There is no shortage of AI video tutorials online. Most of them end where real work begins: with a beautiful ten-second clip that no brand could ever run.
I work as an in-house AI creative director at a major FMCG company. My job is not to make impressive clips — it’s to ship commercials that a real brand, with real guidelines and a real legal department, will actually put in front of customers. Over the past years I’ve compressed that process into a repeatable six-day workflow, and in this article I’ll walk you through it exactly as it runs, day by day, with the numbers attached.
No hype. Just the process.
Why six days and not six hours?
Because generation is the fast part, and generation is maybe 20% of the job.
Anyone can produce footage in an afternoon. What takes time is everything an ad needs before and after the footage: positioning, story structure, brand accuracy, revisions, and the judgment calls that decide whether the film sells a product or just decorates a feed. When someone promises you a finished commercial in a day, one of two things is true — either the creative thinking was skipped, or it was never going to happen.
Here is where the six days actually go:
| Day | Phase | Main output | Share of effort |
|---|---|---|---|
| 1 | Brief & positioning | One-page creative brief, single key message | ~15% |
| 2 | Concept & storyboard | 3 routes → 1 approved, shot-by-shot board | ~20% |
| 3 | Visual development | Style frames, product reference pipeline | ~15% |
| 4 | Generation | Full shot coverage, 3–5 takes per shot | ~20% |
| 5 | Edit & compositing | Assembled film, product-accuracy passes | ~20% |
| 6 | Revisions & delivery | Client round, final masters + cutdowns | ~10% |
Notice what that table says: four of the six days are thinking, structuring, and correcting. Two are generating. That ratio has stayed stable across dozens of projects, and it’s the single most useful thing I can tell you about AI production.
What happens on Day 1 — the brief?
Everything that goes wrong in AI advertising goes wrong on Day 1, quietly.
I start every project by forcing the brief down to one page: who is this for, what single message must survive, where will it run, and what does the brand refuse to look like. For an FMCG product that last question matters more than people think — a biscuit brand that’s playful-but-premium dies instantly in footage that reads as cheap fantasy.
The deliverable is one sentence I call the spine: “After watching, the viewer should think ___.” If the client can’t agree on that sentence, we don’t move to Day 2. In one of our projects, the team wanted three messages in a 30-second spot — freshness, heritage, and a promotion. We spent half a day negotiating it down to one. That half-day saved the entire film, because AI will happily generate a confused ad faster than any human crew ever could.
How do I develop the concept without falling in love with AI tricks?
Day 2 is deliberately low-tech: paper, references, and three written concept routes — usually one safe, one bold, one culturally sharp. Each route is a paragraph plus a rough shot list, not a mood film. Clients choose faster and more honestly when they’re reacting to ideas, not renders.
Then comes the storyboard, and here’s my rule: every shot must earn its place in the story before it’s allowed to be beautiful. I storyboard 12–18 shots for a 30-second spot, each with its purpose written next to it — establish desire, show texture, land the product, close. Shots that exist only because “AI can do this now” get cut. That single discipline is what separates ads from demos.
Day 3 is visual development: I lock the film’s look with 5–10 style frames and — critically for FMCG — build the product reference pipeline. The pack shot, the exact label, the true colors. This is unglamorous work, and it’s precisely what the cheap providers skip. I covered what that skipping costs in my breakdown of AI commercial pricing.
What does the generation day actually look like?
Day 4 is the day everyone imagines the whole job to be.
I generate full coverage for the board — typically 3 to 5 takes per shot, which for a 15-shot film means 45–75 generated clips reviewed that day. I select fast and ruthlessly, with the storyboard open next to me. The question is never “is this take stunning?” It’s “does this take do its job in the story?”
Two honest numbers from my own projects:
- Roughly 60–70% of generated takes get rejected. Not because the tools are bad — because brand work has narrow tolerances. A hand holding the product slightly wrong, a label that drifted, lighting that breaks continuity with the previous shot. Rejection is the process working, not failing.
