Class 72 taught the paid engine: testable cards, destinations that convert, borrowed trust, and honest numbers. Class 73 turns to the fuel that engine burns — advertising creative — and to the specific way AI changes how a small team produces it. The promise is real: one person can now explore more angles, draft more variants, and organize more assets in an afternoon than a team once managed in a week. The trap is equally real: a folder of fluent, good-looking variations with no brief, no selection, and no review gate.

The line, before and after AI

The traditional sequence is unchanged in shape:

strategy/brief → copy/image/video/voice variations → selected creative
→ landing page → review → launch

What AI changes is throughput inside each stage, not the stages themselves. A traditional small team wrote three headlines, shot one cut, and argued about both for a week. An AI-enabled small team can draft thirty headlines, generate visual directions, cut caption versions, plan resize and crop variants, sketch landing-page sections, and file everything before lunch. That speed is only useful if selection, truth, and review keep pace — otherwise the team ships confusion faster.

Reuse the principle from Part XIII, the canonical home for making media: the model produces drafts; the human directs the promise, proof, audience fit, and selection. Part XIV, where you now stand, is the canonical home for getting that creative to the right people. Keep the boundary clean: generation belongs to the machine's speed, judgment belongs to human ownership.

What AI accelerates, what humans own

Be explicit, because every creative failure in this class comes from swapping the two columns:

A concrete pass: the tailor's "ready before Friday" angle becomes ten AI-drafted hooks, three caption lengths, two visual directions (garment close-up versus tailor at work), and a draft booking-section headline. The human then discards nine hooks, checks that Friday capacity is real, picks the photo showing actual work, and kills any draft claiming "any garment, any deadline." Speed upstream, strictness downstream.

Browse the TikTok Creative Center to see which short-form structures currently hold attention, and the Meta Ad Library to study how competitors frame comparable promises. Treat both as field research for the brief — what hooks, formats, and proof styles recur — never as permission to copy claims or aesthetics wholesale.

The board that prevents the folder-of-randomness

A production board makes every asset answerable: where it came from, what it claims, who approved it, where it points, and how it will be judged. Create CREATIVE-PRODUCTION-BOARD.md with six columns or sections:

• Brief: audience moment, promise, proof, destination, and the claims that are off-limits — one short block everything links back to.
• Assets: each draft with its angle, format, hook, and status (draft, in review, approved, rejected with reason).
• Approvals: who approved what and when; no asset moves to launch without a named approver.
• Versions: variant identifiers stable enough that results can later be traced (this becomes the variation matrix in Lesson 73.3 and the test plan in Class 74).
• Destination: the exact landing page or sample each asset points to, confirming message match from Lesson 72.2.
• Measurement owner: who reads results, on which metric, by which date — named before launch, not after disappointment.

Run Sonariq through the board in miniature. Brief: analysts doubting unverified summaries; promise — a sample brief with every claim linked; proof — one real filing-to-brief sample; off-limits — invented client outcomes, coverage claims beyond supported sources. Assets: two hooks ("Where exactly is that stated?" versus "Cite every line"), one demonstration clip, one static carousel of the citation trail. Destination: the sample page. Measurement owner: one named person, conversion defined as completed sample brief, review in seven days. Everything downstream — angles, variations, review — hangs off this board.

Exercise: map your production line

Build CREATIVE-PRODUCTION-BOARD.md for one real or realistic offer. Fill the brief first, in under 150 words. Then add at least four asset rows in draft status, each naming its angle and format. Assign an approver, a destination, and a measurement owner with a review date — even if every role is you wearing different hats. The discipline is naming the hats separately so review actually happens.

Finish line: a production board with a brief, four tracked assets, named approvals, linked destinations, and a measurement owner with a date — a system, not a folder.

Verify quickly: pick any asset row and trace it backward to the brief's promise and forward to its destination and metric. If either link is missing, the board has a hole.

Common failure mode: generating forty variations before writing the brief, then selecting by taste under deadline. Volume without a brief is noise with rendering costs; the board exists to force brief-first production.

Check your understanding

1. Name the six stages of the production line and explain which ones AI speeds up. 2. Which three responsibilities must never be delegated to the model, and why? 3. How does the production board prevent the "folder of random variations" failure?

Next

The board needs its first input: angles worth producing. Lesson 73.2 turns approved customer evidence into testable creative angles, with every claim sitting beside its proof.