Every few months (or weeks these days), someone publishes a piece declaring that marketers are either doomed or saved by AI. Both camps aren't quite getting it, mostly because they are arguing about the wrong thing. The question isn't whether AI can do your job. The question is which parts of your job were already replaceable before AI arrived. That is a more uncomfortable question, and it's the one worth carefully considering.
Brief writing that could have been a template. Research that was really just Googling with better formatting. Copy that sounded like everyone else's copy because it was built from the same reference points. Social post captions produced at mass volume with minimal thought. Monthly reports that reorganised data without interpreting it. If AI can do those things faster and cheaper, that's not a threat to good marketing. It's an indictment of how those things were being done, and a signal about where the real work actually lives.
What AI can't replace, and won't for a long time, is discernment. The ability to walk into a room of work and know, without being able to fully articulate why, that something feels off. The judgment that comes from having been publicly incorrect a few times and learning what that actually costs. The instinct to know when a brief is actually asking the wrong question, and the confidence to say so before anyone has spent any money answering it (incorrectly). These things aren't what you develop by just reading about them. They accumulate through decisions made under pressure, with incomplete information, in front of people who are watching how you handle it.
I use AI tools regularly. For research, for first-draft structure, for pressure-testing arguments by asking the model to steelman the opposite position. It is genuinely useful for the mechanical and the generative. What I don't use it for is the judgment call — as that can only come off the back of years of experience. Which idea is right for this brand at this moment? What's this customer actually feeling but they have not said? Is this campaign honest, or is it 'clever' in a way that will embarrass us in twelve months? Those questions require context, experience and a point of view that's been built out of years of trial and error. A language model has the first in abundance, and the other two….not at all.
The marketers I have watched thrive in this environment are not the ones who use AI least, or the ones who use it most. They are the ones who treat it like any other tool: ruthlessly useful for the mechanical, and firmly kept away from the decisions that actually matter. Machine speed, human taste. That is the pairing. It's no different to those who were using Google in 2002 instead of time-consuming manual research. It's not complicated, but it does require you to be honest about which parts of your job are genuinely creative and which parts you have just been calling creative because the alternative leaves you too exposed.
The risk is not that AI creates bad marketing. Bad marketing has always been abundant and has never needed technology to help spray it hither and thither. The risk is that AI writes average marketing very efficiently, and that average starts to look like a ceiling rather than a floor. The volume of competent-but-forgettable content has increased dramatically and will continue to do so. In that environment, the brands that stay interesting, that say something distinct, that have an actual point of view (even if it's not a popular one), will be the ones where a human has still decided what interesting means. That human needs to know what they think. Not what the model suggests they think. What they actually, genuinely think.
That turns out to be the resource with the greatest scarcity.