- The most expensive mistake is generating without a locked board. Early on, I once let generation start from a loose shot list. We produced twice the footage and used a fraction of it — an entire day lost to optionality. Never again. Freedom in AI production comes from constraints set the day before.
Where does the film actually get made?
Day 5 — the edit. This is my quiet conviction after years of shipping: AI films are made in the edit, not in the prompt.
Assembly, rhythm, sound design, and the passes that make a brand trust the film: product-accuracy compositing where the generated pack isn’t perfect, color continuity across shots born from different generations, and a frame-by-frame check against brand guidelines. On a typical 30-second FMCG spot, I spend more hours in the edit than in generation — and the client never sees this work directly. They just feel that the film holds together, without being able to say why.
Day 6 is the revision round and delivery. Because regeneration is cheap, I can offer what traditional production can’t: real revisions. A client asking to re-shoot the hero shot of a traditional commercial is asking for a five-figure change order. In my workflow it’s a same-day fix. We close with final masters plus cutdowns — 6s, 15s, vertical for Reels and TikTok — because AI collapses the marginal cost of versions, and a hero film without its cutdowns is half a delivery in 2026.
Where does this workflow still struggle?
An honest section, because this site doesn’t do hype:
- Long dialogue scenes. Sustained on-camera talking with precise lip sync across many shots is still the weakest link. I design around it or go hybrid.
- Precise hand-product interaction. Fingers opening a wrapper exactly right may take many takes or a compositing solution. I budget for it on Day 3, not discover it on Day 5.
- Legal and claims. AI generates images, not permissions. Anything implying a verifiable claim — ingredients, health, comparisons — goes through the same review as any traditional ad. The workflow doesn’t exempt you from advertising law.
Six days is also not a law of physics. A single-product spot with an approved concept can land in four; a campaign with localizations takes two to three weeks. What stays constant is the ratio — thinking outweighs generating, roughly four days to two.
What should you prepare before Day 1?
If you’re a brand considering this process, three inputs cut days off the timeline — and their absence adds days back:
Your brand assets, in usable form. High-resolution pack shots from multiple angles, the exact logo files, and your brand guidelines. I’ve had projects where chasing a printable version of the label took longer than storyboarding the film.
One decision-maker. The six-day clock assumes someone can approve the concept on Day 2 and the look on Day 3 within hours, not committees within weeks. AI removed the production bottleneck; approval loops are the only bottleneck left, and they’re entirely on the client’s side of the table.
A real answer to “what does success look like?” Views, add-to-carts, retail sell-in, internal alignment — these lead to different films. Naming the metric on Day 1 is free. Discovering it after delivery costs a re-edit.
Give me those three things and six days holds. Withhold them and no workflow on earth — AI or otherwise — will save the schedule.
FAQ
Can this 6-day AI ad workflow work for any brand?
For most product and lifestyle categories, yes. Heavy dialogue, celebrity likeness, or footage that must be verifiably real still call for hybrid or traditional production — and a serious studio will say so upfront.
Which AI tools do you use in this workflow?
The honest answer: it changes every few months, and it matters less than people think. The workflow — brief, board, references, coverage, edit — has outlived several generations of tools. Judgment transfers; tool tips expire.
How many revision rounds are included in 6 days?
One structured round on Day 6, and that’s usually enough — because the client already approved the concept on Day 2 and the look on Day 3. Approvals staged early are what make late surprises rare.
Is 6 days really faster than traditional production?
A traditional 30-second commercial typically runs 4–12 weeks from brief to delivery. Six days against that is a 5–10x compression — with test variations included rather than priced as extras.
I’m Mohammad Nasravi — an in-house AI creative director for a major FMCG brand. This workflow is the one I actually run, not a theory. If you want your next commercial built this way, see my services or get in touch — I reply within 24 hours.