AI Video Ad Production Workflow for a GenZ Skincare Brand

Industry

Beauty, Wellness & Personal Care

Geography

United States

Cooperation Period

6 Months (ongoing)

Overview

A United States based GenZ focused direct to consumer (D2C) skincare brand was spending roughly $300,000/mo on Meta and TikTok paid social and hitting a wall. This brand sells to a challenging audience segment that scrolls past anything that looks like an ad, burns through creative faster than any other demographic, and trusts ingredient lists more than influencer endorsements.

Their marketing lead claimed that most of their winning ads decayed within 10 to 14 days. The internal creative team could produce 15 to 20 new ads per month, nowhere close to the 200+ variants the account needed to keep testing at the pace GenZ audiences demand.

The brand carries six hero SKUs, each with different variants of claims (clinical results, ingredient breakdowns, before and after outcomes, peer reviews). They wanted those claims packaged across different hooks, body structures, CTAs, aspect ratios, and durations.

We built a repeating production workflow powered by Google Flow, operated by a content and a video editing team, along with a department head managing delivery through Zoho Projects.

Within the first production quarter, the workflow was producing over 200+ tracked ad variants each month, and the client’s media buyer was feeding performance observations back into each cycle.

About the Client

The client is a fast growing GenZ focused direct to consumer skincare brand selling 6 different hero products through a Shopify storefront.

The product line runs narrow and deep. Serums, treatments, and moisturizers, each backed by clinical trial data, full ingredient transparency, and a growing base of repeat buyers between 18 and 27 who research every product before buying.

We call this a “skintellectual” buyer, the consumer who reads the INCI list on TikTok, cross references ingredients on INCIDecoder, and will not purchase a serum without knowing the exact percentage of niacinamide in the formula.

Paid social is the primary growth channel, and TikTok accounts for the larger share of spend. By the time the brand approached us, monthly spend on Meta and TikTok sat around $300,000, with a dedicated inhouse media buyer managing the ad accounts.

The client’s internal creative team, two people, handled product photography, static ad cards, and the occasional video. The marketing director sat across both functions, translating media buying signals into creative direction.

That setup stopped working once the media budget crossed a threshold where two designers simply could not feed the ad account fast enough to prevent the same ads from burning out in front of GenZ audiences who punish repetition harder than any other age group.

Fig 1 – The same brand, the same budget, the same products, with an ad account that went from starving for creative to more variants than the media buyer could test in a single flight

The Challenge

Four constraints were competing against each other, and each one fed the others. All four hit harder because the audience was GenZ.

Creative fatigue outran production capacity. A winning ad on Meta or TikTok now fatigues in five to seven days on average, with click through rate dropping 30 to 50 percent by day eight to ten. For GenZ audiences specifically, that window is even tighter. From our internal data, we derive that 58% of GenZ skip or scroll past ads within two seconds.

This brand’s decay window was 10 to 14 days, slightly better than average but still faster than what their internal lean team could replace. Every week the brand was spending $37,500 against ads the audience had already seen too many times.

The variant math was impossible at current headcount. The brand’s six hero SKUs each carry three to four claims.

clinical results
ingredient stories
before and after outcomes
peer reviews

Multiply those claims across different hook styles (the first two to three seconds, which is where GenZ decides to watch or scroll), body structures (testimonial, problem agitation solution, demonstration, ingredient education), CTAs, three aspect ratios (9:16, 1:1, 4:5), and three durations (15s, 30s, 60s), and the full testing surface runs into the thousands.

Even narrowing to the highest priority claim pairings, the media buyer needed 200 or more variants a month to test meaningfully. Two designers producing 15 to 20 videos a month were covering roughly 8% of that.

No structured way to learn from what worked. The media buyer could see that ad 47 beat ad 52 in the Meta dashboard, but there was no system to record why. Was it the hook? The claim? The CTA? The length? This matters more with GenZ because their preferences shift faster. A hook style that works in January can feel stale by March. Without that resolution, every new production cycle started from intuition, and at $300,000 a month the gap between a good guess and a data informed decision adds up fast.

AI video tools had already burned them once. The brand had experimented with generic AI video generation before coming to us. The results looked fine on first glance. Skin tones shifted between clips, product packaging colors drifted, and label text blurred.

The brand pulled two ads after they went live because the serum bottle in the video did not match the actual product. For a GenZ audience that prizes authenticity and ingredient transparency, that kind of visual inconsistency is very close to a trust issue.

The Solution

We built a production workflow that repeats every week, operated by three groups.

a content writing team that handles scripting and prompt engineering
a video editing team that handles clip generation, assembly, and quality checks
a department head who manages the full delivery cycle and client feedback

The variant math works because of how the workflow is structured. We produce 8 to 12 master concepts per month (each one a fully scripted, edited video for a specific SKU and claim combination), and each master fans out into variants through hook swaps, CTA changes, aspect ratio exports, and duration cuts.

