Brez Fulfillment Core — AI Arbitrage, Copy, Creative, Design, Targeting (Lessons 3.7–3.10)

3.7 The AI Arbitrage System (course's core mechanism)

The arbitrage: traditional agency = 15-20 hrs/client → caps at 3-5 clients ($6-10K/mo ceiling). AI-agent system = 3-5 hrs/client → 10-15 clients ($20-30K/mo). Pitch line: "Same price to the client. Same results. 4x less work."

Five named agents (roles, not magic — each is a persona'd workflow):

• Aurelius — launches campaigns; 50+ copy variations across PAS/AIDA/BAB/Story
• Vortex — creative diversity review; flags duplicate angles + weak format variety
• Nexus — QA: ads approved, pixel firing, placements correct, Ad Library competitor research
• Zenith — weekly client reports from raw Ads Manager data in ~2 min
• Stratos — performance diagnosis + data-driven optimization recs

7-day client launch (4-5 hrs total): D1 onboard + asset collection (90m) → D2 competitor research + 20+ copy variations (60m) → D3 15-25 creatives + diversity check (60m) → D4 build campaign + QA audit (45m) → D5 client approval (30m) → D6 launch + verify live (10m) → D7 first data + diagnosis (15m).

Performance diagnostics table:

Golden rule: "When in doubt, add more diverse creatives. 90% of performance problems in the Andromeda era are creative problems."

Why it works (Andromeda, Dec 2024): 10,000x more ad candidates per auction, +22% ROAS with Advantage+, 100x faster processing. Structure shift: 5+ campaigns × 3-6 creatives → 1-2 campaigns × 15-25+ diverse creatives. "Creative IS the new targeting."

Prospect-facing sales language (verbatim):

"We test 50+ ad variations versus traditional agencies testing 5-10. We find what works faster, lowering your cost per lead and enabling faster scaling. You get big-agency quality without overhead."

🎯 Rio: You already RUN this arbitrage (Claude Code + skills = the five agents). The NEW asset is the positioning language: "50+ variations vs their 5-10" and "same price, 4x less work" are sales lines for Uprise pitches. The diagnostics table is a clean triage layer on top of your KILL/WATCH/SCALE rules — those tell you WHEN to act, this tells you WHAT's broken. Steal the 7-day timeline as a client-facing onboarding promise.

3.8 AI Ad Copy Generator

Core principle: One Ad = One Reason to Buy. Not headline variations — each ad targets a completely different fear/desire/angle. Example matrix (pizza, 12 angles): exhaustion, family time, value, authenticity, speed, social proof, new-in-area, occasions, health, simplicity, late-night, urgency.

Frameworks: PAS (pain audiences), AIDA (offers/promos), BAB (transformations), STORY (emotional/UGC). Output: 12 ads × 3 lengths — short <50 words (Feed/Story), medium 50-100 ("See More"), long 100+ (high-intent storytelling).

Mega prompt inputs: business name, what they sell, target customer, USP, current offer, tone, platform.

Six AI anti-patterns to kill: (1) "Not X — it's Y" contrast, (2) triple structure "X. Y. And Z.", (3) "Imagine this" opener, (4) filler words (elevate, transform, seamless, game-changer), (5) rhetorical question-then-answer, (6) perfect-grammar robot (add contractions/slang). Test: read every ad out loud — if you'd cringe hearing a real person say it, rewrite.

60-second edit checklist per ad: read aloud → check hook → kill AI words → add specifics ("$12.99" not "affordable") → match brand voice. 12 ads = 12 min.

Swipe file categories: scroll-stopping formats, hooks that worked, competitor ads running 60+ days (= proven winners), video structures.

🎯 Rio: Your 4-master system (Schwartz/Georgi/Halbert/Bencivenga) + Ralph Wiggum 8-pass QA is strictly stronger than this — don't adopt their generation method. The anti-pattern list overlaps ~90% with your QA passes and no-em-dash rule. What IS new: the one-reason-per-ad angle matrix discipline — before writing, list 10-12 distinct buying motivations for the niche, then one ad per motivation. That's your Schwartz mass-desire work operationalized as a creative-volume plan (feeds the 15-25 creative quota from 3.7). Also steal: 60+ days in Ad Library = proven winner as a swipe filter.

3.9 AI Creative Builder

2-Second Rule: if they can't get it in 2 seconds, they scroll. Scroll-stoppers: human faces (eyes), bold contrast, ONE message, real photos, numbers/prices, before/after, movement. Killers: stock photos, text clutter, low contrast, big logos, ads that look like ads.

Six formats — every client needs 4+ of 6:

1. Single image (the workhorse) 2. Carousel (2-10 cards, story/multi-product) 3. Video 15-30s (most frequent outperformer) 4. UGC-style (phone-filmed look, NOT polished) 5. Before/After (split-screen — "extremely powerful for services") 6. Meme/relatable (shares + saves).

