 · The 20-minute weekly review (≈ 1:30)

[SCREEN: card "THE 20-MINUTE REVIEW". A simple weekly agenda]

Here's how this becomes a habit instead of a one-off. Twenty minutes a week.

[SCREEN: steps appear]

Monday, the AI runs the prompt on every call from last week. Automatic. In our own business it pulls from Fathom into a sheet. One row per call, seven yes/no columns, the moment, the fix.

Then I open the sheet and read one column: the fix. Not the transcripts. Not the scores. The fix column, fifty one-line sentences.

[SCREEN: the sheet, the "fix" column highlighted, a pattern visible: the same fix repeated]

Then I look for the one fix that repeats. There's always one. Last time it was "restate their numbers before the price". That's the only thing the closer works on this week. One thing. Next week, run it again, see if that column changed.

[SCREEN: our own sheet, two weeks side by side, the repeated fix gone from week two] (our own business; show real sheet with names blurred, no close-rate claim unless the number is on the sheet)

That's the whole review. Twenty minutes, one fix a week. And the closer stops getting "do better" as feedback and starts getting a timestamp.

(≈ 240 words)

9:52 · Where this goes wrong (≈ 1:00)

[SCREEN: three cards]

Three mistakes with AI call review.

One. Using it to punish. The second your closer thinks the sheet is a surveillance tool, they'll start performing for the AI instead of talking to the lead. Review the fix column with them, not at them.

Two. Fixing all seven at once. You'll fix none. One fix a week.

Three. Trusting the score without the quote. The AI gets the yes/no wrong sometimes. The quote with the timestamp is the part you can check in ten seconds. Always read the quote.

(≈ 120 words)

10:52 · Recap + the one move this week (≈ 0:50)

[SCREEN: the seven cards, then the prompt]

Seven moments. Frame, number, real problem, monologue, bridge, objection, next step.

Your one move this week. Take your last five lost calls. Run the prompt from the description on each one. Read the five fixes. If the same fix shows up twice, that's what your closer works on this week. Twenty minutes, and you'll know more about why you lose deals than you've known all year.

(≈ 110 words)

11:42 · CTA (≈ 0:35)

[SCREEN: Nik, session link lower third] [CTA]

If you want to watch me run this scorecard on real calls, see the sheet, and get the follow-up messages for each of the seven moments, there's a free session linked in the description. My team is in the chat answering questions.

It's free. Click the link, pick a time, and bring one lost-call recording. That's the only prep.

I'm Nikhil Sai. Stop guessing why deals die.

(≈ 90 words)

4. Proof and screen-recording list

#	What	Source	Notes
P1	Pipeline board, five cards sliding to Lost, reason fields empty	demo sub-account
P2	Fathom call library scroll	our own Fathom, names blurred	confirm the real call count for the title
P3	The prompt, full screen	PACKAGE.md §6
P4	A real AI output: 7 rows, quoted line + timestamp, one fix	run the prompt on one of our own calls, names blurred
P5	Seven moment cards with timestamp badges	editor build
P6	Talk-time bar (closer 80 / lead 20)	from a real transcript, or built from the AI output	no invented %; use the real one
P7	Transcript excerpt: monologue → "so what's the investment"	our own call, names blurred
P8	The weekly review sheet, fix column, repeated fix highlighted	our own sheet
P9	Two weeks side by side	our own sheet	no close-rate claim unless on the sheet
P10	Ellis testimonial card	`brand-assets/proof/testimonial-ellis-hammond-card.png`	footer + [consent pending]
P11	Award wall cutaway	C01

5. Shorts cut points

#	In → out	First line on screen	Why
S1	5:02 → 6:00 (the monologue, "so what's the investment" is an exit signal)	"'What's the investment' is not a buying signal"	contrarian, complete, ends on "flag it"
S2	3:02 → 3:45 (moment 1, no frame, + moment 2, no number)	"Your deal died in the first 6 minutes"	two checks in 43 s
S3	10:52 → 11:42 (the one move: five lost calls, the prompt)	"Run this on your last 5 lost calls"	instruction, ends on "all year"

LF#5 package · "Why Your Sales Calls Don't Close (AI Reviewed 50 of Them)"

Script: SCRIPT.md. Engine: CLOSE. Nothing uploaded. Proof: C01, C39 Ellis [consent pending], own call library. The "50" must match the real reviewed count on shoot day. Humanizer pass done.

