Cancun Modular Shoot v2 — AI-First–Led · 50 Hooks · 5 Bodies · 7 CTAs

2026-07-23 · v2 (supersedes this morning's v1: re-weighted around the AI-First angle per Akash + ad data) · Alex St. Pierre on camera · 9:16 vertical · foundation & waterproofing spoken as the niche (clone winners per niche later)

What the ad data actually says (Jul 8–20, first rounds)

Honest sample-size warning first: total tracked spend is ~$360 with 7 leads. Directional, not statistical. Within that:

• - The manifesto-funnel statics won. "SP AO | {niche} | MF | static" ads took ~$150 of spend and produced 5 of the 7 leads (DR niche alone: $108 → 4 leads). The manifesto IS the AI-First playbook piece — the data and your instinct agree.
• - Statics beat everything else outright (Jul-20 weekend readout: $154 → 3 leads, ~$51 CPL). Video has effectively never been tested: only text-on-screen VSLs ran, 10 of them sit wrongly disapproved, and zero real filmed footage exists beyond the original VSL. This shoot is the first true video test.
• - The funnel bottleneck is LPV → form-start, not clicks — hooks are getting attention; the handoff to the page is where it dies. That's why hook↔caption↔LP alignment (triplet contract) matters more than more hook volume.
• - The champion selection already crowned AI-First: all 5 champion hooks in the registry are ai-first-vs-ai-added-split (R4/c1–c5), cloned into the live CTRL pack. R4 (27 ads, AI-First × 9 niches) and R2 relit Jul-22 — in learning until ~Jul-25, no readable signal yet.

Conclusion: AI-First is the lead angle for this shoot. Every other family below is trimmed and gets an AI-First bridge line so any hook can stitch into the AI-First bodies.

How this works (why you don't shoot permutations)

You shoot ~70 modular clips — hooks, bodies, CTAs as separate takes with matched framing/audio — and permutations are assembled in edit. Meta reads main effects, not interactions (fractional factorial, Jul-20 session): each hook rotates across bodies and CTAs so nothing is confounded. ~100 live ads max; Meta's allocation is the signal.

Test waves: 1) hook read (every hook once, bodies/CTAs balanced) → 2) body read on top ~10 hooks → 3) destination read on winners. Paid ads ship as the v3 triplet: hook + "If you…" ATF caption + matching LP dynamic headline.

Batch A — 50 hooks (3–10 sec each, 2+ takes, slate every take with its ID)

★ = not yet in sp_creative_angles — register before ads are built.

A1 · AI-FIRST SPLIT — the lead block, 12 hooks

The champion family. Shoot these first, freshest energy, best locations.

1. 1. "Some foundation companies are adding AI tools. One company in your market is about to be AI-First. Those are not the same thing." (r4/c1)
2. 2. "Every market gets one AI-First foundation & waterproofing company. It's either you, or the guy you keep losing bids to." (r4/c2)
3. 3. "I screened 805 contractor websites. I found the same 8 leaks on almost all of them." (r4/c3)
4. 4. "Want 10 or more new foundation jobs in the next 90 days? I build the whole system. You pay only if it works." (r4/c4)
5. 5. "Becoming the AI-First foundation company in your market takes 90 days. Or you don't pay." (r4/c5)
6. 6. "Looking for 5 foundation & waterproofing companies that want 10+ new jobs in the next 90 days." (w5/opener-lf5-scarcity)
7. 7. "Most foundation companies are adding AI. The ones pulling away are AI-First. The difference is the whole game." (ai-first-split)
8. 8. "Adding AI is a purchase. Becoming AI-First is a rebuild. Your competitor can copy the first one." ★
9. 9. "In two years every foundation company will say they use AI. One per market will have been built on it." ★
10. 10. "The AI-First contractor doesn't buy leads. He owns the machine that makes them." ★
11. 11. "What happens to your foundation company the day a competitor goes AI-First? I'll show you, because I build them." ★
12. 12. "I run this entire company with Claude. Now I build the same machine for foundation & waterproofing contractors." ★

A2 · Founder / The Bet — 6 (reframed to land on AI-First)

Each closes able to bridge: "…so I rebuilt the whole engine with Claude — AI-First from the ground up."

