The last decade's advice came down like gospel.
Learn to code.
Parents repeated it. Universities sold it. Mid-career professionals swallowed it whole and signed up for online classes. STEM was the future. The humanities were a luxury. Nice for conversation, useless for survival.
Now the ground has shifted.
The coders are watching parts of their jobs get automated.
The English majors are getting interesting phone calls.
The Wrong Bet
T
he corporate world spent fifteen years drawing a straight line from technical skill to economic value. Code ran the systems, the systems ran the business, and the people who could build them got paid.
Everyone else scrambled to keep up. Add Python to the resume. Take a quick course. Get “data literate.”
The logic felt airtight. Too bad it was never permanent.
MIT economist David Autor pointed this out this decades ago when computers and spreadsheets began started eating bookkeeping and clerical work. Routine tasks disappeared. Non-routine thinking went up in value. Judgment, interpretation, and human interaction became the scarce resource.
Same pattern now but at a larger scale and faster pace.
With AI now going after anything that can be reduced to pattern, sequence, and output, it turns out a remarkable amount of what passed for “knowledge work” can now be done by chips.
What Actually Separates People Now
A
recent field experiment led by researchers from MIT, Tulane, and Rice gave 250 employees at a technology consulting firm access to ChatGPT for a week. Only 26 percent saw meaningful improvement in their output.
The rest saw nothing useful from it.
The difference had nothing to do with technical skill. It came down to how people think about their own thinking.
Some questioned the output. Pushed it. Refined it. Treated it like a draft, not an answer.
Others took the first result and moved on.
One group got leverage while the other got noise dressed up as insight.
Metacognition’s Moment
Met
acognition isn’t some shiny new concept cooked up in a venture lab. It’s been around, lurking in the bloodstream, waiting for a moment like this when the machines started talking back. Now it’s finally getting dragged into the daylight where it belongs.
It’s looking at an AI-generated draft and groaning, “Clean, polished, and dead wrong,” and then rolling up your sleeves and starting to cut and revise.
Metacognition asks the uncomfortable questions: What game are we actually playing here? Is this thing doing real work or just impersonating it? What’s missing, and why does it feel like something important just slipped out the back door?
Microsoft’s Chief Scientist Jaime Teevan put it in plain English in an interview with the The Wall Street Journal. Metacognitive skills are going to matter. Flexibility. Adaptability. The willingness to experiment, to challenge, to think critically when everyone else is nodding along to something that smells off. And here’s the part that makes the spreadsheet crowd nervous. A good old liberal arts education turns out to be rocket fuel for that kind of thinking.
Because nobody builds this muscle by sitting through a tidy little training module. You earn it the hard way. Four years in the arena, getting your ideas ripped to pieces in writing seminars that feel more like bare-knuckle fights than classrooms. Arguing about whether Hamlet is a coward or a philosopher while someone across the table is trying to take your head off with a better argument. Learning to stand your ground, then having the sense to abandon it when the facts start closing in. Reading fiction until you can crawl inside someone else’s worldview and see the cracks forming before they speak.
That kind of perspective doesn’t come out of a data structures class. Not even close.
And this is where the split happens in the wild. The real operators versus the hall monitors. The CIO who can actually move a business from the one who manages a process; the difference I've described elsewhere as the Translator versus the Traffic Cop. The former being the one who listens to understand what's actually being survived while the latter forwards it to the steering committee.
The Machines Are Hiring Storytellers
The
clearest signal in any market is money.
The companies building AI are choosing to feed and grow their communications teams instead of letting them die on the vine of their magical machines.
Anthropic has expanded aggressively to a team of almost 80 and is still hiring communicators with $200,000 starting salaries. OpenAI is posting similar comms roles that clear $400,000. LinkedIn job postings mentioning “storyteller” have surged. Venture firms are building media arms to help founders shape narratives, and Netflix put out a communications role with a $750,000 ceiling that would make most CFOs blink.
For some perspective, the national average for those communications roles sits closer to $100,000.
Nobody pays a 4x premium to schedule press calls. They are paying for people who can take something complex, incomplete, and often confusing and turn it into something that makes sense to another human being.
Sasha de Marigny, Anthropic's first chief communications officer, described to Business Insider what she's actually looking for: "Critical thinking is still a huge comparative advantage for humans. I'm looking for excellent strategists. People who understand the new world order and know how to develop plans to cut through to the audiences we care about."
Anthropic's own job posting for an Editorial, Economic and Societal Impacts role says it requires a "flair for new ways to communicate complex ideas," the ability to "find compelling narratives within technical work," and "critical thinking applied to all aspects of communications."
The company that built Claude cannot prompt Claude into doing this. That's the whole point.