Launch minimum: 15+ creatives across 4+ formats. Export every creative in BOTH 1080×1080 (feed) and 1080×1920 (Stories/Reels).

45-min build: pull angles (5m) → review swipe file (5m) → 8-10 statics (20m) → 3-5 videos (15m) → diversity check (5m). Course tools: Canva, CapCut, Claude, Gemini Nano Banana.

🎯 Rio: Your HTML→screenshot pipeline + Higgsfield beats their Canva/CapCut stack — keep yours. The NEW operational rules: (1) the 4-of-6 format quota — you default to polished statics; UGC-style, meme, and video are gaps Higgsfield covers, (2) the 15+ creative launch minimum slots directly into your ABO test structure, (3) dual-export 1:1 AND 9:16 for everything — bake into the screenshot script. Before/After is flagged as the killer format for services = pools/cleaning (see authentic before/after memory note — real screenshots, not AI fakes).

3.9b AI Ad Creative Design

Workflow (sub-10 min/ad): Google Stitch (plain-English → full ad design, ~60s generation) → iterate ("make text bigger", "add before/after split", "warmer colors") → Canva polish (logo, brand color, mobile preview, PNG). Prep first: what client does, the offer, audience, logo + brand color.

Five universal ad angles: Direct Offer / Before-After / Social Proof (stars + review quotes) / Problem Call-Out / Urgency-Seasonal. Test 3 ads on different angles.

Niche prompt templates (verbatim — adapt for Higgsfield):

Roofing: "Design a Facebook ad for a roofing company in [CITY]. Show a beautiful home with a brand new roof, bright natural lighting, suburban neighborhood. Text says "Free Roof Inspection" in bold white letters. Dark blue brand color. Professional, clean, trustworthy feel. 1080x1080."

Dental: "Design a Facebook ad for a dental clinic in [CITY]. Show a bright, modern dental office with a patient smiling confidently. Text says "When Was Your Last Cleaning?" in clean white font. Light teal brand color. Warm, welcoming, professional. 1080x1080."

HVAC: "Design a Facebook ad for an HVAC company in [CITY]. Show a comfortable family relaxing at home on a hot day, cool and happy. Text says "$49 AC Tune-Up — Stay Cool This Summer" in bold. Brand color: [COLOR]. Clean, modern layout. 1080x1080."

Plumbing: "Design a Facebook ad for a plumbing company in [CITY]. Show a clean, modern kitchen with no water damage, everything working perfectly. Text says "Don't Wait for the Flood — Call Now" in bold. Brand color: [COLOR]. Professional and urgent feel. 1080x1080."

Gym: "Design a Facebook ad for a local gym in [CITY]. Show an energetic group fitness class with diverse people working out, bright lighting, positive energy. Text says "First Week Free — No Catch" in bold white. Brand color: [COLOR]. Motivating and welcoming. 1080x1080."

Restaurant: "Design a Facebook ad for a restaurant in [CITY]. Show a beautifully plated dish, close-up food photography, warm ambient lighting, inviting atmosphere. Text says "Tonight's Special Won't Last" in elegant white font. Brand color: [COLOR]. Premium and appetizing. 1080x1080."

Salon: "Design a Facebook ad for a hair salon in [CITY]. Show a stylish, modern salon interior with a client looking confident after a fresh cut. Text says "Walk In Good. Walk Out Great." in bold. Brand color: [COLOR]. Trendy and inviting. 1080x1080."

Landscaping: "Design a Facebook ad for a landscaping company in [CITY]. Show a beautiful, manicured front yard with green grass, trimmed hedges, and flowers. Before-and-after split layout. Text says "Your Yard Could This" [sic] in bold. Brand color: [COLOR]. 1080x1080."

Design rules: max 5-8 words of on-image text · no fake stock photography · always include logo + brand color · ALWAYS preview at mobile size (95% of viewers) · health before/afters need "Results may vary" per Meta policy.

🎯 Rio: Overlaps heavily with your niche templates (7 niches) + HTML→screenshot + Higgsfield — you don't need Stitch. The keepers: the template anatomy (scene + offer text verbatim + brand color + feel + dimensions) is a tight prompt spec for Higgsfield image gen; the 5-8 word on-image text cap and mobile-size preview check are hard QA rules worth adding to your creative checklist. No pools/cleaning template here — write one in this anatomy for the first client.

3.10 Targeting in the Andromeda Era

Doctrine: "The algorithm finds your buyers through your creatives, not your targeting settings." Interest stacking, lookalikes (1/3/5%), behavior stacking, detailed targeting expansion, saved old audiences = all obsolete. Old way: 2-3 hrs of audience building per client. New way: 5 minutes, better performance (+22% ROAS with Advantage+ broad; Andromeda evaluates 10,000x more candidates across 3.1B users).

