Vector 1
Organizations & Work

The Personas Who Will Run Circles Around Everyone Else in the AI-First Future

Brian J. Curry
Brian J. Curry
April 20, 2026 · 38 min read

Why the 50-year-old with a messy career and an AI subscription is about to outperform the 25-year-old prodigy — and the data proving it’s already happening

There is a quiet revolution happening in the labor market that almost nobody is talking about honestly.

It is the inversion of one of the most durable myths in the knowledge economy — the belief that the young, by virtue of being young, are better positioned to use new technology than the old. For forty years this was Silicon Valley’s operating theology. Hire the 22-year-old with a computer science degree and a willingness to sleep under their desk. Age out the 50-year-old with two mortgages and a family. Call it “culture fit” and “grit” and “digital nativism” and a dozen other euphemisms for what it actually was — a preference for cheap, interchangeable throughput over expensive, accumulated wisdom.

The AI-first future is tearing this framework apart, and the data is no longer ambiguous.

In August 2025, a team of Stanford economists led by Erik Brynjolfsson — arguably the most prominent labor economist working on AI — published a working paper that should have been front-page news for a week. Using payroll data from ADP covering roughly 25 million American workers, they found that employment for workers aged 22 to 25 in the most AI-exposed occupations — software development, customer service, entry-level analytics — had fallen by 13 percent since late 2022. In the same occupations, employment among workers aged 30 and older had grown by 6 to 9 percent. Brynjolfsson put it bluntly in the Wall Street Journal: “There’s a clear, evident change when you specifically look at young workers who are highly exposed to AI.” His colleague Bharat Chandar called it “the canary in the coal mine.”

This is not a story about displacement. This is a story about redistribution — of output, of value, of economic leverage — toward people with the one thing AI cannot manufacture at any price.

Context.

The future of work in an AI-first world belongs to the people with the largest, deepest, messiest, most associative mental models of how the world actually operates. And those people, for the most part, are not 25 years old. They have careers that zigzagged through three industries. They have watched at least two technology cycles rise and collapse. They have been wrong enough times to recognize the shape of being wrong. And when they finally sit down with a state-of-the-art AI tool, they do not ask it to write a function — they ask it to help them architect a system, and they know within three seconds whether its answer is any good.

This is the persona that will win the AI-first future. Let me describe who they are, why the data supports it, and what this means for how organizations should be thinking about talent right now.

The Collapsing Value of Book Knowledge

Start with why the Stanford data looks the way it does.

Brynjolfsson offered the cleanest explanation of the mechanism in an interview with CBS MoneyWatch: large language models are trained on books, articles, and written material scraped from the internet — “that’s the kind of book learning that a lot of people get at universities before they enter the job market, so there is a lot of overlap between these LLMs and the knowledge young people have.”

This is worth sitting with for a moment.

The entire value proposition of hiring a recent college graduate in a knowledge-work role has historically rested on the assumption that their book learning was scarce. A computer science graduate from a top-ten school had, in their head, a reasonably comprehensive model of data structures, algorithms, system design patterns, and the syntax of several languages. That model was expensive to produce, took four years and six figures of tuition to construct, and was therefore worth paying a starting salary of $85,000 to rent access to.

Large language models have reduced the marginal cost of that knowledge to roughly zero.

Anyone with a $20-a-month Claude or ChatGPT subscription now has on-demand access to a synthesized version of every textbook ever written, every Stack Overflow answer ever posted, and every technical blog post ever indexed. The junior developer’s core product — “I know things from books” — has been commoditized overnight. What has not been commoditized is everything the junior developer does not yet have: pattern recognition across failed projects, intuition about which stakeholders will kill an initiative six months in, a felt sense of when a vendor is lying, the ability to recognize that the problem the customer described is not actually the problem the customer has.

None of this is in any textbook. None of it can be scraped from the open web. It lives in the heads of people who have been doing the work for twenty or thirty years, and its value — precisely because AI has commoditized everything else — has just gone vertical.

What McKinsey Is Telling Its Clients (Quietly)

The McKinsey consulting machine, which has a financial interest in being on the right side of this shift, has begun publishing the quiet part out loud.

In November 2025, McKinsey released an analysis titled Redesigning technology workforce for the agentic AI era that reads less like management consulting and more like an obituary for the conventional tech org chart. The key passage, which almost nobody has written about:

“In our experience, with agents as tools, expert technologists can be several times more productive than less experienced peers, and the pay differential between them is modest relative to the output gap. Companies that continue to hire for volume rather than expertise risk inflating costs without increasing impact.”