The Judgment Economy
We ar
e sliding, fast and a little sideways, into what Nils Gilman in Noema calls the judgment economy, where nobody pays you for how much you can crank out anymore, they pay you for whether you can make a call that holds up when the lights come on.
Three crucial judgment skills crawl out of the wreckage of this machine-fed economy like survivors who know the terrain better than the map ever did.
First, the dark art of lining people up and getting them to move in the same direction without a threats. Persuasion, negotiation, a little backroom diplomacy, the ability to herd a pack of loud, overpaid skeptics toward a single decision while they’re still arguing about the menu.
Second, the instinct to read a data-soaked landscape without freezing up like a rookie waiting for perfect information that never shows. You look at the mess, trust your gut, and call the play anyway.
Third, the real trick, the cross-pollination of ideas that shouldn’t logically belong together but somehow crack the whole thing open. Seeing how a move in one industry rewires another, how a flicker of consumer behavior bends a technical system, how a nagging ethical itch forces a product team to rethink the entire roadmap. That’s the game now, and it does not reward the timid.
These talents have historically gotten labeled as soft skills.
There is nothing soft about them.
Follow the Leaders
Anthro
pic co-founder Daniela Amodei took one look at the machine age barreling toward us and said the quiet part out loud to ABC News: the more silicon we stack, the more human stuff starts to matter.
Meanwhile, her co-conspirator and co-founder Jack Clark, an English major loose in a room full of engineers, told Ezra Klein at The New York Times that the real scarcity isn’t code or compute, it’s judgment, taste, instinct, that gut-level sense of what’s worth doing before the spreadsheet shows up with excuses. The machines can crank and flood the zone with output, but they can’t tell you what matters without an experienced human in the loop willing to make the call and own the consequences.
The people building, owning, and running the machines are telling you this. The question is whether you're listening with enough care to act on it.
Taste and Intuition Are the Scarce Resources
None of
this is a nostalgia tour for liberal arts degrees. Technical skill still matters and anyone who can't think clearly about data is going to struggle in these times.
But the specific thing the last decade got wrong, badly and at scale, was its contempt for the skills that can't be easily credentialed. The skills that live in the capacity to read a room, hold a position under pressure, tell the difference between an answer that's technically correct and one that's actually true. To know what good looks like without being handed a rubric.
Those sometimes intrinsic skills were built in seminars and in regular argument and in reading things that resisted easy summary. They were built by people who were told, repeatedly, that they were choosing a path without a future.
I know something about this from a business direction most corporate arguments don't travel. After graduating with a History degree and before joining corporate conference rooms, I was in the music business, marketing, promoting, and making records for Def Jam and Columbia Records at a time when not a soul in the building had an MBA. Many hadn’t even attended college.
Lyor Cohen and his team didn’t grow Def Jam to unimagined heights because he had an MBA. Don Ienner and the rest of Big Red didn’t build Columbia’s most successful decade with a credentialed degrees on walls.
They ran those places on taste, intuition, judgment, and grit.
They knew what would hit and what would die on arrival. They could feel the difference between something real and something manufactured to look real. They were wrong sometimes, but they were wrong fast and corrected.
The whole operation ran on judgment and taste refined by proximity to the real thing.
Meanwhile, the last decade dismissed much that couldn’t be easily measured, quantified, or proved. If it did not fit in a dashboard, if it didn’t have an avalanche of data, if it didn't come from smart, highly credentialed, but ultimately inexperienced Big 3 consultants it didn’t count.
But the internet has a long history of taking convention and turning it on its head. This time builing a $34B influencer economy on attention and taste. Which is its own kind of proof.
The influencers who have built durable audiences and real commercial leverage are not the most credentialed. They are the ones with a original points of view. The ones who know what is good before the metrics confirm it. The ones who can communicate that instinct in a way others recognize.
AI can’t manufacture that. It can approximate taste's surface features in the same way it can approximate a professional's output: accurately enough to fool someone who can't tell the difference, and badly enough to embarrass someone who can.
It cannot decide what is worth choosing.
That requires a human who knows what good looks like.
The Gap Where the Work Lives
The iron
y is that the tool those people were supposedly made obsolete by has turned out to be the most powerful advertisement for what they actually learned.
Generative AI cannot know, without someone telling it, whether the story it just told is the right one.
Of course, AI will keep getting better at producing. That was never the hard part. The hard part is knowing what to keep, what to discard, and what to push further. The hard part is holding a contra position when the easy answer is sitting in front of you, polished and ready to ship.
That gap is where the work lives now.
The people who can operate in that gap are not the ones who learned to produce faster.
They are the ones who learned how to think.
The traffic cops are still standing in the intersection, waving things through.
The people with intuition, taste, and judgment are already down the road, making calls that move the business.
And they are not waiting for permission.