The writing job didn't disappear. It unbundled.

Kimberlee Meier wrote a post this week about what happened to her content consultancy when ChatGPT started gaining momentum.

In March 2023, a long-standing client told her they would no longer need her to write content. They would generate articles with AI instead and pay her half her normal rate to edit them. The next day, her biggest client paused all work. By the end of the month, she had lost three clients.

Three years later, she is still working in marketing. But the job she describes now is much wider: long-form and short-form content, campaigns, outbound, sales, customer success, design, LinkedIn, internal marketing and defending performance numbers.

That struck me because it makes the AI-and-writing argument much more concrete.

We spend a lot of time asking whether AI can write. Her experience shows that the first thing AI changed was not necessarily the existence of the work.

It changed the price of one part of it.

The first thing AI changed was the unit of value

For years, content writing had a fairly understandable commercial unit.

An article. A report. A customer story. A landing page. A batch of social posts.

A client could ask for a 1,500-word article and attach a price to it because producing that article required a bundle of activities that were difficult to separate. Somebody had to understand the brief, research the subject, find the angle, make decisions about evidence, produce the words and get the piece into publishable shape.

The visible deliverable was the article, so the article became the thing that was bought.

Generative AI attacked that visible layer first. A business could suddenly produce something that looked like an article without paying for every hour that had previously gone into producing one. Whether the result was any good was almost secondary to the immediate economic signal: the production cost had changed.

That is why Kimberlee’s story matters. The client did not say the subject no longer mattered or that content had stopped being useful. They changed what they were prepared to pay for the production layer.

The unit of value moved before the profession had agreed where it was moving to.

AI didn’t remove the bundle. It split it apart.

This is the bit I think gets lost when we describe writing as a single task.

A professional article has always contained more than words. There is research, interviewing, judgement, fact checking, audience understanding, argument, stakeholder management, editing, distribution, performance analysis and a lot of tacit knowledge about what should not be written at all.

Those things used to arrive together because one person often did most of them while producing the asset. AI makes it easier to separate them.

The first draft can now be generated cheaply. The research can be partially automated. Variations can be produced in seconds. A transcript can become a rough article before the interview has even finished cooling down.

I have written before that the AI draft is not the article. What I was really describing there was this separation. The draft is one component of the job, and increasingly it is not the component that determines whether the finished work is worth publishing.

Once that component becomes abundant, the other parts become much more visible.

Freelancers felt the economics before in-house teams felt the reorganisation

Freelancers probably experienced this change earlier and more brutally than many people inside organisations because the price signal was immediate.

If a client decides that drafting should now cost half as much, somebody’s income changes that month.

Inside a company, the same transition can be harder to see. The writer often remains employed while the boundaries around the job expand. They start using AI. Then they are expected to work across more formats. Then campaigns. Then distribution. Then closer collaboration with sales, customer teams and design. Then measurement. Then internal enablement.

The title may still say content writer or content marketer while the actual unit of work has become something much larger.

That is close to what I have been arguing in The content job is moving up a level. The finished asset is becoming one part of a wider system of questions, evidence, retrieval, decisions, distribution and memory.

Kimberlee’s post shows the same shift from the other direction. Instead of starting with the emerging strategic role and working backwards, it starts with the moment the old economic model broke.

That is useful because it stops this becoming an abstract argument about the future of work. Somebody actually received the email saying the writing would now be generated elsewhere and the editing would be worth half the fee.

This is not automatically a good outcome

There is an optimistic version of this story where writers become more strategic, more influential and more valuable because AI removes routine production.

Sometimes that will happen.

There is also a much less attractive version where one person is expected to become a writer, campaign manager, distribution team, analyst, designer, AI operator and sales enablement function because the technology supposedly makes everything faster.

I do not think those are the same thing.

A wider role can be genuinely more valuable when it comes with more authority over the questions, evidence, narrative and decisions that shape the work. It is simply workload expansion when the organisation keeps all the old output expectations and adds five new responsibilities on top.

I made a similar distinction in The content job is moving up a level: the interesting opportunity is not becoming a “full-stack content creator” who personally does every production task. It is owning more of the editorial decisions connecting the work.

That distinction matters for pay as well as job design.

If the profession responds to AI by saying writers should just become more strategic, more commercial, more technical and better at distribution, it risks accepting a strange bargain: the production part of the role is repriced downwards while everything added around it is treated as free.

Unbundling can reveal where the real value sits. It can also hide how much extra work has been bundled back into one person’s job.

The writing job was always bigger than writing

This is where the post connects most directly to Editorial Intelligence for me.

