The Visual Coherence of Corporate Control

Sarah Chen watches the Firefly training bar crawl toward completion. Ninety-three percent. The AI has consumed her brand's visual identity — twenty-seven logos, fifteen campaign images, the complete style guide she spent three years perfecting. In thirty minutes, it'll spit back a custom model that can generate infinite variations of her work.

"This is what we've been waiting for," her creative director says. "Brand consistency at scale." Sarah nods, but something feels wrong. Like watching a photocopier memorize your handwriting.

Adobe released Firefly Custom Models to public beta this week. Feed it 10-30 of your images, pay 500 credits, and it learns your creative signature. "A creator's style is their signature," Adobe's Deepa Subramaniam writes. "For creative pros and brands, it's their identity."

But identity for whom?

The Talent Pipeline Apocalypse

While Adobe promises brand salvation, the creative execution jobs are hemorrhaging. Computer graphic artists down 33% this year. Photographers down 28%. Writers — the actual humans who used to craft the words — down 28%. These aren't seasonal dips. They're structural collapses.

"Creative work is splitting between strategic roles and execution roles," reports a new analysis of 180 million job postings. Strategy survives. Execution dies. Creative directors endure while junior designers vanish.

The math is brutal: Adobe's Custom Models cost 500 credits. A junior designer costs $45,000 annually. Plus benefits. Plus the risk they'll leave for TikTok. The algorithmic apprentice never calls in sick, never demands a raise, never questions the brief.

"Job postings for 'AI Creative Director' increased by 120% since 2022," according to new industry data. Meanwhile, 68% of creative freelancers report reduced job security due to AI. The pyramid inverts — all chiefs, no Indians.

Brand DNA as Intellectual Property

Adobe's promise feels seductive: "Your models are private by default." Your training data won't contaminate the shared pool. Your aesthetic remains yours alone. It's creative nationalism — build walls around your visual identity, defend against generic AI slop.

But this creates a new creative caste system. Enterprise brands with licensing budgets can train bespoke models on Getty archives. Indie agencies scrape Instagram and pray they don't get sued. "Adobe regularly reviews users' custom models and will remove those that don't play by the rules," the company warns.

The result? Visual monocultures. Every McDonald's ad looks like every other McDonald's ad, generated by the same Custom Model trained on the same focus-grouped creative brief. Rebellion becomes impossible when the machine has memorized your brand guidelines.

The Agency Death Spiral

"CFO pressure, not CMO leadership, will drive AI adoption in agencies," veteran Martin Sorrell predicts. He's right. When Meta's AI agents can "analyze your ad campaigns, match you with creators, and draft client replies" inside Ads Manager, what exactly does a junior account executive do?

U.S. advertisers will push $57 billion through AI-powered platforms this year — 63% growth while human-managed spend limps at 5%. Performance Max and Advantage+ automate targeting and budgets. The humans become expensive ornaments.

Sara watches her Custom Model finish training. Score: 87 out of 100. "Ideal," according to Adobe's standards. She tests it with a simple prompt: "Generate a summer campaign hero image in our brand style."

The result is perfect. Technically flawless. On-brand. Focus-grouped optimized. And completely, utterly soulless.

The Question Nobody Wants to Answer

"Workers with advanced AI skills earn 56% more than peers in the same roles," PwC reports. But what happens when there are no peers left? When the junior designers who would have developed those advanced skills never get hired in the first place?

Adobe's Custom Models solve the wrong problem. Brands don't need more efficiency. They need human weirdness, cultural insight, the happy accidents that happen when a junior designer misreads the brief and creates something unexpected.

But efficiency scales. Weirdness doesn't. And in the race to the bottom of cost-per-impression, guess which one wins?

The custom model generates another perfect image. Sara stares at her reflection in the monitor. Tomorrow, she'll train it on her latest campaign. Next week, it'll know her style better than she does. Next year, will it need her at all?