1. Titles (title.py, top 5 of 10)

#	Title	Chars	Score
⭐ 1	Why Your Sales Calls Don't Close (AI Reviewed 50 of Them)	57	98
2	I Let AI Grade 50 Sales Calls. Here's Why Deals Die.	52	94
3	7 Reasons High-Ticket Calls Don't Close (and the AI Fix)	56	94
4	The AI Call Scorecard That Raised Our Close Rate (7 Checks)	59	94
5	How to Review Sales Calls With AI in 20 Minutes a Week	54	94

Pick #1: the viewer's problem in the first five words, the receipt in brackets. #3 is the list challenger. #4 is only usable if the sheet shows a close-rate change (otherwise it's a claim we can't show).

2. Thumbnail briefs (≤3 words). Pick: A.

A · "MINUTE 6" (pick)

Text: DIED AT MINUTE 6 reads as 3 units: DIED red, MINUTE 6 white. Tag: 50 CALLS.
Photo: Nik, flat stare, brand-assets/nik/nik-photo-014.jpg crop or a shoot-day frame with headphones on.
Right: a transcript block with one line highlighted yellow and a timestamp badge "06:12".
Why: the title says calls don't close; the thumb says when, which is the surprise.

B · "NOT THE PRICE"

Text: NOT THE PRICE, PRICE struck through.
Photo: pink-blazer talking head, caption cropped.
Right: a pipeline "Lost" column with "price" written on every card and a red X over it.
Why: the contrarian promise, pairs with title #2.

C · "7 CHECKS"

Text: 7 CHECKS huge, every call small.
Photo: award wall crop.
Right: a seven-row scorecard, yes/no ticks, one row red.
Why: deliverable thumbnail, pairs with #3/#4.

3. Description (paste-ready)

I ran an AI over every sales call in our own library and asked it one question: where did the deal actually die? Seven moments came back, over and over, and only one of them is about the price. Here are all seven, the exact prompt (below), and the 20-minute weekly review that replaces "do better" with a timestamp.

I'm Nikhil Sai. ClickFunnels gave me their $100M Two Comma Club C award, I spoke at the HighLevel Summit, and I've trained 100,000+ business owners to implement AI live. In our own business we collected $100,928 in 30 days through our own funnels (our own business, not typical).

Your move this week: run the prompt on your last 5 lost calls and read the 5 fixes.

THE PROMPT (paste the transcript after it):
"This is a sales call transcript for [your offer, e.g. a high-ticket coaching program]. The goal of the call was a decision on the call.
Score these 7 checks yes/no, each with a timestamp:
1. Frame: did the closer say in the first 2 minutes what the call is for and that a decision comes at the end?
2. Number: did the closer get a current number and a target number before minute 6?
3. Real problem: did the stated problem change between minute 6 and 12 after a follow-up question?
4. Monologue: longest closer talk stretch without a question (flag if over 3 minutes).
5. Bridge: in the 30 seconds before the price, did the closer restate the lead's own numbers?
6. Objection: when the lead objected, did the closer ask a question or give a speech?
7. Next step: did the call end with a time on the calendar?
Then: find the single moment where the lead's energy dropped or the decision was lost. Quote the line with the timestamp.
Then: in one sentence, what should the closer have said at that moment instead."

⏱ Chapters
{{CHAPTERS}}

👉 Watch me run the scorecard on real calls in the free session: {{SESSION_URL}}
(My team is in the chat. Bring one lost-call recording.)

