KLIK/PROCESS/2026
▶ HOW THE PIPELINE RUNS

Eighty ad concepts a week doesn't come from working harder. It comes from rebuilding the loop. Four steps run every week — and at step three, you choose: we click the button, or you do.

Four steps.
Two [endings].

WEEK 0
01
DISCOVER
WEEK 1
02
BRIEF
WEEK 2+ · WEEKLY
03
GENERATE · FORK
WEEKLY · ONGOING
04
SHIP & LEARN
// 01 · THE LOOP

One pipeline.
Two endings.

The same four modules run every week. The first two install once. The last two repeat forever — us shipping for you, or your team shipping for itself. We didn't invent a new process; we engineered the one your performance team has been asking for since the algorithm started eating creative.

Step 01 — Discover. Week zero. We tag every dollar of your last six months of spend, score it against 200 creative axes, and write the gap brief. You see what worked and what didn't, and why. One-time. Two weeks.

Step 02 — Brief. Week one. We turn the gap brief into a hypothesis tree: angles, hooks, formats, audiences. Ten copy agents iterate on each line until it scores 90+ against your brand voice. Output is a ranked weekly slate. One-time install of the framework, weekly refresh of the slate.

Step 03 — Generate. Week two onward, every Monday. The slate goes into image / video / motion prompt chains. Eighty concepts come out the other side, ranked and tagged. This is the fork. Path A: our studio runs it for you. Path B: your team runs it on the rails we installed.

Step 04 — Ship & Learn. Weekly. Concepts go live. Their tags come back in. Next week's slate is recut against what just won. The loop closes here. The longer it runs, the smarter the brief gets.

klik / pipeline.loop
  ┌──────────────────────────────────────┐
  │  01 DISCOVER    →  audit + tag map  │
  │       │                              │
  │       ▼                              │
  │  02 BRIEF       →  ranked slate     │
  │       │                              │
  │       ▼                              │
  │  03 GENERATE    →  80 concepts/wk   │
  │       │                              │
  │      ╱╲    ← fork                  │
  │     ╱  ╲                             │
  │  A·WE B·YOU                           │
  │     ╲  ╱                             │
  │      ╲╱                              │
  │       │                              │
  │       ▼                              │
  │  04 SHIP & LEARN →  tags back in    │
  │       │                              │
  │       └──────────► loops to 02       │
  └───────────────────────────────────────┘

  // install once: 01–02   //  loop weekly: 03–04
01
WEEK 0 · ONE-TIME · 7-10 BUSINESS DAYS

Discover

▶ TAG THE SPEND. FIND THE GAP.

Before we make anything, we read everything you've already made. Six months of spend, every variant, every audience, every result — scored on the same 200 axes the pipeline will later generate against.

// INPUT — FROM YOU
  • ›Read-only access to ad accounts
  • ›Brand kit · voice doc · existing system
  • ›3 top competitors to watch
  • ›One 45-min kickoff call
// PROCESS — ON US
  • ›Tag every ad on 200 creative axes
  • ›Cluster by performance × format × hook
  • ›Map the gap — what's missing, what's saturated
  • ›Score your voice against 3 reference brands
// OUTPUT — TO YOU
  • ›Tagged ad library · sortable
  • ›Gap brief · ranked opportunities
  • ›Voice profile · scored 0-100
  • ›Pipeline build proposal
★ DELIVERABLE·Audit doc + tag map + opportunity brief·SHARED IN NOTION · YOURS TO KEEP
02
WEEK 1 · INSTALL + WEEKLY REFRESH

Brief

▶ HYPOTHESIZE. SCORE. RANK.

A brief isn't a vibe. It's a hypothesis tree. We take the gap from step one and grow it into 40-60 testable angles, each with a hook, a format, and a target audience. Ten copy agents iterate on every line until it scores ninety against your voice. The ones that pass make the slate.