I started building Editorial Intelligence because I was frustrated by how often good content work disappeared into the asset. You did the research, interviewed the customer, found the proof point, discovered the useful contradiction, published the article and then started again from a fresh brief three weeks later.

The article was treated as the value. Much of the intelligence that made the article possible was discarded.

I now think that is almost exactly backwards.

The useful thing left behind by good editorial work might be an argument, a piece of evidence, a customer insight, a decision, a reusable question, a narrative, a source relationship or an understanding of where the organisation’s knowledge is weak. The published asset matters, but it is not necessarily the only thing worth keeping.

That is the argument behind Why content work needs to compound. Good work should improve the next piece of work rather than resetting after publication.

AI makes that more important because it reduces the scarcity of the output itself. If I can generate another competent draft tomorrow, the differentiator is increasingly what the organisation knows that the model did not already know, how well that knowledge has been tested and whether somebody can turn it into a defensible point of view.

The article becomes temporary packaging for something more durable.

My own comfort with AI is partly a seniority story

There is an uncomfortable part of this argument that I do not want to skip.

I have written about why being an ageing content writer may have made me comfortable with AI. After nearly twenty years of writing professionally, I am quite happy for a machine to help with the production layer because I have already spent years doing it the slow way.

I know what a weak draft looks like. I know when the argument is thin. I know when a source does not support the claim. I know when a piece is technically correct but not worth reading.

That makes “move upstream” a very convenient argument for somebody in my position.

It is much less useful advice to someone trying to get their first writing job.

The routine work AI absorbs is also some of the work people used to learn through. I have been following that problem in Raising the floor is not building a ladder: AI can help an early-career person produce stronger work sooner while making it harder to see whether the judgement underneath the output is developing at the same rate.

So I do not think the answer is simply that writers should stop worrying about writing and become strategists.

The profession still needs a way for people to acquire the judgement that makes the upstream work valuable in the first place.

What becomes scarce when words become cheap?

“Human creativity” is too vague an answer for me.

The things that look scarce in the work I am doing are much more specific.

Access to real customers. Evidence that competitors cannot simply scrape from the same public sources. The ability to interview somebody properly. Knowing which contradiction deserves investigation. Understanding the organisation well enough to spot when two separate projects are actually about the same problem. Knowing what can be claimed and what cannot. Remembering why a decision was made six months ago. Recognising when another article is unnecessary.

AI can help with almost all of those activities. It can search, compare, synthesise, challenge and retrieve. But it still needs something worth working on.

A thin brief plus a powerful model can produce a polished piece very quickly.

It cannot magically turn a thin organisation into a knowledgeable one.

That is why I keep coming back to evidence, editorial judgement and memory rather than writing quality alone. The harder problem is increasingly not producing the answer. It is establishing whether the organisation has earned the right to give it.

The job may now be rebundling around a different centre

There is a second stage to the unbundling.

Once the old writing bundle breaks apart, organisations have to decide how to put the work back together.

The job adverts I have been collecting suggest one answer. In The editorial jobs are not disappearing. They are moving upstream, I looked at roles asking editors and content leaders to own narrative architecture, customer evidence, thought leadership, executive messaging, governance, commercial priorities and the systems underneath the work.

Those jobs still involve writing.

Writing just is not the centre of gravity anymore.

The centre is closer to judgement and orchestration: deciding what is worth saying, establishing what evidence supports it, connecting the argument across different formats and functions, and making sure the organisation learns something from the response.

That may be where part of the profession is heading.

It does not mean the disruption was harmless. Kimberlee’s post is a useful reminder that careers had to absorb the shock long before the new role was clearly defined, and plenty of people will have left the industry rather than successfully rebundled themselves around a wider remit.

It also does not mean every organisation will value the wider work properly.

That is probably the next argument.

The question is no longer just whether AI can write

For a lot of routine marketing work, AI can produce something recognisable as content. That question is becoming less interesting every month.

The more useful question is what has to happen around the writing for the output to be worth producing at all.

Who found the evidence? Who understood the customer? Who decided that this was the right question? Who challenged the convenient claim? Who connected this piece to what the organisation already knows? Who made sure the insight survived the campaign? Who watched what happened afterwards and changed the next piece of work?

Those activities were always inside the writing job somewhere.

AI is making them easier to see because it has pulled the most visible layer away from the rest.

Kimberlee’s clients felt that first as a pricing decision. In-house teams are increasingly feeling it as a scope decision. Job adverts are starting to show it as an organisational design decision.

The writing job did not simply disappear. It unbundled, and the next question is whether writers get to help decide how it is put back together.