📺 Next video: {{NEXT_VIDEO_URL}}
🎬 The AI setter build (LF#3): {{LF3_URL}}

Results shown are from our own businesses and real clients. They are not typical and not a promise of your results.

#SalesCalls #CloseRate #AISalesCoaching #HighTicketSales #NikhilSai

4. Chapters

0:00 Your last five lost deals
0:42 Why nobody reviews calls
1:42 The prompt (4 parts)
3:02 Moments 1–3: The first six minutes
5:02 Moment 4: The monologue
6:22 Moments 5–7: The close
8:22 The 20-minute weekly review
9:52 Where this goes wrong
10:52 Your one move this week
11:42 See it run on real calls

5. Pinned comment

Your move this week: run the prompt (in the description) on your last 5 lost calls and read the 5 one-line fixes. Reply with the fix that showed up more than once. I'll tell you the follow-up message for that moment.

See the scorecard run on real calls: {{SESSION_URL}}

6. One CTA

Free evergreen session, soft at 6:05, full at 11:42.

7. Open items

Confirm the real number of calls reviewed (Fathom library) and update the title/thumb tag to match.
P4/P6/P7/P8/P9 come from our own calls: Karan runs the prompt on last week's calls and exports the sheet before the shoot.
Ellis card: consent pending.

LF6 · ≈1570 spoken words · ≈10 min at 150 wpm

LF#6 · Full script · "The Exact AI System I Install in 7 Days (ABCDE Walkthrough)"

Engine: all five (the full install, day by day). This is the video that feeds the apply funnel hardest: it shows what "done for you" looks like without naming a tier or a price. Draft only, nothing posted. Package: PACKAGE.md.

Runtime ≈ 12:40 spoken (1,900 words at 150 wpm), ≈ 14 min finished with screen holds. Setup A, landscape, award wall. Film in 10 chunks, 2 s pause between chunks.

Rules baked in: evergreen, no prices, no phone tiers, no "DFY" or "DWY" words (say "when my team installs this"), one CTA (free session, mid-roll + end), proof only from PROOF-INDEX, first names, "Individual result. Not typical." aloud, "Nikhil Sai"/"Nik", "you" never "you guys". The business in the walkthrough is a composite ("a coaching business", no name, no numbers claimed as theirs). Humanizer pass done.

1. Hook (hookscore.py, 5 written, top 2 kept)

Score	Hook	Formula
⭐ 44	You don't need a year of AI experiments. You need seven days and the right order. Here's the install, day by day, nothing skipped.	The Reveal
43	If your AI setup took longer than a week and still doesn't book calls, you built it in the wrong order. Here's the order.	The Mistake

Winner leads; runner-up is the second line of the cold open (they're the same argument from two sides).

2. First 30 seconds, word for word

[SCREEN: black. A seven-day strip, Mon to Sun, empty. Each day fills as he speaks: "Map · Book · Attract · Close · Deliver · Expand · Numbers"]

You don't need a year of AI experiments. You need seven days and the right order.

If your AI setup took longer than a week and still doesn't book calls, you built it in the wrong order. Here's the order.

[SCREEN: cut to Nik, award wall. Lower third: "Nikhil Sai · AI implementation"] (PROOF: C01)

I'm Nikhil Sai. I've trained over a hundred thousand business owners to implement AI live, and when my team installs this in a business, it takes seven days. Not because we're fast. Because the order is fixed and nothing gets built before the thing it depends on.

In this video I walk through all seven days for a coaching business, what gets built each day, what it's wired to, and what it does to booked calls. Day two is the day that pays for the rest, and day six is the one almost nobody builds. Stay for both.

(≈ 0:00–0:44, 170 words)

3. Full script with retention beats

0:00 · Cold open + hook + promise + plan (0:44)

(the 170 words above, verbatim)

0:44 · Day 1: The map (≈ 1:30)

[SCREEN: card "DAY 1 · THE MAP". The five-box strip: ATTRACT · BOOK · CLOSE · DELIVER · EXPAND]

Day one, nothing gets built. Day one is the map. Five boxes, Attract, Book, Close, Deliver, Expand, and for each one, two questions. What's the number right now, and where does it leak.