// INPUT — FROM STEP 01
  • ›Ranked gap opportunities
  • ›Locked voice profile
  • ›Your offer + promo calendar
  • ›Format mix target (static / motion / UGC)
// PROCESS — IN THE LAB
  • ›Branch each gap into 4-6 angles
  • ›Generate hook variants per angle (10×)
  • ›Multi-agent scoring on voice + clarity + CTR-proxy
  • ›Cull below 90; ladder the rest by predicted lift
// OUTPUT — EVERY MONDAY
  • ›Weekly slate · 40-60 ranked briefs
  • ›Each brief: hook · angle · format · audience
  • ›Predicted-lift score · 0-100
  • ›Notion drop · 9am Monday, on the dot
★ DELIVERABLE·Hypothesis-tree framework (install) + ranked slate (weekly)·SAME FRAMEWORK, BOTH PATHS
03
WEEK 2+ · WEEKLY · THIS IS WHERE YOU CHOOSE

Generate

▶ EIGHTY CONCEPTS. RANKED.

Every ranked brief goes into a generator chain — image prompts, motion stills, script-to-video, scene cuts. Eighty concepts come out the other side, scored and tagged, ready to upload. This is the only step where the two paths diverge.

// INPUT — MONDAY MORNING
  • ›Ranked slate from step 02
  • ›Brand asset library (logos, products, models)
  • ›Reference shots · style anchors
  • ›Format quotas per channel
// PROCESS — THE GENERATORS
  • ›Image · Midjourney + fal.ai (Flux, Nano Banana)
  • ›Motion · Runway + Higgsfield + Seedance
  • ›Audio · Suno · ElevenLabs · curated SFX
  • ›Edit · n8n · Adobe pipeline · Premiere templates
// OUTPUT — BY FRIDAY
  • ›80 ranked concepts · scored 0-100
  • ›Statics · motion stills · scripts · cuts
  • ›Tagged on the same 200 axes
  • ›Routed to your Drive · ready to upload
▶ PATH A · WE SHIP

Our studio runs it.

Our two-person team plus the model orchestra spends the week generating, editing, and approving. You get a slack ping when the slate's ready and a Loom walkthrough of the top picks. You upload, your buyer launches.

OUTPUT TO YOU · FINAL ASSETS IN YOUR DRIVE · NO TOOLS TO LEARN
▶ PATH B · YOU SHIP

Your team runs it.

We hand over the prompt chains, the scoring rubric, the n8n graphs, and the Claude skills. We train two of your people on the pipeline over four weeks. After that you generate your own slate every Monday — same engine, your hands on the wheel.

OUTPUT TO YOU · FULL SOURCE + 90-DAY SHADOW SUPPORT · NO LOCK-IN
★ DELIVERABLE·80 ranked concepts every week·SHIPPED BY US, OR BY YOU — SAME ENGINE
04
EVERY WEEK · ONGOING · THE LOOP CLOSES

Ship & Learn

▶ THE LOOP CLOSES.

Concepts go live. Performance tags come back in. Next Monday's slate is recut against what just won, what got skipped, and what no one swiped past. The longer it runs, the smarter the brief gets.

▶ PATH A · WE OWN THE LOOP

We close it for you.

  • ✓Slate uploaded to your ad accounts
  • ✓Daily performance pull · tagged + scored
  • ✓Friday review call · what won, what didn't
  • ✓Monday's slate reflects last week's signal
  • ✓Monthly performance read-out · written
Output: A self-sharpening creative engine you don't run. Roughly 320 concepts a month, each one informed by the last.
▶ PATH B · YOU OWN THE LOOP

You close it yourself.

  • ✓Your team uploads to your ad accounts
  • ✓We co-pilot the first 12 weekly cycles
  • ✓Bi-weekly office hours · 60 min · for 90 days
  • ✓Tag schema + scoring rubric · yours to evolve
  • ✓Quarterly engine tune-up · optional retainer
Output: A pipeline your team owns end to end. Source code, prompts, agents — all on your infrastructure. We're optional after day 90.
★ DELIVERABLE·The loop, closed — by us, or by you·WEEK ON WEEK · COMPOUNDING
// 06 · FIRST 30 DAYS

What the
first month
actually
looks like.

Both paths share the first two weeks. They diverge at day fifteen. By day thirty you've either received your first weekly slate (Path A) or run your first slate yourselves with us on the call (Path B).

