The Companies Building the Machines That Write Are Paying $400K for Humans Who Can. That Should Tell You Everything.

On Monday, OpenAI announced that Colin Fleming, most recently CMO of ServiceNow and a 13-year Salesforce veteran, is joining as Chief Marketing Officer of its business unit. Not CMO of the company. CMO of the business unit. The title alone tells you where their head is at.

Meanwhile, Anthropic's careers page is running a full hiring sprint for its Creative Studio: a Head of Copy & Content role listed at $320,000 to $400,000, a Copy Lead for Enterprise at $255,000 to $320,000, a Copy & Content Lead for Launches, a Motion Designer, and a Head of Anthropic Creative. This is not a side project. This is a full in-house creative agency being built from scratch at a company valued north of $30 billion.

The two companies that did more than anyone else to convince the world that creative jobs are toast are now spending like they have never needed creative people more.

We are supposed to pretend that is a coincidence. It is not.

And as of this week, the hiring data is no longer doing the talking alone. The CEOs are saying it themselves. Sam Altman told Commonwealth Bank of Australia CEO Matt Comyn on Tuesday that he was "pretty wrong" about AI's economic impact on jobs, walking back his June 2025 warnings that entry-level roles were at serious risk. Dario Amodei, who told Axios last year that AI could eliminate 50% of white-collar jobs, now frames automation as a productivity multiplier rather than a wrecking ball. The companies that scared the world into thinking creative work was over are now admitting that prediction was wrong, and they are proving it with their checkbooks at the same time.

The Job Descriptions Say It All

Read the Anthropic listings carefully. The Head of Copy & Content role sits "at the intersection of brand strategy, creative leadership, and storytelling" and owns the full voice, tone, and written creative across both the Anthropic and Claude brands. This person hires and grows a team of writers. They build the system that determines how a company on the edge of an IPO sounds every time it opens its mouth.

And here is the line from their Creative Studio job descriptions that should stop every creative professional mid-scroll: "We care about craft, making things by hand and using AI when it makes sense."

The company that builds Claude is telling you, in a public job posting, that handmade creative work matters. That AI is one tool inside a larger process, not the process itself. That is a deliberate, considered position from people who understand what these models can and cannot do better than anyone on earth.

On the OpenAI side, the build-out is just as aggressive. ThinkPol reported in April that OpenAI has been quietly hiring human writers, designers, and editors at salaries above $200,000. A Copywriter/Creative Director role in their Design Studio was posted last year at $240K to $295K. Their startup and enterprise marketing teams are posting new content roles every week. The Fleming hire guarantees that the entire B2B marketing machine is about to scale hard, which means more creative headcount, not less.

Stop Celebrating. Start Paying Attention.

There is a lazy, comforting version of this story that a lot of people want to tell. It goes like this: "See? Writers are safe. Even the AI companies need us."

That version is wrong, and it is dangerous if you believe it.

The same goes for the walkbacks. A lot of people are going to read Altman admitting he was "pretty wrong" and Amodei reframing automation as augmentation and feel like they won the argument. That is the wrong takeaway. The doomsday predictions missed on the timeline and the totality of it. They did not miss on the disruption itself.

What is actually happening is a violent split in the creative labor market, and these hires are proof that the split is accelerating.

At the bottom, the work that involved turning a brief into a first draft, the basic product descriptions, the templated ad variations, the SEO blog posts built around a keyword cluster, all of that is compressing fast. AI does it adequately, and adequately is all you need when the content is transactional. We wrote about this in The Great Flattening. The data has only gotten worse since.

At the top, something completely different is happening. The demand for people who can define a brand voice from zero, who can architect a narrative that holds across enterprise buyers and consumer users simultaneously, who can translate insanely technical AI products into language that moves a buying committee of six people who all want different things, that demand is surging. The comp is surging with it.

Anthropic is not paying $400K for someone who writes blog posts. They are paying $400K for someone who builds the system that governs how every word the company produces sounds, feels, and lands. That is brand infrastructure. It requires judgment about audience, competitive positioning, cultural context, and organizational identity that gets built up over years of doing the thing. No model can replicate it. Not because the technology is not good enough yet, but because the work is definitionally human.

OpenAI did not hire Colin Fleming to run ChatGPT's Twitter account. They hired him to build an enterprise brand that goes head to head with Salesforce, ServiceNow, and Microsoft for the trust of CIOs spending eight figures on software. That is a 15-year body of strategic and creative knowledge walking through the door. You cannot prompt that into existence.

The Karpathy Problem

This is the part that should make every creative person uncomfortable and motivated at the same time.

Andrej Karpathy, the former head of AI at Tesla and a founding researcher at OpenAI, rated copywriting as an 8 or 9 on his job exposure scale. "Very high exposure." He then joined Anthropic. The same Anthropic that is paying $400K for a head of copy.

Those two data points are not in conflict. They are the same data point observed from two different altitudes.

At the task level, AI writes copy. It generates headlines, drafts emails, and produces social posts that are passable. The exposure at that layer is real, and pretending otherwise is denial.

But at the strategic layer, the question changes entirely. The question stops being "can AI write a sentence" and becomes "can AI decide what this company should say, to whom, in what sequence, in what tone, through which channels, and why." The answer is no. The people building the models know that more clearly than anyone in a LinkedIn comment section arguing about whether ChatGPT can replace a copywriter.