[SCREEN: the map filled in for the coaching business, numbers as blanks the owner fills: "Leads/mo ___ · Booked ___ · Showed ___ · Closed ___ · Renewed ___"]

For this coaching business: leads come in from ads and a podcast. Some book. Fewer show. Some close. Clients stay about as long as the program and then drift. Nothing's broken. Every box just leaks a little.

[SCREEN: the map with one box circled red: BOOK]

Then we pick the one box bleeding the most. Here it was Book. Leads came in all day and got called back the next afternoon. That's the same leak I see in most businesses doing twenty to fifty thousand a month, and it's the reason day two is day two.

Day one also sets the one number we'll watch all week. For this business, leads to booked calls. One number. Everything else is noise until that moves.

[SCREEN: Nik to camera] [PATTERN BREAK]

Hear me out. If you skip the map and start with the tools, you'll build the thing that's most fun, not the thing that's bleeding. Every owner who tells me "AI didn't work for us" skipped day one.

(≈ 230 words)

2:14 · Day 2: Book (≈ 1:40)

[SCREEN: card "DAY 2 · BOOK". GHL workflow: form → 60 s wait → AI conversation → calendar → pipeline]

Day two is the AI setter. Every form, every ad, every podcast link now lands in one place, and within sixty seconds my girl Cassie texts them. She asks three questions, one at a time, and offers two call times. If they ask about price, she says the call covers it. If they're confused, a human takes over.

I did a whole video on this build, it's linked below. The short version is four pieces: a job description, a five-block prompt, a workflow in the CRM, and four stop rules.

[SCREEN: Cassie conversation, lead name blurred, "Booked" chip]

By the end of day two, the coaching business has something it never had: a reply to every lead in under a minute, at any hour. That's why it's day two. It's the fastest number to move, and when the owner sees booked calls go up on day three, every other day gets easier, because now they believe the map.

[SCREEN: Morgan's testimonial card, "Individual result. Not typical."] [PROOF: C39 Morgan] [consent pending]

Morgan got his appointments automated with this piece, in his words. Individual result, not typical.

Day three is where most people would have started. [LOOP OPEN: why Attract comes after Book]

(≈ 250 words)

3:54 · Day 3: Attract (≈ 1:30)

[SCREEN: card "DAY 3 · ATTRACT". Two columns: "Ads + content" / "The list you already have"]

Day three is Attract, and it comes after Book on purpose. Turning on more leads before the follow-up works just means more leads get ignored faster. [LOOP CLOSE]

Attract has two halves. The new leads, which for this business is ads and the podcast. And the old leads, the list sitting in the CRM. On day three we start with the old list, because it's free.

[SCREEN: the five-day reactivation strip: Permission · Reply desk · Reason · Breakup · Email]

That's the reactivation campaign. Five messages over five days, Cassie on the replies. Also its own video, linked below. It starts on day three and runs on its own while we build the rest of the week.

[SCREEN: an ad creative, the landing page, the form, an arrow into the day-two workflow]

For the new leads, day three is one thing: every ad and every page points at the same form, and that form fires the day-two workflow. No separate funnels, no lead that lands somewhere Cassie can't see. One door.

[SCREEN: Aaron's card, "Individual result. Not typical."] [PROOF: C39 Aaron] [consent pending]

Aaron tenxed his leads with us, in his words. Individual result, not typical. One door, every lead answered.

(≈ 220 words)

5:24 · Day 4: Close (≈ 1:40)

[SCREEN: card "DAY 4 · CLOSE". Three pieces: "Pre-call" · "Call review" · "Post-call follow-up"]

Day four is Close, and it has three pieces, none of which replace the closer.

