Every time a production cost collapses, the premium moves somewhere else. Cheap printing made the handwritten note valuable. Cheap photography made the real photographer expensive. Cheap content is about to do the same thing to human craft.
Rand Fishkin made this argument in a recent SparkToro post that starts, of all places, with a Pokemon game. His prediction is that "made entirely by humans" will become a premium signal, and that people will pay more for work that shows care and taste. I think he is right, and I think most marketing teams are organizing themselves in exactly the opposite direction.
The evidence is in the stated preferences
Start with the simplest data point. Very few people say they want to watch AI-generated movies, read AI-generated books or follow AI-generated blogs. That preference shows up again and again in surveys, even among people who use AI tools daily.
I covered that contradiction last week in people hate AI and use it every day. The resolution is that people are happy to use AI as a tool and unhappy to receive its output as a product. They want the calculator, not a novel written by the calculator.
That gap is where the price tier forms. If buyers prefer human-made work and AI makes machine-made work nearly free, then human-made work becomes the scarce good. Scarce goods carry a premium.
Cheap output converges
The second piece of evidence is less obvious. Fishkin cites a 2024 study in which writers assisted by generative AI produced stories that were rated as more creative individually. But those stories were also more similar to one another.
Read that at market scale. Each marketer using AI gets a slightly better first draft. The whole market gets a flood of slightly better first drafts that all look the same. Individual quality rises, collective novelty falls, and the audience stops noticing any of it.
That is the trap. A team that measures itself on output per person will see its numbers improve while its work becomes interchangeable with every competitor using the same tools. I wrote about this in taste is the last thing AI cannot copy, and the convergence data makes the point sharper. Sameness is now the default setting.
Busier is not the same as better
Fishkin also points to a Harvard document arguing that AI does not make teams more productive so much as it makes them busier. That matches what I see in agencies and in-house teams alike.
When AI frees up an hour, the hour rarely goes to thinking. It goes to producing more: more variations, more posts, more emails. The calendar fills to the same level it was before, now with higher volume and less reflection.
His recommendation is to protect time for experimentation instead of filling every freed hour with output. That sounds soft. It is actually an operating decision with commercial consequences, because experimentation is where the differentiated ideas come from, and differentiated ideas are the only thing that will hold a premium.
The skill you lose by skipping the work
The part of the argument I find most important is about how taste develops. It comes from practice: building campaigns, getting them wrong, improvising under pressure, seeing what actually moves people. A marketer who has done that a hundred times can feel a weak brief before reading the second paragraph.
AI workflows can let people skip that practice. A junior marketer who never writes a first draft, never builds a media plan from scratch and never sits through a failed launch may produce acceptable work for years. They will also never develop the judgment to produce great work, or to know which of the AI's ten options is the right one.
Fishkin gives a sharp example. Delegating creative brief intake to AI might feel efficient, but it can leave the team less informed about the business it is supposed to serve. The brief is not admin. It is where the marketer learns what the client actually needs.
How I would position a team for the premium
This is not an argument against AI. I build AI systems for a living. It is an argument for deciding deliberately where humans stay in the work.
Automate the parts where sameness is acceptable: reporting, formatting, versioning, research summaries, first-pass analysis. Keep humans in the parts where sameness is fatal: the strategic idea, the creative concept, the voice, the brief, the final call. Then say so. If a campaign, an article or a design was made by people with real judgment, that is now a selling point, not an assumption.
Budget for it too. Clients will increasingly pay two prices: a low one for AI-assisted volume and a higher one for human-led work with a point of view. Agencies and teams that cannot tell those two apart in their own process will be priced as the first.
Cheap content is a commodity now. Care is the new scarce resource, and scarce resources get paid.