← back to index

The Workflow / The Cold Start

How the brief became 40 scripts

Sync to Submission

47min 40 scripts

One call · About two days · Eight creators · Nine products

40
Scripts Delivered
6
Ad Formats Used
8
Creators
~2d
Build Time
72
Ad Ideas Mapped
9
Products
11
Small Tools Used
1
Working Brief

The first 24 scripts were submitted inside 48 hours. The program expanded to 40 scripts across the full brief.

The Thesis

What This Case Study Is About

A cold handoff became a working brief.

A 47-minute call, several months of account data, and almost no handoff became 40 regulated scripts in about two days. The brief covered eight creators and nine finance products with different claims, audiences, and legal requirements.

Client LendingTree / Role Creative Strategist / Approach AI-assisted, human-checked

Stage 01 / Score

Four Questions

I score the buyer before I write a word.

I start with four questions. The answers tell me what kind of ad fits the buyer and what the script needs to explain first.

How expensive is it?

Under $50
Over $300
110

How skeptical is the buyer?

Easy trust
Prove it
110

Feeling or facts?

All feeling
All facts
110

How long do they research?

Impulse buy
Weeks
110

What the score tells me

4 to 16

Tell a story

Buyer decides on feeling. Let them feel understood before the product shows up. Rushing it kills it.

17 to 28

Mix both

Short story, then proof. Could go either way. Let the numbers tell us which works.

29 to 40

Lead with proof

Buyer is skeptical from second one. Product in frame one. Data right after. No warmup.

Stage 02 / Map

Ad Types × Buyer Problems

60+ ad ideas mapped before I write a word.

Ad types across the top. Buyer problems down the side. Every square is an ad I could make. One map per client, good for 3 to 6 months. No blank page. No guessing what to write next.

Rate
Anxiety
Approval
Fear
Comparison
Paralysis
Monthly
Budget
Debt
Shame
Podcast talking head -
Whiteboard explainer - -
Side-by-side compare - - -
Founder to camera - - -
Before / after - - -

Sample shown. The first map held 72 viable ideas across six ad types, four creator types, and three products. I shipped 24 in the first batch; the full program expanded to 40 scripts.

Why it works

01

No duplicate ads.

The map forces every ad to come from a different angle. No two ads feel the same. Meta rewards this with more spend.

02

No blank page.

The map replaces decision fatigue. Pick a square, write the ad. Output scales without the blank-page bottleneck.

03

Gets smarter over time.

Every winning square tells me what works. Every loser tells me what to stop testing. The map sharpens.

Stage 03 / Compound

The Learning Loop

Every result improves the next brief.

I save the useful parts of each project: what performed, what legal changed, what creators could execute, and what broke in production. The next brief starts with that record instead of a blank page.

01

Find

Winning ads, competitor moves, legal feedback, performance data.

02

Analyze

Why it worked. What made it different. What's worth keeping.

03

Save

Wins become rules. Failures become guardrails. All of it gets filed.

04

Use

Every new brief pulls from the whole library. Not just the last job.

05

Learn

Results come back in. The library updates. Next brief is sharper.

A deodorant ad taught me how to write whiteboard ads for fintech.

Different products. Same buyer psychology. That's what the system catches.

My AI Stack

Three tools. Same playbook. Not locked into any one platform.

Claude

Where the briefs get built. 11 custom workflows loaded. Handles scripts, hooks, QA, and formatting.

Perplexity

Where I research. Competitor ads, real-time trends, source checking.

Manus

Where complex tasks run. Multi-step research and doc generation in one shot.

Stage 04 / Build

11 Workflows Loaded

Eleven small tools, each built for one job.

Together they turn the account read into scripts, shot lists, text overlays, compliance checks, and editor notes. The first pass gets faster; the final call still stays with me.

Brief builder

Writes full ad scripts with shot lists and creator notes.

Testimonial builder

Turns a call transcript into a ready-to-film brief.

Ads strategy

Knows how Meta's algorithm picks winners right now.

Hook selector

Picks the right text overlay so it works with audio off.

Production planner

Maps out what to make this month and what to skip.

Research extractor

Turns any winning ad into a reusable lesson.

