Coca-Cola ran their 2023 holiday campaign using AI-generated imagery. The response from creative professionals ranged from mild discomfort to open hostility. The campaign performed fine commercially. The brand's creative reputation took a hit that's harder to measure and slower to recover from.

What marketers actually use ChatGPT for

A 2024 Salesforce survey of 1,000+ marketers found that 51% already use generative AI in their workflow. The top use cases: writing first drafts of copy, generating social media variations, brainstorming campaign angles, and summarizing research reports.

These are production tasks. They're valuable time-savers. They're also not creative direction.

Creative direction means deciding what the brand should say, why, and how it should feel. It's choosing between ten possible campaign concepts and knowing which one will land with the audience six months from now. ChatGPT generates the ten options. It can't choose between them.

Where ChatGPT performs well

Volume. Speed. Variation. If you need 40 subject lines for an email A/B test, ChatGPT produces them in 30 seconds. If you need five angles on the same product benefit, it delivers instantly. These are real productivity gains — Jasper.ai reported that their enterprise clients reduced content production time by 40%.

ChatGPT also works as a decent sparring partner for early-stage ideation. You describe a brief, it generates directions, you react to them. The reaction is where the value lives — your gut response to AI output tells you something about what you actually want. Many creative directors at agencies already use it this way, as a rapid first-draft machine that they then shape, redirect, and refine.

Wieden+Kennedy's global chief creative officer Colleen DeCourcy described AI as "the most powerful brief ever." The output itself isn't the creative. The creative is what a human does in response to it.

The creative direction gap

Apple's "Shot on iPhone" campaign wasn't the obvious choice in 2015. User-generated content for a premium brand? The creative director who greenlit that — Tor Myhren — made a judgment call that contradicted conventional wisdom for luxury positioning. That judgment came from 20 years of pattern recognition about what audiences actually respond to versus what they say they want.

The ChatGPT creative director problem isn't about output quality. It's about selection quality. AI produces average-of-everything output because that's what training on the entire internet gives you. A creative director's job is to find the specific, non-obvious angle that cuts through the average.

Wieden+Kennedy's "Old Spice Man" campaign, Ogilvy's "Dove Real Beauty," BBH's "Levi's Go Forth" — all of these were counterintuitive strategic bets. Each one required someone to say "this weird idea is the right one" when safer options sat on the table. That's judgment. You can't prompt your way to judgment.

What AI can't do in creative work

AI can't feel cultural tension. It can't sense that an audience is tired of a particular visual style before the data proves it. It can't make the call to kill a campaign that tests well in focus groups but feels stale to someone who has watched consumer taste shift over a decade.

Margaret Boden, a cognitive science researcher at the University of Sussex, distinguishes between "combinational" creativity (mixing existing elements) and "transformational" creativity (changing the rules of the space itself). AI does combinational work — often impressively. Transformational creativity — the kind that redefines a category — requires understanding what rules exist and choosing which one to break. That requires knowing the game.

When Liquid Death wrapped water in heavy metal branding, they broke the rule that water brands must communicate purity and wellness. That rule-break was the entire business. ChatGPT, trained on decades of water branding, would have generated purity-and-wellness concepts because that's what the training data contained.

The ChatGPT creative director question reframed

The useful question isn't "can ChatGPT replace a creative director?" It's "which parts of creative direction are automatable and which require human judgment?"

Automatable: generating options, producing copy variations, mocking up rough concepts, research summarization, competitive audits. These tasks consume 40-60% of a creative team's time. Handing them to AI frees humans for the work only humans can do.

Not automatable: reading cultural currents, making taste-based decisions under uncertainty, killing good-enough ideas in favor of great ones, knowing when a brief is wrong and needs to be rewritten. These are pattern recognition skills built through years of practice and failure.

Training humans vs automating output

The better investment is both. Use ChatGPT for production. Invest in the human judgment skills that AI can't replicate.

Sparks trains the cognitive patterns behind creative direction: seeing alternatives, breaking assumptions, connecting unrelated domains, thinking from unusual perspectives. The same divergent and lateral thinking that Boden describes as distinctly human. These skills atrophy without practice, the same way a musician's ear dulls without regular training.

Marketers who use ChatGPT for drafts and Sparks for thinking skill development end up with both: faster production and sharper creative instincts. That combination is harder to hire for than either skill alone.