AI Made the First Draft Cheap. Now What?
I’ve been thinking about AI a lot lately. Not in the “AI is going to replace all marketers by Tuesday” way. More in the much less dramatic, much more inquisitive way: What does this actually change about the job?
Because I think we’ve spent a lot of time asking whether AI can write. It can. Can it come up with 10 campaign ideas? Absolutely. Can it turn a few bullet points into something that looks suspiciously like a LinkedIn post? Very easily. Can it give you a first draft before you’ve finished your coffee? Fortunately and unfortunately, yes.
The cost of a first draft is approaching zero. For a long time, getting from nothing to something was a significant part of creative work. You had to sit down. Think. Write a terrible first sentence. Delete it. Write another one. Open Instagram for “research.” Spend 25 minutes there. Eventually come back. Write something slightly less terrible. That first draft had a cost. Time, mostly. Sometimes money. Often both.
AI has dramatically reduced that cost. Give it a brief, and you can have five directions in seconds. Ask for ten hooks, and you’ll get ten. Ask for a campaign concept, and you’ll get one before you’ve had time to wonder whether you actually like the idea. And that’s useful. I don’t want to pretend it isn’t.
I’ve used AI to get unstuck, explore angles, organise thoughts, interrogate an idea, rewrite something I’ve already written, and occasionally produce the thing I knew I wanted but couldn’t quite articulate.
It is very good at getting you from blank page to something on the page. But that creates a new problem.
When everyone can make something, making something isn’t the whole skill anymore If AI can produce a perfectly serviceable first draft for almost anyone, then the ability to produce it becomes less differentiating. And “serviceable” is doing a lot of work there. AI is very good at making things sound like they belong. A LinkedIn post that sounds like a LinkedIn post. A campaign headline that sounds like a campaign headline. A brand caption that sounds like a brand caption. A strategy document that has all the headings a strategy document is apparently supposed to have.
Sometimes you read it and think: This is… fine. And that’s the problem. There is now an almost infinite supply of fine. So what becomes valuable? Knowing when something is actually good. Knowing when an idea is boring before you’ve spent three weeks building it. Knowing that the clever headline is, in fact, trying far too hard. Knowing that the fifth carousel on “5 lessons from failure” does not need to exist. Knowing that your brand doesn’t need to join every conversation simply because the algorithm invited it.
In other words: taste.
Taste is harder to automate Taste is not just knowing what looks nice. It’s judgment. It’s knowing what to keep and what to throw away. It’s understanding context. It’s knowing why something works, not just recognising that it looks like something that has worked before. And, perhaps most importantly, it’s being willing to say: No. No, this isn’t the right idea. No, this doesn’t sound like us. No, our audience doesn’t need this. No, just because everyone is doing it doesn’t mean we should. No, the AI-generated version is technically better written, but the original has more personality - and this is increasingly important.
Because AI is very good at smoothing things out, including what makes marketing memorable. The slightly strange phrase. The unexpected observation. The opinion that isn’t perfectly balanced. The sentence that sounds like a human being actually said it.
So what should marketers do with AI? Use it, obviously. I don’t think the answer is to become romantically attached to doing everything manually. If a machine can take a tedious two-hour task and turn it into a ten-minute task, let the machine do it. The interesting question is what you do with the two hours you just got back. I’d rather spend that time understanding the audience, looking at the numbers, reading comments, studying competitors, talking to customers, thinking about the actual problem we’re trying to solve, or finding the strange little insight that nobody else has noticed yet.
Because AI can give you 20 campaign ideas. It cannot tell you which one your audience is secretly waiting for. Well, it can suggest one. But somebody still has to know whether it’s right. And that, increasingly, is the job.
The first draft is not the product. This is where the conversation about AI and creativity gets confused. We keep asking whether AI can do the creative work. But a first draft was never the whole creative process. The first draft is just the beginning. The interesting work happens afterwards when you have to decide what survives, what gets cut, what gets sharper, what feels too obvious, what doesn’t sound like the brand, what would make someone stop scrolling, what would make them care, and what would make them remember it tomorrow.
AI can help with all of those questions. But you still need someone capable of asking the right ones. And that’s why my current theory is pretty simple: AI is making the cost of a first draft approach zero. Which, inconveniently, makes taste much more valuable.
So perhaps the future of marketing isn’t about who can produce the most content. There will be plenty of that. It’s about who can look at all that content and say: “This one. Not that one. And here’s why.” I’m still figuring out what that means for the way I work. But I suspect I’ll be thinking about it for a while.
Field Notes
New notes land here when I have something worth sending.