Two things people say about AI and creative work are both wrong. One is that it replaces creatives. The other is that it’s all slop. The truth is more specific, and more useful: it changes some parts of the work a lot, and other parts barely at all. The skill now is knowing which is which.
Start with the part that is hardest to argue with. By 2025, adoption was already wide enough that the useful question had moved from whether teams would use AI to where it actually improved the work. McKinsey’s 2025 survey has most organizations using the technology in at least one function, with marketing among the most common. A 2025 study of brand marketers found more than half already use it to generate content and campaign ideas. This isn’t a coming wave to prepare for. It’s the current baseline.
Where it genuinely changes the work
The real change is in the production layer: research, iteration, volume, versioning, localization. The repetitive, scale-bound parts. McKinsey describes campaigns that used to take months of content design now shipping in weeks. One consumer brand broke production into hundreds of small tasks and built around a hundred small agents to handle things like short-form copy and content versioning. Take one master asset and spin out a hundred localized variants from it, and the value is obvious and real.
The productivity research backs this up, with a catch worth holding onto. A large study of BCG consultants using GPT-4 found they finished roughly 12 percent more tasks, about a quarter faster, with notably higher quality, and the biggest gains went to the people who started below average. The tool lifts the floor. It’s very good at making an okay first pass faster.
Where it’s overhyped
Here’s the catch. The same studies that prove the productivity also show the limit. In that consultant study, on work that sat outside what the model was actually good at, people using it were meaningfully more likely to land on the wrong answer, and the ideas it produced got more similar to each other. A separate study in Science Advances found the same pattern with writers: assistance raised the creativity of any individual story by around 8 percent, but the stories themselves became measurably more alike. Individually better, collectively flatter. The researchers called it a social dilemma. Everyone reaches for the same tool, and the work converges on the same answers.
And audiences notice. Deloitte found about two-thirds of people familiar with generative AI worry about being fooled by it. Enthusiasm for AI-made creator content dropped from 60 percent to 26 percent in two years as the slop set in. Researchers have found that simply labeling an identical ad as AI-generated measurably lowers how people rate it. The brand risk isn’t hypothetical.
That doesn’t mean machine-made equals bad. When System1 tested a set of AI-assisted ads, they scored above the average ad in its database, and Coca-Cola’s AI-assisted holiday spot hit a top score. So the lesson isn’t that the technology is good or bad. It’s that emotion, story, and brand fluency still decide whether the work lands, and those are still human calls.
The frame that holds
So the frame I use is simple. It’s a feature, not the pitch. A tool with a human in the loop, not an autopilot you walk away from. In my own concept work, the automated parts draft and a person approves, the system proposes routes and a person picks the one with an actual point of view. That’s not a hedge against the technology. It’s where the evidence points. It’s best at volume and speed, worst at taste and judgment, and taste and judgment are the whole job.
None of this changes the job of having something to say. Use the tooling to do more of the repetitive work faster, so the human hours go where they count: the idea, the brand, the decision about what’s even worth making. The teams that get this right won’t be the ones using the most AI. They’ll be the ones who kept a person in the loop where it mattered.
Sources
- McKinsey, “The State of AI” (2025): mckinsey.com
- McKinsey, generative AI in marketing (months to days; agents): mckinsey.com
- Dell’Acqua et al., “Navigating the Jagged Technological Frontier” (Harvard / BCG, 2023): SSRN
- Doshi & Hauser, “Generative AI enhances individual creativity but reduces the collective diversity of novel content,” Science Advances (2024): science.org
- Deloitte Connected Consumer Survey (2024): deloitte.com
- System1 / Jellyfish AI ad testing, via Marketing Week: marketingweek.com