Average Is Over
The most dangerous place in the AI economy is the comfortable middle.
Quick note before we start.
I’m an official ambassador for B2B Marketing Expo US, and if you’re in B2B, this one is worth your time.
We’re sponsoring it this year, which made this edition feel especially timely.
📍 Miami Beach Convention Center 📅 April 29-30, 2026
100+ exhibitors. 50 speakers. 5,500+ attendees.
Two days of real conversations, practical sessions, and people who are actually building things, not just talking about it.
It’s free to attend for subscribers of this newsletter.
See you in Miami.
Hi,
Over the last few weeks, I’ve heard the same sentence in different voices. Sometimes from founders. Sometimes from marketers. Sometimes from people who are objectively smart and successful and still somehow manage to say it with complete confidence:
“I know AI matters, but I still think most of it is hype.”
That sentence fascinates me. Not because it is stupid. Because it is half true.
And half-truths are dangerous. They let you feel intelligent while standing still.
Yes, AI hallucinates. Yes, the internet is filling with slop. Yes, a lot of people are confusing a prompt with a skill and a demo with a business. Yes, there is a weird religious tone in parts of the discourse, like every new release is either the second coming or the end of civilisation.
All true.
But markets do not reward you for correctly identifying a tool’s flaws. They reward you for recognising when a flawed tool still changes the game.
You can be completely right about AI’s weaknesses and still be completely wrong about what it is doing to the market.
That is where a lot of smart people are getting trapped right now. They are being correct in theory while becoming obsolete in practice.
And that brings me to the real point.
Average is over.
Not in the motivational poster sense. Not in the “just hustle harder” sense. In the Tyler Cowen sense. In the economic sense. In the brutal sense that the comfortable middle is losing its old protection while machine-amplified people pull away from everyone else.
The comfortable middle just lost its insurance policy
For a long time, being reasonably good at your job bought you a decent life. You did not need to be elite. You did not need to be world class. You just needed to be competent, reliable, and present. That was enough to stay in the safe middle.
That middle is getting carved out.
A while ago, I wrote
At the time, I framed it as a mindset problem, a warning about complacency, a reminder that “doing okay” can be more dangerous than outright failure because failure forces honesty while mediocre success lets you hide.
Now I think it was accidentally an economic forecast.
Because mediocre success used to be survivable. Now it is being repriced in public. And AI is the thing speeding that process up.
For years, average had an insurance policy. Slow companies paid the premium. Bloated org charts kept renewing it. Messy ownership blurred accountability. Meetings disguised weak output. Information moved slowly enough that ordinary competence could still look exceptional.
You could survive by being useful-ish. By coordinating instead of creating. By formatting instead of deciding. By summarising instead of owning. By sounding smart in rooms where nobody was really measuring leverage.
That was enough.
It is not enough anymore.
The people in the most danger are not the obvious ones
A lot of people still imagine this transition incorrectly. They think the first people in trouble are the clearly obsolete ones, the people doing obviously repetitive work, the people whose roles already sound fragile.
Some of them are exposed, yes.
But I do not think that is the most dangerous place to be.
The most dangerous place is the polished middle. Good salary. Good title. Good meetings. Good reputation. Enough output to look solid. Not enough leverage to be undeniable.
Because AI does not need to replace you fully to destroy your economics. It only needs to make the person next to you faster. Clearer. More productive. Less dependent on extra headcount. Less dependent on time. Less dependent on permission.
You are not competing with AI. You are competing with the person who learned to use it before you did.
That is the part people still do not want to hear. They think criticism is a shield. It isn’t. It is often camouflage for delay.
Tyler Cowen wrote the memo years ago
I went back to Average Is Over by Tyler Cowen, and it hit differently this time. Less like economics. More like a warning people filed in the wrong section.
His point was never that every human becomes obsolete. His point was that the people who learn to complement machines get rewarded, while the people stuck in the generic middle get squeezed. High performers become more leveraged. The comfortable average loses altitude.
Back then, this sounded like a sharp economic thesis. In 2026, it sounds like a hiring memo.
One of the most useful frames in the book is freestyle chess. For a while, the strongest unit was not human alone. It was not machine alone. It was human plus machine, assuming the human knew what they were doing. That does not feel like a chess metaphor anymore. It feels like the job description of the next decade.
