Field doctrine · #ad-creative-autopsy

Strategy With
Fingers On
The Keyboard

Media buying is not dead. Creative strategy is not dead. Meta ads are not dead. What is dying is the layer that only manages the work. The next 6–12 months belong to people who can set the strategy and ship the cut — offer, creative, and landing — in the same week.

ICnot people-manager
5skills that compound
1weekly autopsy ritual
Working definition
2–3× Output of a strategist who executes with AI
0 Value in prompt libraries alone
Full Funnel owned: offer → ads → LP
Kill 80% of AI output on taste

The lethal hire is not the person who writes the best brief. It is the person who writes the brief, generates the variants, kills the weak ones, and lands the page change before the week ends.

DissolveMiddle management
KeepJunior ICs + mentors
OwnOffer + creative + LP
DropPrompt theater
RaiseJudgment rate
The shift

The wrong prediction was “juniors vanish”

The early AI narrative said entry-level creative roles would disappear first. The sharper read — from how agencies and brands are actually reorganizing — is that management seats dissolve first. Execution-capable ICs get paid more.

01
Agencies were built for oversight

Classic org charts promote people out of the work into workflow management: QA, status, client narration. That layer made sense when reporting ate a week and briefs moved through five hands. Automations now cover large parts of review volume, reporting, and project theater. Oversight without output is the first role to lose budget.

02
Law-firm logic arrives late

Other professional services already price people who can think and produce. Agencies are catching up: the premium is on strategists who stay fingers-on-keyboard. Mentorship of juniors still matters — but as craft apprenticeship, not as a permanent middle seat between founders and doers.

03
Reverse the “creative lead” hire

A common mistake in 2025–26: hire a creative-strategy leader to manage the strategists. The better move when automations already multiply review capacity: hire more junior and mid ICs, tighten mentor contact, and let automation absorb the managerial hours that used to justify the lead seat.

04
Salaries follow lethality

Highest packages go to people who can do the work of two or three: strategy plus execution with AI. That is not “knows ChatGPT.” That is ships acquisition outcomes without a handoff queue.

Expanded scope

What the IC creative strategist actually owns

At the top of the craft, the job is no longer “write briefs and QA edits.” AI collapsed the distance between ideation and shipped surface. The strategist who waits for a separate growth lead, a separate LP owner, and a separate editor will lose to the one who owns the loop.

Old brief

  • Messaging angles and hook hypotheses
  • Creative brief to production
  • QA before client or brand review
  • Comment on performance after the fact

2026 brief

  • Which offer / SKU / bundle is actually scalable
  • Hook matrix + variant generation killed by taste
  • Landing experience that matches the promise in the first 3s
  • Same-week iteration from Ad Library survivors and account data
Operator translation

If your title still says “creative strategist” but you cannot change the offer framing, ship a LP section, and kill a hook set in one sitting, you are practicing a 2022 job description inside a 2026 cost structure. The market will not wait for your org chart to catch up.

EX
Systems ownership · personality → SKU → retail
One cut that owns the loop
Why this is IC work

You are watching a host, then a product, then a purchase path — sequenced as one system. That is the expanded brief: not “make a nice spot,” but hold attention long enough that offer + shelf + brand identity stay attached. Steal the sequencing grammar at 1/1000th the budget; do not wait for a separate “brand film” team to invent it.

Reference cut — MrBeast × Feastables · personality-led retail stunt

Host first, SKU second, reward framing throughout. Partnership Ads and UGC try to clone this energy at Meta scale — handheld chaos, direct camera, identity before inventory.

Also mine live Meta creatives · Ad Library → Feastables
Skill stack

Five skills that beat “being good at AI”

Prompt libraries are a comfort object. Modern models are conversational. The scarce capacity is driving them — deciding what good looks like before the machine finishes a sentence. Systems thinking keeps showing up in platform leadership language for a reason: creative is not a isolated file; it is a node in acquisition. Each skill below has a public reference cut — watch it with the pass/fail test open.