Ad set config (the whole thing):

1. Location: pin at business address, "People living in or recently in this location"
2. Radius by business type: restaurant 5-10mi · dentist/doctor 10-15mi · gym 5-8mi · real estate 15-25mi · home services 10-20mi · e-com/coaching = entire country
3. Age range only if contextually necessary
4. Interests/behaviors: leave blank
5. Retargeting: only after 2+ weeks of traffic

Retargeting WORTH building: site visitors 30-90d, 50%+ video viewers, customer email lists, IG/FB engagers 30d, incomplete lead-form openers. NOT worth: interests, lookalikes, behavior stacks.

Nexus audience research prompt (verbatim — research feeds COPY, not targeting):

"I'm running Facebook/Instagram ads for a local business. I need to deeply understand their target audience — NOT for targeting settings (I'm using broad targeting), but to write better ad copy and create better visuals. BUSINESS: [Business name] WHAT THEY SELL: [Product/service] LOCATION: [City, state] PRICE RANGE: [Average ticket] Give me: 1. The top 10 REASONS someone would buy this (not features — emotional motivations) 2. The top 5 OBJECTIONS they'd have before buying 3. The top 5 MOMENTS when they'd think about this product (triggers) 4. What language/slang does this audience actually use? 5. Who are they comparing this business to? (competitors + alternatives) 6. What does their day look like? (daily routine relevant to the product) Be specific. Use real examples. No generic marketing talk."

🎯 Rio: This is the missing layer UNDER your campaign-structure playbook — the ABO→CBO system tells you structure, this tells you what goes in each ad set: pin + radius + maybe age + nothing else. For the pools/cleaning client: home services = 10-20 mile radius, blank interests, retargeting from week 3. The Nexus prompt is redundant with your Schwartz market-awareness research BUT its 6-question shape (reasons/objections/triggers/slang/comparisons/daily routine) is a fast intake form — run it per client, pipe output into the 4-master system and the angle matrix from 3.8.

🔍 Fidelity-audit additions (7/6 overnight — recovered from the verbatim archive)

3.7 — sales framing + timing rules the summary dropped:

• ⚠️ Never say "AI does everything." Clients want a human managing their money. The right framing: "I use AI tools to do MORE for you, FASTER." Not: "A robot handles your ads." (Guardrail on the 50-vs-5-10 pitch line.)
• Day 7 = look, don't touch. First REAL optimization = Day 10. Expected Day-7 diagnosis: "Learning phase active. Normal volatility. Wait."
• Post-launch load math: 15 min/day per client → at 10 clients = 2.5 hrs/day. (The retainer-scalability number for capacity planning.)
• Nexus launch-check has two items the note's QA list missed: ✅ Audience Network excluded · ⚠️ flags Advantage+ creative features ON — review if intended (pairs with the toggles-OFF doctrine).
• Zenith weekly-report anatomy (the template shape, ~2 min to produce): 6 headline metrics each with week-over-week delta (Spend/Impressions/CPM/CTR/Leads/CPL vs target) → Top 3 performers (each tagged angle + format + framework + CPL) → Bottom 3 kill list (same tags) → next-week recommendations in 5 verbs: Scale / Create / Kill / Replace / Monitor. Feed this shape into Command's report generator — the angle+format+framework tagging per ad is what makes the report diagnostic instead of just numbers.

3.8 — one habit line: swipe-file collection is scheduled, not ad hoc — search the client's niche in Ad Library weekly, filter "longest running," screenshot winners. For hooks, "save the pattern, not the exact words."

3.9 — the niche creative playbooks (what to shoot / style / top formats — absent from the note):

• Static mix inside the 8-10 Canva builds: 4 single images, 2 carousels, 2 before/after. Video scripts = hook → body → CTA, 15s.
• Nano Banana 2 sketch-to-design (free, Google AI Studio): sketch the ad on paper → photo → professional design. The no-photos-no-designer client unlock; Higgsfield equivalent = describe the layout in words, but the sketch route is faster for spatial ideas.

3.9b — the hooks quick-reference (2 niches + hook lines the note missed):

• Auto repair: clean shop, diagnostic screen, under hood · hook: "Check engine light? We'll check it free."
• Real estate: property tour, neighborhood aerial, staging · hook: "Just listed in [neighborhood]"
• Roofing hook variant not in the 8 prompts: "Your neighbors already called us"
• Angle→niche mapping for the 5 universal angles: Before/After = roofing, dental, landscaping, cleaning, renovation, salon · Problem Call-Out = HVAC, plumbing, auto repair, pest control · Direct Offer + Urgency/Seasonal = every niche · Social Proof = any business with real reviews.
• ✏️ Correction: the landscaping prompt text in the archive reads "Your Yard Could Look Like This" — the "[sic] Your Yard Could This" in the note above was a capture typo, not the course's.

3.10 — the research→creative mapping (the operational bridge): each of the 10 "reasons to buy" = an ad angle for the matrix · each objection = a PAS or BAB ad · each trigger moment = a hook. Age-range examples: bar = 21+, retirement services = 55+, everything else stays 18-65+.