Piece one, pre-call. The night before every call, the AI writes a one-page brief from the Cassie conversation and anything on the lead's website. What they sell, their numbers, their words for the problem. The closer walks in knowing the lead.

[SCREEN: a pre-call brief, one page, names blurred]

Piece two, call review. Every recorded call gets the seven-check scorecard. Frame, number, real problem, monologue, bridge, objection, next step. One quoted line, one fix. That's also its own video, linked below.

[SCREEN: the scorecard sheet, fix column]

Piece three, post-call follow-up. If the call ended without a decision, the AI drafts the follow-up from the transcript, in the closer's voice, referencing what the lead actually said. The closer reads it, edits one line, sends it. The deal that used to die in the week after the call now has a reason to come back.

[SCREEN: a drafted follow-up, "Draft, closer approves" badge]

Notice what's human here. The call. The decision. The send. What's AI is the prep, the review, and the first draft. Day four doesn't fire anyone. It makes the closer walk in prepared and walk out with a next step.

[CTA, soft] I show the full Close setup on screen in the free session linked below. Back to it at the end. Day five.

(≈ 280 words)

7:04 · Day 5: Deliver (≈ 1:30)

[SCREEN: card "DAY 5 · DELIVER". Onboarding flow: "Paid → welcome → intake → kickoff booked → week-1 check-in"]

Day five is Deliver, and it's the day owners think they don't need. They do. The fastest way to lose a client is the first week after they pay, when the excitement drops and nothing happens for four days.

[SCREEN: the onboarding workflow in GHL]

So day five builds the first fourteen days of the client's life. The second they pay: welcome, intake form, kickoff call booked, all automatic. Day three, an AI check-in that asks one question, "did you finish the intake", and nudges if not. Day seven, another. Day fourteen, a human call.

[SCREEN: a support inbox, AI-drafted replies with a "review" badge]

And for the questions clients send during the program, the AI drafts the answer from the program material, and a human approves it. Clients get answers in an hour instead of a day, and the team isn't typing the same answer forty times.

For this coaching business, day five is what made the program feel bigger than it was, because the client heard from someone every few days without the owner doing anything.

(≈ 220 words)

8:34 · Day 6: Expand (≈ 1:30)

[SCREEN: card "DAY 6 · EXPAND". Three triggers: "Result logged" · "Day 60" · "Renewal minus 30"]

Day six is Expand, and this is the one almost nobody builds, because it's the one that pays last. It's three triggers.

[SCREEN: trigger one]

Trigger one, a result. The moment a client logs a win, a booked call, a sale, whatever the program promised, the AI drafts two things: a one-line testimonial request, and a referral ask. "Who else do you know with this problem?" Sent when they're happiest, not at the end when they're tired.

[SCREEN: trigger two]

Trigger two, day sixty. A check-in that asks what they want next. The answer tells you who's ready for the bigger program before they ever ask.

[SCREEN: trigger three]

Trigger three, renewal minus thirty days. A conversation starts a month before the end, with their results pulled in, so renewal is a conversation about what's next instead of an invoice that surprises them.

[SCREEN: Nik to camera]

Understand this. Every one of these is a message the owner always meant to send and never did. Day six is just the owner's good intentions, scheduled.

(≈ 220 words)

10:04 · Day 7: The numbers (≈ 1:10)

[SCREEN: card "DAY 7 · THE NUMBERS". A five-row dashboard: one number per box]

Day seven, nothing gets built. Day seven is the dashboard. One number per box. Leads. Booked. Showed and closed. Clients active past day fourteen. Renewals and referrals. Five numbers, one screen, refreshed every morning.

[SCREEN: the dashboard, our own, numbers blurred except the structure] (our own; "Our own business. Not typical.")

And the one number from day one is at the top. For this business, leads to booked calls. If that moved in the week, the install worked. If it didn't, we know which of the five boxes to open, because each one has a number.

Then the owner gets one page: what was built, where it lives, what each number means, and what to fix first if it drops. That page is the whole handover.