The cage is open. The tools are real. The models are getting dependable.
But once the cage opens, average loses its protection. That is the harsher conclusion nobody wants to sit with.
Why marketing matters more, not less
One of Cowen’s most underrated calls was that marketing would rise in relative importance. A lot of people hear that and think it sounds backwards. Shouldn’t AI make marketing cheaper, easier, less important?
Only if you misunderstand what is happening.
AI makes production cheaper. And when production gets cheaper, distribution matters more. When content becomes infinite, attention matters more. When everyone can write, positioning matters more. When everybody can generate “good enough”, taste becomes commercially valuable.
Taste is not decoration anymore. It is filtration at scale. It is decision quality. It is knowing what not to publish. It is knowing which message deserves repetition and which one deserves deletion.
And this is why B2B should care early. B2B has always been a game of trust, narrative clarity, and risk reduction. You are not just selling software or service. You are selling confidence. You are selling fewer bad decisions. You are selling a cleaner future than the one your buyer is stuck inside now.
AI does not kill that. It magnifies it.
It makes mediocre marketing cheaper. But it makes great marketing more valuable. Because when noise becomes abundant, signal gets expensive.
What gets overlooked here: most B2B companies are still treating AI as a content factory, a way to produce more. The real edge is the opposite, using AI to produce less, but sharper. The companies that figure out filtration before amplification will own their categories.
You are not competing against AI
This is where most people still lose the plot.
You are not competing against ChatGPT. You are competing against a human being who knows how to use AI better than you do. That person might be younger than you. Cheaper than you. Less experienced than you. And still more dangerous.
Why? Because the market does not pay for nostalgia. It pays for output. It pays for speed. It pays for judgment. It pays for people who can turn tools into outcomes without turning the result into generic sludge.
Think of it like two photographers with the same camera. Same lens. Same light. One produces art. The other produces stock photos nobody remembers. The difference was never the equipment. It was always the eye behind it.
Human with machine plus taste versus human with machine plus none.
The upside is not becoming “the AI guy” or “the AI girl.” The upside is becoming the person who can orchestrate research, writing, systems, analysis, and execution faster than traditional roles were ever designed to move.
Work is still the loop. Iteration still wins. Discipline still matters. Hard work did not die. It just got rerouted.
Hard work still compounds. Low-leverage hard work does not.
That is a different sentence. And a very important one.
Busy is not a moat
One of the strangest bugs in modern work culture is that people still think busyness is a form of protection. A full calendar feels important. A flooded inbox feels important. A thousand Slack messages feel important.
But often that is not proof of value. It is proof that your work is made of fragments. And fragments are exactly what AI compresses first.
If your week is mostly made of status updates, note clean-up, deck polishing, surface-level research, reporting, first drafts, coordination, scheduling, chasing people, and moving information from one place to another, then you are not in a safe zone. You are in a compression zone.
That does not mean you are useless. It means your current mix of work is vulnerable. There is a difference. And that difference matters because it tells you what to do next.
Panic is useless. Repricing is useful.
An enormous amount of modern white-collar work is not sacred craft. It is information movement with a nice title on top. And information work is where AI gets vicious.
The marketer who can go from rough idea to positioning to landing page to campaign assets in one afternoon will beat the team that needs three meetings just to agree on a brief. The operator who can map a process, automate the repetitive middle, and keep human judgment where it matters will start to look superhuman. Not because they became a genius overnight. Because they stopped doing everything at the same altitude.
The anti-average audit
You do not need another vague instruction to “learn AI.” You need a harsher mirror.
Ask yourself these questions honestly:
If a sharp competitor in my field had the same tools I have and 90 days of focused experimentation, how much of my current output could they replicate?
Can I explain, in plain English, how my work affects revenue, margin, trust, retention, product speed, or pipeline?
What part of my work depends on judgment, not just execution?
What still takes me four hours that should only take one?
Am I known for a title, or for solving a specific painful problem?
Have I turned any of my recent learning into visible proof?
Sit with those longer than feels comfortable. That discomfort is data.
What to do now
I am not interested in giving you a cute framework here. This is what I would actually do.