01

Creating

You specify the visual, the first-frame conflict, the line that must land before second three — without waiting for the model to invent the brief. AI accelerates implementation. You own the hypothesis. If every “idea” originates in the chat, you are interchangeable with anyone who pastes the same account context.

Pass

You can name the hook before opening a tool.

Fail

“Make me 20 hooks about our product” is your entire process.

02

Curating & editing

Taste is not vibes. It is selection under constraint: what attaches to your name, what survives the kill pass, what elevation of finish you refuse to ship. Volume without curation is how accounts drown in near-duplicates and train the algorithm on mush.

Pass

You delete more than you post, and can say why in one sentence.

Fail

Every AI variant goes live “to let the algorithm decide.”

03

Articulation

Working with models is speech, not spellcasting. Audience state, promise, proof, and success metric must be sayable out loud. Vague prompts produce polished nonsense that looks done and converts like a brochure.

Pass

You can brief a stranger in 45 seconds without slides.

Fail

Your brief is a moodboard and “make it scroll-stopping.”

04

Judgment

The room belongs to people who decide. Approve, reject, or revise — on the spot. Separate the 80% slop from the 20% worth editing into something excellent. Hemming and hawing is now an expensive latency tax.

Pass

You make a kill/keep call in under a minute per asset.

Fail

Everything waits for a group review “next week.”

05

Systems thinking

Change the hook and you change who the algorithm invites. Change the offer and the LP must keep the promise. Change reporting cadence and mentoring capacity moves. Creative strategists who only optimize the file in front of them miss second-order effects across acquisition.

Pass

You map A→B→C before shipping a “simple” creative test.

Fail

Creative, media, and site are three tickets owned by three moods.

Reference cuts · skill → evidence

Watch these with the kill pass open

Same embeds pattern as our formats guide — public reference films, not screenshots of a talk. Use Ad Library links to see what is still running paid today.

01
Creating · name the enemy before second five
Hypothesis before production
Reference — Dollar Shave Club · “Our Blades Are F***ing Great” (2012)

Enemy named in the open, CEO as ally, offer clear. ~$4.5k shoot, 12k subscribers in 48 hours. The IC move: decide the conflict before anyone opens CapCut or a chat.

02
Curating · refuse safe beverage advertising
Taste as selection under constraint
Reference — Liquid Death · Big Game · absurdism as brand filter

Every objection preempted with humour before the viewer weaponises it. That is curation: a long list of “normal water ads” killed so one voice stays attached to the brand.

Live Meta · Ad Library → Liquid Death · many UGC taste-tests run 30+ days
03
Articulation · odd pacing that still says one thing
Say the problem out loud
Reference — Oatly · Toni TV · pattern interruption with a clear brief

Strange on purpose — but the brief is speakable: interrupt, then product logic. If you cannot restate that in one breath, you cannot supervise an LLM or a junior editor.

04
Judgment · which hooks clear the bar before you scale
Kill rate is the skill
Reference — Rapid hook testing · operator breakdown of what actually wins

Mental model for Micro-POV batches: ship dozens of two-line angles, read Hook Rate + CTR, rebuild winners as higher-fidelity cuts. Judgment is the filter between volume and spend.

05
Systems · caption / dwell / algorithm as one machine
Creative is a node, not a file
Reference — Caption-heavy vertical edits · “read the screen” mechanic

Editing grammar that forces linger and loop — the same behaviours paid social treats as high intent. Pair with mined customer language on low-distraction B-roll; LP copy must keep the same promise the captions sell.

Tooling without theater

Abandon the prompt museum

Side-by-side conversational workflows beat static prompt libraries. The model is a junior you talk to, not a vending machine you feed sacred strings. Spreadsheets, research docs, and status artifacts that used to require a specialist are now speech → structure. That does not make you “technical.” It removes the old tax that kept creative people out of their own data.

Expect periodic spiral — tools jump, defaults shift, yesterday’s workflow feels obsolete. The response is not to freeze. The response is to re-articulate your craft in the new interface and keep deciding.