Amodei's new framing from earlier this month lands right here. He said that automating 90% of a job does not eliminate it. The remaining 10% expands to become 100% of what people do, multiplying their productivity tenfold. That is not a prediction about some distant future. That is a description of what is already happening in creative right now. The commodity layer compresses, the strategic layer expands to fill the space, and the person standing in that strategic layer becomes ten times more valuable rather than redundant. The $400K comp band is the market pricing that shift in real time.

And then there is Altman's personal experiment. He told Comyn that he tried delegating his Slack and email responses to AI, then went back to writing them himself. "We really do care about our interactions with people," he said. "This thing is not something that I can imagine myself outsourcing to an AI anytime soon." Think about what that means. The CEO of OpenAI tried to automate his own communication and could not stomach the result. He arrived at the exact same conclusion Anthropic put in a job listing: we care about craft, making things by hand, and using AI when it makes sense. When the people who build the models cannot bring themselves to outsource their own words to the models, the wholesale-replacement argument falls apart.

Salesforce. Slack. Instacart. ServiceNow. Read the Room.

Pull back further and these hires reveal how the frontier labs see their own futures.

OpenAI is becoming an enterprise software company, full stop. Fleming is just the latest in a run of hires that tells you exactly where this is going. Denise Dresser came from Slack as CRO in late 2025. Fidji Simo joined from Instacart as CEO of Applications. Sarah Friar from Nextdoor has been running finance since 2024. Salesforce. Slack. Instacart. ServiceNow. These are not research lab hires. These are people who sell to enterprises, who understand procurement cycles, who know how to build trust with a CISO. Enterprise revenue already makes up more than 40% of OpenAI's total and is climbing fast. Every one of those executives needs creative and marketing infrastructure underneath them to do their jobs, and that is where the headcount goes next.

Anthropic is playing a different game entirely. Their Creative Studio model, in-house, craft-first, writer-led, is a bet that brand voice will be a moat when every AI product sounds the same. And they might be right. They just won Ad Age's Best B2B Campaign after building a $30 billion company without running a single ad. When benchmarks converge and feature sets blur, the company that sounds like something, that has a point of view, that earns trust through how it communicates rather than just what it ships, wins. Anthropic is building that from the ground up with humans, not prompts.

Both companies arrived at the same conclusion from different directions: you cannot outsource brand to a model.

It is also worth being honest about the timing. Both OpenAI and Anthropic are reportedly preparing IPOs this year, each with estimated valuations near $1 trillion. You do not go public telling potential investors and enterprise customers that your product will eliminate their workforce. The walkbacks and the $400K creative hires serve that positioning. But that does not make them false. If anything, the fact that the market demands this narrative, that trillion-dollar valuations require acknowledging the limits of automation, tells you something real about where the value sits.

What This Actually Means If You Work in Creative

We are going to say this as plainly as we can, because the nuance matters and most of the discourse is getting it wrong.

The middle of the creative market is getting eaten. The "write me 10 variations of this ad" work, the "produce a monthly content calendar from this keyword list" work, the "turn this product brief into 800 words" work. If that is the core of what you do, the competition is now a model that works for free, never gets tired, and ships in seconds. Klarna learned this when they cut 700 jobs for AI and then had to hire people back because the output was not good enough. But the compression on that layer of work is real, and it is not reversing.

The top of the market, though, is expanding rapidly. The people who can think strategically about brand, who can build creative systems that hold across channels and audiences, who can make judgment calls about voice and positioning that shape how a company is perceived for years, those people have never been more in demand. The proof is in the comp bands: $255K, $320K, $400K, from companies that have more reason than anyone to automate this work and are choosing not to.

Daniela Amodei, Anthropic's president, has a literature degree. She has said publicly that she does not regret it. She said in February that humanities skills will become more valuable as AI gets more capable. "In a world where AI is very smart and capable of doing so many things, the things that make us human will become much more important."

She is not being sentimental. She is running a company that just proved it with a $400K job listing.

Two Truths, No Comfort

The frontier labs are telling you two things at once, and both are true.

AI is coming for the commodity layer of creative work. That is already happening. The BLS still projects 4% growth in writing jobs through 2034, but the composition of those jobs is shifting underneath the topline number in ways the statistics do not capture yet. A Yale Budget Lab study from May found that AI was likely not the primary driver of labor market softening through March 2026 and that there has not been a meaningful spike in unemployment for workers in high-AI-exposure jobs. That sounds reassuring until you realize the disruption is not showing up as mass layoffs. It is showing up as a market bifurcating quietly, the middle hollowing out while the top pulls further away.

The strategic layer of creative work, the brand thinking, the voice architecture, the narrative leadership, is becoming more valuable, not less. The proof is not in a think piece or a LinkedIn hot take. The proof is in the hiring decisions of the two companies that understand AI's capabilities better than anyone alive, and it is now in their own public admissions that they got the doom predictions wrong. The same CEOs who spent a year telling the world that creative jobs were finished are now saying they overshot while posting six-figure salaries for the humans who do the work their models cannot.

If your reaction to all of this is relief, you are reading it wrong. If your reaction is to double down on the strategic, the human, the unteachable parts of creative work, you are reading it exactly right.

Anthropic's Creative Studio job descriptions all end with the same line: "The stakes are real, and there's room to shape what all this becomes."

We agree. The question is whether you are going to be in the room when it gets shaped, or outside it wondering what happened.