One master concept with 3 hook options, 2 CTA options, 3 aspect ratios, and 3 durations produces 54 trackable variants. Ten masters a month puts the account past 200 unique assets the media buyer can test independently.

Every step in the workflow has a named owner, a defined handoff, and a Zoho Projects task that tracks it. Here is how a video moves from brief to published ad, and how feedback from the client’s media buyer cycles back into the next production round.

Fig 2 – Every video passes through the same seven stage workflow, and both types of client feedback reenter the cycle at defined points

Step 1. Client Brief and Requirements

Each production cycle almost always starts with a brief the client submits. The brief specifies which SKUs need new creative, which claims to feature, any seasonal or promotional angle, and the formats and durations required.

The Clixlogix team reviews the brief, clarifies anything ambiguous (we learned early that “make it feel more authentic” needs to be translated into specific visual and copy direction before it reaches the writing team), and creates individual task tickets for each master concept with deadlines, assigned team members, and deliverable specs.

Everything is agile at best, meaning, if the client’s media buyer has performance feedback from the previous cycle, it enters here. That feedback comes in a specific format. The media buyer flags which variants performed well and which underperformed, notes which hook type, claim, or CTA they believe drove the result, and attaches screenshots from Meta or TikTok Ads Manager.

Our team translates client feedback into production priorities for the next cycle. If ingredient education hooks outperformed problem agitation hooks on a specific SKU last month, that SKU’s brief this month will lean heavier on ingredient education structures, because GenZ audiences consistently respond better to content that teaches them something they can verify.

Step 2. Script Writing and Prompt Engineering

Clixlogix’s content team scripts each master concept using direct response copywriting structures the brand has approved which includes artifacts like testimonial, problem agitation solution, demonstration, ingredient education, before and after, and social proof.

Each script specifies the hook (the first two to three seconds, which is where the ad wins or loses with GenZ viewers), the body structure, the closing CTA, and the on screen text overlays. The scripts also flag which product claims are approved for each SKU and which need softer language, because skincare claims around clinical results carry serious compliance risk.

Alongside the script, the content team writes the prompts using a library of proprietary skills that team Clixlogix uses for such projects. A script says “show the serum bottle rotating against a soft gradient background with light particles.” The Flow prompt needs to specify camera angle, lighting direction, background color hex, motion speed, and shot duration to get a usable clip on the first or second generation. Bad prompts waste hours of regeneration. A shared prompt library was built over the first two months and new concepts pulled from that library as a starting point.

What did not work as expected

Scripts written for 30 second ads do not compress into 15 second ads by cutting words. The rhythm is completely different. A 15 second TikTok ad gets one hook, one claim, one CTA, and that is the entire script. GenZ does not give you time to warm up. We ended up writing separate scripts for each duration bracket, which added to the upfront workload but eliminated the problem of 15 second videos that felt rushed and 60 second videos that felt padded.

Step 3. AI Clip Generation in Google Flow

Google Flow is built on Google DeepMind’s Veo model. It generates video from text prompts and reference images with control over camera angles, motion, pacing, and scene consistency. We use the client’s actual product photography as reference input, which keeps the generated clips anchored to real product appearance. The content team submits the prompts, and the video editing team reviews the raw output for usability before it moves to assembly.

Clixlogix has written about the technical side of building AI video generation workflows in a separate blog, and the principles there apply directly to this engagement. The short version is that AI video generation is fast, but the output is inconsistent. You generate three to five versions of each scene and pick the best one.

That selection step is where the video editing team’s judgment matters most, because they are evaluating whether the clip matches the script’s intended pacing and emotional beat.

Step 4. Editing, Assembly, and Compliance Check

This is where raw AI clips become actual ads. Clixlogix’s video editing team assembles each master concept from the selected clips by cutting transitions, timing scenes to the script, layering text overlays, applying the brand’s color grade, adding background music (subject to client licensing) and syncing everything to the intended pacing for each duration (15, 30, and 60 seconds).

The compliance check happens here, inside the editing step, before the video leaves the team. Every clip is reviewed against four criteria the brand set after the earlier AI video failure.

skin tone accuracy against the brand’s approved reference set
product packaging and label consistency
approved typography
compliant claim language in all text overlays

The video editing team has a checklist attached to every task. If a clip fails any of the four criteria, it goes back to generation. Two clips in the first production month were caught for skin tone drift, one for a label color shift. All three were regenerated and replaced before the client ever saw them.

An honest note on AI generated product shots

The serum bottle cap color drifted from a matte to a glossy in about 20 percent of initial generations. Our editors learned to catch this, but it required building a visual reference sheet with zoomed in details of each product's packaging. That reference sheet now lives in the project oracle of shared files and every editor on the team references it during generation. For a GenZ audience that will screenshot a product inconsistency and post it on Reddit, this level of visual fidelity is not optional when dealing with AI generated video workflows.