A

Personal on-ramp

If work on-ramps feel dead, start with a personal “board of directors”: roles you’d hire for yourself. Turn each into an agent/workflow. Habit of articulation transfers to client work faster than another SaaS tutorial.

B

Craft made explicit

Speak your research process until it becomes a checklist the model can instantiate in Notion or Sheets. “Shooting in the dark” becomes “opinions with receipts” when the path is verbalized once.

C

Human edges kept

Keep at least one surface deliberately human if that is your distribution edge — writing, voice, or taste calls that must not sound like everyone else’s completions.

D

Adoption is not optional

Environmental and labor concerns about AI are real. Using them as a permanent permission slip to not learn the interface accelerates a K-shaped split: adopters compound, hesitators wait for clarity that will not arrive. The humane move is approachable on-ramps — not purity freezes.

#ad-creative-autopsy

Weekly ritual: tear one cut open

This channel is not for dunking on brands. It is for training judgment. Pick one live Ad Library survivor (30+ days running is the cheap proxy) or one of your own winners. Write three blocks only.

First 3s promise

What does the opening frame + line claim the viewer will get? Who is it for? What emotion is assumed already present?

What the cut delivers

Proof, demo, offer, identity — what actually arrives after the hook? Where does it keep or break the promise?

What you would reshoot

One change only: first frame, line, proof order, or CTA. Not a redesign. A surgical retest.

Map the five skills onto the cut

Creating: what hypothesis did they bet on? Curating: what did they refuse to include? Articulation: can you restate the brief in one breath? Judgment: would you kill or scale it today? Systems: what LP / offer / audience must be true for this cut to work?

DEMO
Worked autopsy · three lines only
Liquid Death · transparency through absurdism
Same cut as skill 02 — now run the Slack ritual on it

First 3s promise

You will not get a polite beverage ad. The brand claims permission to be loud, weird, and anti-“healthy water” cliché — immediately.

What the cut delivers

Objections (taste, positioning, premium scepticism) get laughed down before they harden. Product logic rides inside the joke, not after a soft brand film.

What we’d reshoot

One surgical retest for Meta: open on a single UGC face tasting the can, keep the absurd punchline as end-card — same thesis, cheaper first-frame for feed.

Post this shape in Slack · then swap the brand · pick a 30+ day survivor in Ad Library
30-day implementation

From title to lethal IC

Four phases. No certification. Evidence is shipped work.

I Week 1

Scope audit

Own the loop
  • List every handoff between brief, edit, LP, and offer
  • Mark which handoffs are latency vs actual skill gaps
  • Pick one product/offer you will own end-to-end this month
II Week 2

Judgment reps

Autopsy ×5
  • Five Ad Library survivors through the 3-block ritual
  • Force good/bad + why on each — no hedging
  • Post one in #ad-creative-autopsy without product CTA
III Week 3

Craft system

Verbalize
  • Dictate your research → brief → kill → ship process
  • Instantiate it as a checklist the model can refill
  • Define “excellence vs slop” for your category in writing
IV Week 4

Economic hedge

You as a business
  • Package one offer: audit, static bundle, or retainer slice
  • As management seats thin, IC freelancers fill demand
  • Safest path is also the highest-agency path
First principles

What this doctrine assumes

01

Creative is targeting

Meta reads faces, language, and structure as audience signals. The IC who owns creative owns a large share of who enters the account — whether media buys “broad” or not.

02

Latency is a CPA tax

Every day a decision waits for a manager is a day competitors rotate. Judgment speed is a performance lever, not a soft skill.

03

Messaging outlives format

Formats fatigue. The line that names the pain does not. Treat converting language as IP and redeploy it across formats before you retire the idea with the edit.

04

Survivors are paid research

Ads running 30–90+ days in the Ad Library are someone else’s validated concept. Autopsy structure — do not clone skin.

05

Excellence is a brand

In a world of infinite near-duplicates, being associated with excellent work is a decade-long differentiator. Define the bar; refuse to ship under it.

06

Opt-out is a career bet

Refusing the interface is allowed. It is also a wager that your current seat outlasts the reorganization. Price that wager honestly.