// DAY
01
05
10
15
20
25
30
// SHARED · INSTALL
Kickoff + access
D1-D2
01 · Discover · audit
TAG + GAP BRIEF
02 · Brief · framework
HYPOTHESIS TREE
// PATH A · WE SHIP
03 · Generate · run 1
80 CONCEPTS · RUN 1
03 · Generate · run 2
80 CONCEPTS · RUN 2
04 · Ship + read-out
LIVE + LEARN
// PATH B · YOU SHIP
03 · Install + handover
PROMPT CHAINS + AGENTS
03 · Train your team
2 PPL · 4 SESSIONS
03 · You run · we co-pilot
FIRST WEEKLY · SUPERVISED
SHARED · INSTALLPATH A · WE SHIPPATH B · YOU SHIPEACH CELL = 1 DAY
// 07 · THE STACK

Tools the
pipeline runs on.

No proprietary black box. Everything underneath is named, swappable, and (on Path B) yours to extend. We pick the best tool per job and rotate as the frontier moves — these are what we're running today.

// ORCHESTRATION
Claude
Brief expansion, scoring, copy iteration. Custom skills per client voice.
// AUTOMATION
n8n
The pipeline glue. Tag-in, slate-out, every Monday. Self-hosted on your infra (Path B).
// IMAGE
Midjourney
Hero shots, lifestyle, brand-consistent style refs. Style anchors locked per client.
// IMAGE · UTILITY
fal.ai
Flux, Nano Banana, batch variations. API-first, plugged straight into n8n.
// VIDEO · MOTION
Runway
Image-to-video, scene continuity, controlled camera moves. Gen-3 / Gen-4.
// VIDEO · SCENES
Higgsfield
Seedance scene-by-scene, batch submission, queue management. The 80-a-week workhorse.
// AUDIO
Suno + 11Labs
Background score, VO, brand voice cloning. Mixed in-line with the cut.
// DELIVERY
Notion + Drive
Slate, scoring, sortable history. Lives in your workspace, not ours.
// 08 · QUESTIONS ABOUT THE PROCESS

The five
we get asked.

How fast from kickoff to first live ad?+

Path A: 14 days to first slate, ~18 days to first live ad assuming your buyer queues normally. The first two weeks are install (audit + brief framework) and can't be compressed — they're the asset every later week stands on.

Path B: same 14 days to install, plus a 4-week training window before your team is generating solo. First live ad lands at day ~25 (we run the first slate together).

How many revisions per concept?+

The honest answer: we don't think in revisions. The pipeline generates eighty concepts a week — if one's off, we kill it and the next twenty are already queued. You pick winners, not patch losers.

On the rare custom hero piece (product launch, founder shoot, hero film), we cap at three rounds and quote separately. Those are exceptions.

Who owns the IP — assets, prompts, system?+

Assets: you. Everything generated for you is yours, full commercial rights, no carve-outs.

Path A prompts & tag schema: ours during the engagement, fully exportable when you leave.

Path B everything: yours. Prompt chains, n8n graphs, Claude skills, scoring rubric — transferred on day one of the install, no license, no kill switch.

What if a tool in the stack disappears or changes?+

The stack rotates every quarter. We swap tools as the frontier moves — Suno replaced ElevenLabs in some VO slots last quarter; Seedance replaced an earlier scene tool the quarter before.

For Path B clients, every swap ships as a versioned update with a 30-day deprecation window on the old tool. No surprise breakage.

Can we start on Path A and switch to Path B later?+

Yes — this is how most clients actually work with us. You hire the shop for a quarter while you decide whether your team should own the pipeline; if you decide they should, we credit your Path A spend against the Path B install.

The reverse also works. If you start on Path B and decide running it in-house isn't the move, we'll pick the keys back up for a quarter and re-evaluate together.

▶ READY TO RUN IT

Pick a path.
We'll do the rest.

One 30-minute call. We look at your last quarter's spend together, sketch the gap brief, and tell you which path fits — straight up, no demo theater.

HIRE THE SHOP ▶BUY THE SYSTEM ▶SEE THE WORK