Example AI Video Ad, Concept Piece

A concept ad Clixlogix produced with the same workflow to demonstrate the output, from product reference image through AI video clip generation to a finished, jingle scored ad. It is not one of the variants delivered to this client. The Gemini watermark is left intact to mark the video as AI generated.

Step 5. Client Review and Revisions in Zoho Projects

Master concepts go to the client for review through the project board. The project lead uploads finished videos to each task, tags the client’s marketing director and media buyer as reviewers, and sets a 48 hour review window. The client can approve, request changes, or reject.

Revision requests come back as comments with timestamps and specific notes. “The hook on SKU 3 is too slow, speed up the first two seconds” is a clear revision. “Make it pop more” gets a follow up question from the department head before it goes back to the editing team. We set this expectation during onboarding with our clients, specific feedback gets a 24 hour turnaround, vague feedback gets a clarification round first.

When revisions are needed, the task routes back to the appropriate team. Script changes go to the content writing team. Visual and pacing changes go to the video editing team. We track every revision cycle with comments, timestamps, and version uploads so nothing gets lost and nobody argues about what was requested versus what was delivered.

Step 6. Variant Multiplication and Tagging

Once a master concept is approved, the video editing team produces the full variant set. This is the step that turns 10 master concepts into 200+ assets. Each master gets rebuilt with alternate hooks (typically two to three per concept), alternate CTAs, and then exported across three aspect ratios (9:16 for Stories, Reels, and TikTok; 1:1 for feed; 4:5 for Meta feed placements) and three durations (15, 30, and 60 seconds).

The department head tags every variant in a naming convention the media buyer can filter. Each file name carries the SKU, hook type, claim, CTA variant, aspect ratio, duration, and version number. The naming convention lives in a shared project oracle document so the media buyer is never confused. When the media buyer opens the shared delivery folder, they can sort and filter by any attribute with very little uncertainty.

The math in practice across a recent month includes 11 approved master concepts, each producing an average of 22 variants (range was 12 to 36 depending on how many hook and CTA options the brief called for), for a total of 242 delivered assets.

Step 7. Delivery and the Performance Feedback Cycle

Final variants are delivered with a delivery manifest that lists every asset by SKU, hook type, claim, and format. The client’s media buyer downloads the batch and loads them into Meta and TikTok Ads Manager.

This is where our workflow ends and the client’s workflow begins. The media buyer runs the ads, monitors performance, and decides what is working and what is not. When they have observations, they post them back as a feedback brief for the next production cycle. That feedback reenters the workflow at Step 1, closing the loop.

Fig 3 – Scripts and product assets go in at the bottom, finished tagged variants come out at the top, and the media buyer feedback cycles back down through Zoho Projects

Results

Results reflect the first production quarter. The brand’s $300,000 monthly media budget and internal team size stayed constant throughout. These numbers measure the production workflow only.

242 Tracked Variants Delivered in the Most Recent Production Month

From 11 approved master concepts, the workflow produced 242 tagged ad variants in a single production month. Each variant carries a unique combination of hook, claim, CTA, aspect ratio, and duration, so the media buyer can test any single creative variable independently.

6 Hero SKUs Each Running Independent Creative Testing

Each hero product has its own script library, claim set, and variant rotation. Before the workflow existed, the flagship serum absorbed most of the creative budget because it was the only product with video.

3 to 5 Business Day Turnaround Per Weekly Production Batch

A production batch moves from brief to delivered variants within a single business week. The media buyer never waits more than five business days for fresh creative, which keeps the production cycle ahead of the GenZ fatigue cycle.

0 Compliance Failures Reached the Client's Ad Account

Every variant passes the four point compliance check covering skin tone, packaging, typography, and claims before it leaves the editing team. Five clips were caught and regenerated across the first quarter, all replaced before the client ever saw them.

1 Zoho Projects Board Manages the Full Production and Feedback Cycle

Every brief, script, revision comment, approval, delivery, and performance feedback note lives in one board. The media buyer's performance feedback posts directly to the next cycle's brief task, so production priorities carry forward without a handoff meeting.

Technologies and Tools

Layer	Tools
AI Video Generation	Google Flow
Video Editing	Adobe Premiere Pro
Project and Delivery Management	Zoho Projects
Ecommerce Platform	Shopify
Paid Media Platforms (client operated)	Meta Ads Manager, TikTok Ads Manager
Scriptwriting	Google Docs, Google Sheets

Services Delivered

AI Video Production, Creative & Design, Direct Response Copywriting, Digital Marketing, Creative Strategy

Team Composition

Content Writing Team, Video Editing Team, Project Manager

Have a question for our team or need help with your project?

Our team can share client references, scope your project, and answer any question about your delivery.

Related Case Studies

More engagements where our delivery teams shipped similar outcomes for clients across industries. Read on for context on the patterns we reused, the trade offs we navigated, and the metrics that landed in production.

Social Media Crisis Management and Revenue Recovery for a Brooklyn Fine Dining Restaurant

Digital Marketing

Social Media Strategy and Creator Program for a US Certified Organic Herbal Skincare Brand