1. Compress the boring 30 percent. The biggest waste of capable models is treating them like novelties. Pick one repetitive workflow tied to something real: research synthesis, client reporting, outbound prep, proposal writing, meeting-to-action conversion, content repurposing, competitor monitoring, internal knowledge retrieval. Then compress it aggressively. Not for fun. For leverage.
2. Reinvest, do not relax. This is where people mess up. They save two hours and then donate them to distraction. Bad trade. Reinvest the saved time into work that is harder to copy: customer conversations, sharper offers, better positioning, better taste, product decisions, relationship building, actual thinking. The win is not “AI saved me time.” The win is “I moved my time into a higher-value layer.”
3. Turn hidden learning into public proof. Quiet capability compounds slower than visible proof. This is one of the biggest errors smart people still make. They improve privately and stay invisible publicly. Then they wonder why the market is not repricing them. Publish the before and after. Show the workflow. Write the lesson. Tell people what changed. You do not need to become a content machine. You do need evidence.
4. Become known for a problem, not a title. “I work in marketing” is weak. “I help B2B companies turn confusing demand into clear pipeline” is stronger. “I manage operations” is weak. “I remove friction from how revenue teams move” is stronger. Titles belong to internal org charts. Problems travel. Problems sell. Problems compound into reputation.
5. Stack taste on top of tools. Tools are becoming abundant. Taste is not. Judgment is not. Restraint is not. The ability to say “this is not good enough” is not. When everybody can generate, the people who can curate, combine, reject, and refine become disproportionately valuable.
This is not a doom memo
I know this essay sounds harsh. Good. It should.
But it is not a doom memo. It is an opportunity memo disguised as a warning.
Because if average is over, then leverage is just getting started. That is the part pessimists keep missing.
The same forces that remove protection from vague work also create enormous upside for clear thinkers, fast builders, sharp communicators, and people who know how to turn technology into better business reality.
The distance between idea and execution is collapsing. Companies are getting smaller, sharper, and less sentimental. And the middle can be more dangerous than the bottom because it sedates you.
Put those together and the message becomes very clear.
This is a brutal era for protected average. It is also an incredible era for people who know how to build proof, own a problem, and work with these systems instead of arguing with the calendar.
When AI starts swallowing parts of work, it does not just threaten output. It threatens identity. That is why so many reactions to AI feel emotional before they feel analytical. People are not only asking, “Will this help?” They are also asking, quietly:
“Who am I if the thing I do becomes easier, cheaper, or partly automatable?”
That is a real question. But the answer is not to cling harder to the old task stack. The answer is to climb a level. Own the system. Own the direction. Own the composition. Own the judgment. Own the final call.
Bottom line
Average is over.
Not because humans stopped mattering. Because average human output stopped being protected.
That is the real sentence.
So I would stop asking whether AI is overhyped. That question is too lazy now.
A better question is this:
What part of my work becomes more valuable because machines exist?
That is the question that leads somewhere. That is the question that creates positioning, not panic. That is the question that turns fear into design.
And the people who answer it first will take an unreasonable share of the upside. Everybody else will stay very busy explaining why the old system should still count.
It will not.
And if you want to see what this shift looks like in the wild, not as a Twitter argument but as a room full of people already building around it, come to Miami.
Post-credit scene
📖 Read
Average Is Over by Tyler Cowen. Start here. Not as economics. As a user manual for the repricing of work.
Co-Intelligence by Ethan Mollick. A practical companion if you want less abstract philosophy and more real-world AI operating mode.
🎧 Listen
Forecast 2050: Tyler Cowen on AI, Fake Jobs, and the Next 25 Years. Fresh conversation from March 2026. Cowen believes this will be the fastest technological revolution in recent world history. Worth your commute.
Odd Lots: Tyler Cowen on Why AI Hasn’t Changed the World Yet. A good counter-balance if you want the bull case tempered with cost disease economics and honest timelines.
🎬 Watch
Severance (Season 1 & 2, Apple TV+). The most relevant show on television for what modern work does to identity when systems get stronger than self-awareness. Season 3 filming starts this spring.
Moneyball. For what happens when better models beat legacy intuition. Rewatch it with fresh eyes after reading this edition.
🔁 Revisit
Thanks for reading.
Vlad