QA checker

Catches math errors and compliance gaps before submit.

Brief QA

8-point checklist before anything ships to a creator.

Writing filter

Makes sure nothing reads like it was written by AI.

Design system

Builds the branded PDFs and graphics in my case studies and decks.

My context

Knows my voice, my clients, and my preferences.

Stage 05 / Decide

AI Drafts, Human Judges

AI drafts. I decide.

The AI is fast. The judgment is mine. Every script the system drafts goes through me before a creator sees it. Nothing ships until a human decides it's right.

The AI handles

  • • Writing first drafts across the batch
  • • Tagging each script by buyer and product
  • • Assigning creators and building shot lists
  • • Writing the on-screen text overlays
  • • Generating editor notes and CTA copy

I handle

  • • Rewriting every hook that sounds generic
  • • Making sure no two ads feel the same
  • • Checking every number in every math ad
  • • Making sure nothing breaks compliance
  • • Writing the brand context and prod notes

Why this matters for clients

The tools speed up the first pass. I still check every claim, number, creator assignment, and edit note before the brief leaves my hands.

That judgment now comes from 225 ads across 11 client accounts.

The Stress Test

LendingTree · Fintech · $10M+ Meta

Here's what it looks like under pressure.

I took over with almost no handoff: a 47-minute call, the onboarding form, and three months of ad data. The brief was due that week while several other accounts were still moving.

Finance work slows down when claims, disclaimers, and product details are left until the end. I built those rules into the brief before the first script was written.

Three Moves

01

Double down.

One ad format was already eating 50% of their budget. I didn't try to replace it. I wrote 5 new hooks for the same format so Meta treats each as fresh.

02

Own the gap.

Nobody in financial services is running math-heavy whiteboard ads. Their data buyer wants exactly that. Easy lane.

03

Legal-proof it.

Financial services legal review kills timelines. I built every disclaimer into the first draft so there's nothing to push back on.

Stage 01 in action · Both buyers scored

The Researcher

36/40 Lead with proof

Home equity is a $50K+ decision. They research for weeks. They want the math, not the story. Whiteboard ads were built for this buyer.

The Impulse Buyer

20/40 Mix both

Wants a simple monthly payment and no hassle. Fears complexity. Could go either way. Test both and let the data pick.

What happened in order.

01

Listened to the call.

47-minute call. Pulled out their buyer types, product lines, compliance rules, and tone preferences. Fed straight into my context workflow.

02

Scored both buyers.

Researcher 36/40. Impulse buyer 20/40. Two completely different ads needed, not just two different hooks. Stage 01 of the workflow.

03

Read the data.

Three months of ad spend reviewed. Found their one winning format and the formats nobody was testing.

04

Checked competitors.

NerdWallet, SoFi, Figure. None of them running math-heavy ads. Open lane confirmed.

05

Mapped every possible ad.

Built the full map of ad type vs buyer problem before writing a single script. 72 viable ideas. Stage 02 of the workflow.

06

Wrote and checked the first 24 scripts.

Full scripts with shot lists, creator notes, and compliance language went in inside 48 hours. The program later expanded to 40 scripts across the full brief.

The Result

The first 24 cleared creative review.

Legal notes focused on disclaimers. The brief then grew to 40 scripts without rebuilding the process.

Across Industries

The System Holds

Two different stress tests.

LendingTree was the stress test for speed and compliance. Withings was the stress test for scaled performance. Same workflow, different pressure.

Health Tech

Withings · $100M+ revenue

Peak ROAS at scale 4.08x
CPA reduction 69%
$359 → $112

In the recorded non-promotional window, this UGC program produced the US full-price purchases at scale.

Read the full Withings case study →

Fintech

LendingTree · $1B+ public

Scripts / build time 40 / ~2 days
Creators / products 8 / 9
One working brief

A cold handoff became 40 regulated scripts with creator and editor instructions kept separate.

Same process, adapted to different account data and constraints.

Now Taking Clients

Want this system on your account?

Full creative strategy, competitive audit, and 15 to 30 finished scripts every month. Performance diagnostics on what's working, what's fatiguing, what to test next. A system that gets sharper with every campaign, not one that resets.

Book a Discovery Call