13. 13. "I built the ad machines for some of the biggest names in foundation marketing. Then I quit. What I saw inside is costing you money every month." (S8-192)
14. 14. "I wasn't the face. I was the engine room." (S8-193)
15. 15. "I quit one of the biggest names in foundation marketing with 3 weeks of money left." (S8-194)
16. 16. "Everyone said hoard what you know. I had 3 weeks of cash and did the opposite. Watch." (S8-195)
17. 17. "Don't trust my story. Read my terms. The risk sits on my side now." (S8-200)
18. 18. "Contractors have been trained to pay before they can see anything. That is backwards." (S8-199)

A3 · Proof / inside-knowledge — 3

19. 19. "What 10,000 booked inspections taught me that no agency will say out loud."
20. 20. "Two years running lead gen inside one of the biggest foundation & waterproofing operations in the country. One number was always a lie." (reel-09)
21. 21. "I audited hundreds of foundation companies' marketing. Almost every one has the same leak." (reel-07)

A4 · Calendar confidence — 4

22. 22. "You know you need to hire. You're afraid to. Here's the only number that says it's safe."
23. 23. "You don't have a lead problem. You have a backlog problem — a calendar too thin to make the next hire safe." ★
24. 24. "Ran the Sales Rep Calendar Test on 60 foundation companies. Here's the pattern."
25. 25. "Before you hire any marketing agency, ask these 4 questions. Takes 10 minutes."

A5 · Qualify first — 3

26. 26. "Never let your sales rep walk into a house blind."
27. 27. "Never send trucks to unqualified appointments."
28. 28. "A form isn't qualification. A conversation is." ★

A6 · Speed / follow-up — 3

29. 29. "Your lead rep responds in 4 hours on average. Nobody told you the job goes to whoever responds in 5 seconds."
30. 30. "In foundation & waterproofing, a lead that waits an hour isn't a lead anymore. It's a memory." (reel-06)
31. 31. "60% of your booked estimates never confirm. The follow-up system is the job."

A7 · Real cost / the math — 3

32. 32. "That $100 lead costs you $1,400 once your rep drives 240 miles to a no-show. You're optimizing the wrong number."
33. 33. "Your real cost per booked appointment isn't $150. It's $400. Here's the math."
34. 34. "6 estimates, zero bookings. Your sales rep just drove 240 miles for $0."

A8 · Shared leads — 2

35. 35. "Shared leads are shared losses. A homeowner sold to five contractors isn't a lead, it's a race."
36. 36. "Still buying shared leads? You're training your competition's close rate."

A9 · Google ceiling / the 97% — 4

37. 37. "Every contractor on Google is fighting over 3% of the market. Here's where the other 97% live."
38. 38. "97% of homeowners with a structural problem haven't searched for a contractor yet. They have the symptom. Nobody's reaching them."
39. 39. "Every $3M foundation company hits the same wall. Google's the wall."
40. 40. "Facebook was never Google. You're using a creation channel like a capture channel." ★

A10 · The offer — 2

41. 41. "No lead vendors. No depending on referrals. No expensive advertising." ★ (jargon line — verbatim, never simplified)
42. 42. "Zero down. You pay after results. Week to week. Nobody in this industry offers that. Here's why I can." ★

A11 · Premium frame / owner ceiling — 4

43. 43. "If you do permanent structural work and you're losing to patch guys on Angi, you don't have a price problem."
44. 44. "'I'm the best salesman, I'll keep doing the estimates.' That's exactly why you're stuck at $1.4M."
45. 45. "Operators who break $10M aren't better marketers. They stopped running the business and built a machine."
46. 46. "Never assume the same playbook scales from $1M to $10M. There's a tier where 80% of contractors stall."

A12 · Anti-retainer / AI posture / data loop — 4

47. 47. "Never sign another marketing retainer. Not with me. Not with anyone."
48. 48. "An AI receptionist booking your $14K jobs is killing your trust. AI belongs behind the glass, not on the phone."
49. 49. "Feed Meta your closed jobs, not clicks. Send junk events, it buys junk. Send qualified events, it hunts your real buyers."
50. 50. "Home services ad costs are up 30% year over year. The contractors winning are testing faster, not spending more."

Batch B — 5 bodies (20–45 sec, start mid-thought so any hook stitches in)

All five now route through the AI-First frame.