(≈ 170 words)

11:14 · Recap + the one move this week (≈ 0:50)

[SCREEN: the seven-day strip, complete]

Seven days. Map. Book. Attract. Close. Deliver. Expand. Numbers. In that order, every time.

Your one move this week is day one. Draw the five boxes. Put the real number in each one, even if it's a guess. Circle the one bleeding the most. Write down the one number you'll watch. That's an hour, and it's the hour that decides whether the other six days work.

(≈ 100 words)

12:04 · CTA (≈ 0:40)

[SCREEN: Nik, session link lower third] [CTA]

If you want me to walk through all seven days on screen for your kind of business, and show you what each day looks like inside the CRM, there's a free session linked in the description. My team is in the chat answering questions, and you leave with your day-one map filled in.

It's free. Click the link, pick a time, and bring your rough numbers for the five boxes. That's the only prep.

I'm Nikhil Sai. Seven days, right order.

(≈ 100 words)

4. Proof and screen-recording list

#	What	Source	Notes
P1	Seven-day strip, filling	editor build
P2	Five-box map, blank then filled with blanks	editor build (ABCDE strip from LF#2 kit)	no client numbers
P3	GHL workflow (form → wait → AI → calendar → pipeline)	demo sub-account	reuse LF#3 P7
P4	Cassie conversation + Booked chip	demo sub-account	reuse LF#3 P2
P5	Reactivation strip	reuse LF#4 kit
P6	Ad → page → form → workflow arrow	editor build from real creative (no client)
P7	Pre-call brief, one page	generate from a demo Cassie convo	names fake
P8	Scorecard sheet	reuse LF#5 P8
P9	Drafted follow-up with "Draft, closer approves" badge	demo
P10	Onboarding workflow in GHL	demo sub-account
P11	Support inbox with AI drafts + review badge	demo
P12	Three Expand triggers as workflow cards	demo sub-account
P13	Five-number dashboard	our own, numbers blurred	"Our own business. Not typical."
P14	Morgan + Aaron cards	brand-assets/proof	footer + [consent pending]
P15	Award wall cutaway	C01

5. Shorts cut points

#	In → out	First line on screen	Why
S1	0:44 → 1:40 (day one, the map, "skipped day one")	"Day 1 of an AI install: build nothing"	complete idea, ends on the "AI didn't work for us" line
S2	8:34 → 9:50 (day six, three Expand triggers)	"The AI nobody builds (it pays last)"	three triggers as caption cards, ends on "good intentions, scheduled"
S3	3:54 → 4:20 (why Attract comes after Book)	"More leads before follow-up = more ignored leads"	26 s contrarian cut

LF#6 package · "The Exact AI System I Install in 7 Days (ABCDE Walkthrough)"

Script: SCRIPT.md. Engine: all five. Nothing uploaded. No tier names or prices anywhere. Proof: C01, C39 Morgan/Aaron [consent pending], own dashboard. Humanizer pass done.

1. Titles (title.py, top 5 of 10)

#	Title	Chars	Score
⭐ 1	The Exact AI System I Install in 7 Days (ABCDE Walkthrough)	59	98
2	Full AI Sales System in 7 Days (Day-by-Day Walkthrough)	55	94
3	The 7-Day AI Install: What I Build for a Coaching Business	58	94
4	What a $30k/mo Coach Gets When I Install AI in 7 Days	53	94
5	I Install AI Sales Systems in 7 Days. Here's Every Day.	55	94

Pick #1 (59 chars, one under the desktop cut; if it truncates on a test upload, use #2). #4 is the audience-named challenger. "ABCDE" is the brand frame from LF#2; keep it for the search thread across videos.

2. Thumbnail briefs (≤3 words). Pick: A.

A · "7 DAYS" (pick)

Text: 7 DAYS huge, gold. Tag: the whole system.
Photo: brand-assets/nik/nik-photo-014.jpg (plaque) left 40%.
Right: the seven-day strip as seven filled tiles, day 2 and day 6 lit.
Why: the title says what and the thumb shows the shape of the week.