• - B1 · The Bet → AI-First rebuild (~35s). 3 weeks of runway → quit the most established operation in the niche → gave the playbook away in public → "then I rebuilt the entire engine with Claude. Not AI added on top — AI-First from the ground up. That's the machine I install now." *(Fixes the offer-stale Jun-22 VSL close.)*
• - B2 · The Split (mechanism, ~40s). Adding AI vs being AI-First: adding AI = same broken process, faster. AI-First = brand over bought leads, flywheels over campaigns, Claude in the middle of creative, qualification, booking, and the data loop. The 97% rule and "every sold job trains the system" fold in here. "One AI-First company per market. That's the whole model."
• - B3 · The Proof (~30s). 805 contractor sites screened, same 8 leaks · two years running the engine inside the biggest name in the niche · $5M+/10X–23X stats bar. Defensible claims only; no client names.
• - B4 · The Offer (~30s). 10+ new {niche} jobs in 90 days · $0 down, pay after results, week to week · one company per niche per market · "No lead vendors. No depending on referrals. No expensive advertising." · Built with Claude.
• - B5 · The Demo VO (~30s). Voice-over for the demo-bot screen recording — framed as "this is what an AI-First install looks like on day one" (homeowner books at 11 PM, nobody at the office). Record clean VO in the room; screen capture off-trip.

Batch C — CTA closers (5–15 sec; paid + organic variant of each)

Data note: the magnet-funnel statics produced the leads — magnet CTAs (C1/C2) get priority rotation in Wave 1. Paid always runs the Learn More button.

| # | CTA clip | Destination |

|---|---|---|

| C1 | "The free manifesto is the playbook — grab it" | AI-First manifesto LP (the funnel already producing leads) |

| C2 | "Get the 805-site leak report, free" | report LP |

| C3 | "Apply — takes two minutes" | start.stpierre.ai application (Typeform j4wHbfWD) |

| C4 | "Grab the 7-Ad Blueprint, free" | blueprint.stpierre.ai |

| C5 | "Steal the kit for your niche / run your numbers" | /resources kits + calculator |

| C6 | "Comment DEMO and I'll build it for your company, free" | organic → GHL comment→DM rail |

| C7 | "Follow the bet" (soft, organic-only) | profile follow |

Batch 0 — The 10 scripted reels, realigned to AI-First

The Jul-16 chatbot-first scripts stay the shooting spine (9 talking-heads in one sitting + reel 8 screen-record VO), with the campaign thesis upgraded: the booking leak is chapter one of becoming the AI-First company, and the free demo bot is your first AI-First install — built with Claude. CTA stays comment DEMO. Per-reel adjustments:

1. 1. The Manifesto — becomes the AI-First manifesto reel: leak → "none of the five systems matter until chapter one" → "this is the first piece of the AI-First install. Comment DEMO."
2. 2. Not a Calendar Problem — unchanged body; close: "the qualified-booking bot is what AI-First looks like at the front door."
3. 3. The 8:47 PM Homeowner — unchanged; close names the machine: "an AI-First company never sleeps on a wet basement."
4. 4. Phone tag — unchanged; close: "AI-First companies don't play phone tag."
5. 5. Tire-kickers objection — unchanged; the qualification bot = AI behind the glass, on brand.
6. 6. Lead that waits an hour — unchanged; "speed is the first thing AI-First buys you."
7. 7. The audit leak — bridge to the 805 report ("I screened 805 sites — same leak").
8. 8. 11 PM booking screen-record — reframe VO per B5: "day one of an AI-First install."
9. 9. One number was a lie — bridge: "that lie is why I rebuilt the engine AI-First with Claude."
10. 10. My insane deal — already the c4/c5 offer; say it verbatim to the current offer language.

Batch D — Cancun b-roll

Laptop working shots (terrace/café, neutral not resort-flaunt), walking backdrops for hook variety, phone-call b-roll, water/sunrise ambient for pattern interrupts. Tone check: The Bet runs on "3 weeks of runway" — keep backgrounds neutral or address it head-on as a bet-update beat. Decide before posting anything beach-forward.

Production rules (canon)

• - 9:16, same mic and room tone across ALL batches — stitch seams must be invisible. 2+ takes, slate by ID, vary outfit/location per family block.
• - Every hook names the niche (hook contract): "foundation & waterproofing" now; niche-swap re-records for winners later.
• - "with Claude", never "with AI" for the build attribution ("AI-First" as the category term is canon). Codex never named. Handle only @alexstpierre. Company = stpierre.ai.
• - Plain English ≤5th grade except the verbatim jargon lines. Blunt operator tone: stakes, never desperation.

Before you fly

1. 1. Register the ★ hooks (8–12, 23, 28, 40, 41, 42) in sp_creative_angles.
2. 2. Screen-record the demo bot flow (B5 + reel 8).
3. 3. Decide the 10 disapproved VSL appeals — video cells are dark until then, and this shoot's output lands in the same policy lane.
4. 4. Print/teleprompter deck grouped by outfit/location block, AI-First block first.