B · "RIGHT ORDER"

Text: RIGHT ORDER, with a small numbered 1–7 ladder beside it.
Photo: pink-blazer talking head, caption cropped.
Right: two columns, "wrong order" (scrambled boxes, red) vs "right order" (ABCDE, green).
Why: pairs with the hook. Comment bait for "what's the order".

C · "DAY 2 PAYS"

Text: DAY 2 PAYS, for the rest small.
Photo: award wall crop.
Right: Cassie's booked conversation (LF#3 P2 mockup).
Why: curiosity on one day; weakest for clarity, strongest for returning viewers.

3. Description (paste-ready)

When my team installs AI in a business, it takes 7 days, and the order never changes: Map, Book, Attract, Close, Deliver, Expand, Numbers. Here's every day for a coaching business, what gets built, what it's wired to, and what it does to booked calls.

I'm Nikhil Sai. ClickFunnels gave me their $100M Two Comma Club C award, I spoke at the HighLevel Summit, and I've trained 100,000+ business owners to implement AI live. In our own business we collected $100,928 in 30 days through our own funnels (our own business, not typical).

Your move this week: Day 1. Draw the five boxes (Attract, Book, Close, Deliver, Expand), put a real number in each, circle the one bleeding the most, and write down the one number you'll watch.

THE 7 DAYS:
Day 1 Map: five boxes, two questions each (the number, the leak). Pick one box. Pick one number.
Day 2 Book: AI setter on every form (60 s reply, 3 questions, 2 times, 4 stop rules). Build video: {{LF3_URL}}
Day 3 Attract: reactivate the old list (5 messages, AI on replies) + one door for all new leads. Build video: {{LF4_URL}}
Day 4 Close: pre-call brief, 7-check call review, AI-drafted follow-up the closer approves. Build video: {{LF5_URL}}
Day 5 Deliver: first 14 days automated (welcome, intake, kickoff, day 3/7 check-ins, day 14 human call) + AI-drafted support answers.
Day 6 Expand: three triggers (result logged → testimonial + referral ask; day 60 → what's next; renewal minus 30 → results-led conversation).
Day 7 Numbers: one number per box, one screen, every morning.

⏱ Chapters
{{CHAPTERS}}

👉 Watch all seven days on screen for your kind of business, free session: {{SESSION_URL}}
(My team is in the chat. Bring rough numbers for the five boxes.)

📺 Next video: {{NEXT_VIDEO_URL}}
🎬 95% of owners use AI wrong (LF#2): {{LF2_URL}}

Results shown are from our own businesses and real clients. They are not typical and not a promise of your results.

#AIforBusiness #SalesSystem #GoHighLevel #CoachingBusiness #NikhilSai

4. Chapters

0:00 Seven days, right order
0:44 Day 1: The map (build nothing)
2:14 Day 2: Book (the AI setter)
3:54 Day 3: Attract (old list first, one door)
5:24 Day 4: Close (brief, review, follow-up)
7:04 Day 5: Deliver (the first 14 days)
8:34 Day 6: Expand (the three triggers)
10:04 Day 7: The numbers
11:14 Your one move this week
12:04 See all seven days on screen

5. Pinned comment

Your move this week is Day 1: draw the five boxes, put a real number in each, circle the one bleeding the most. Reply with which box you circled and I'll tell you what day 2 looks like for that box.

See all seven days built on screen: {{SESSION_URL}}

6. One CTA

Free evergreen session, soft at 6:50, full at 12:04. This video is the strongest feeder into the apply path: the session's offer cards do the tier talk, the video never does.

7. Open items

Publish order: LF#3 → LF#4 → LF#5 → LF#6, so the three "build video" links in this description exist by upload day.
Demo sub-account needs the onboarding workflow (P10) and the three Expand triggers (P12) staged (Karan).
Morgan/Aaron cards: consent pending.