Read that again.

McKinsey is telling its clients — the largest employers in the world — that the fundamental economics of engineering team composition have inverted. The classic pyramid model (one senior architect supervising a platoon of mid-level implementers supervising an army of juniors) was a rational response to the bottleneck of senior labor. Seniors cost three or four times what juniors cost and could not physically type fast enough to produce the volume of code a system required. So you paid the premium for their judgment and rented out the typing to less expensive hands.

AI agents have removed the typing bottleneck. A senior engineer with a well-orchestrated set of agents can now produce, in a day, what used to take a team of five a week. The pay differential between that senior engineer and a junior is — as McKinsey notes — modest relative to the new output gap. The entire pyramid is economically upside down.

The same McKinsey research, drawing on their State of AI in 2025 survey of nearly 2,000 organizations, found that only about 6 percent of companies qualify as “AI high performers” — the subset that has captured more than 5 percent of EBIT from AI and whose leaders report significant value. What separates them is not technology. It is organizational design, and specifically, the presence of senior people who know what to point the AI at. The technology is broadly available. The context to wield it well is not.

McKinsey’s parallel workforce research identifies two emerging archetypes who dominate the new org chart:

M-shaped supervisors: broad generalists fluent in AI, orchestrating agents and hybrid human-AI teams across multiple domains. These are people with wide context windows — exposure to multiple industries, functions, and problem types. They tend to be mid-to-late career.

T-shaped experts: deep specialists who reimagine workflows, handle the exceptions AI cannot handle, and safeguard quality. Their depth was built over decades.

Neither archetype describes a 24-year-old.

The Daily-User Pay Premium Goes to the People Who Already Know Things

The argument that older workers are technophobic Luddites who will be left behind by AI is empirically wrong, and the surveys now bear this out.

PwC’s 2025 Global Workforce Hopes and Fears Survey, released in November 2025 and covering nearly 50,000 workers across 48 countries, found that daily generative AI users are dramatically outperforming infrequent users across every dimension that matters:

  • 92 percent of daily users report tangible productivity gains, versus 58 percent of infrequent users.
  • 58 percent report higher job security, versus 36 percent.
  • 52 percent report higher salaries tied to AI use, versus 32 percent.

The question is which workers are in that high-productivity, high-pay, high-security daily-user cohort. McKinsey’s 2025 Superagency in the Workplace report provides the uncomfortable answer. When respondents self-reported AI expertise, 62 percent of workers aged 35 to 44 claimed high expertise — the highest of any age group. Only 50 percent of 18-to-24-year-old Gen Z workers claimed the same. The “digital native” advantage, if it ever existed, does not survive contact with professional AI use.

The reason is not mysterious. Comfort with consumer technology — swiping, tapping, posting — is not the same skill as orchestrating an AI system to produce professional-grade output. The latter requires knowing what good output looks like in your domain, which requires having produced a lot of output in that domain. You cannot prompt your way to expertise you do not already have. The AI can only amplify the signal you bring to it. A 25-year-old with three years of work experience brings a weak signal. A 50-year-old with twenty-five years across three industries brings a signal you can see from space.

Stack Overflow’s 2026 developer pulse survey added an illuminating data point: experienced developers are significantly more likely than early-career developers to consult technical documentation alongside AI, rather than treating AI as their first and only source. Only 1 percent of developers across all cohorts rely on AI alone. The experienced ones know that AI is a starting point, not an ending point — and they have the reference frameworks to know when it’s wrong.

The Cognitive Debt Problem (And Why Your Experience Is a Hedge Against It)

This is the part that turns a sociological observation into something closer to a biological one.

In June 2025, a team of researchers at the MIT Media Lab led by Nataliya Kosmyna published a working paper titled Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task. The methodology was straightforward: 54 participants were divided into three groups — one writing essays with ChatGPT, one writing with only Google search, and one writing with no tools at all — while EEG sensors measured neural activity across 32 brain regions over multiple sessions.

The results, reported across outlets from TIME to The Guardian, were striking. The ChatGPT group showed the weakest neural connectivity and the lowest brain engagement. The brain-only group showed the strongest, most distributed networks. Perhaps most importantly, over successive sessions, the ChatGPT group got worse — their engagement declined, their originality declined, and by the final essays many participants were simply copying and pasting whatever the model produced. The researchers coined the term “cognitive debt” for what was happening: a feedback loop in which AI dependence progressively erodes the cognitive capacities that make AI use valuable in the first place.

The implication for this discussion is pointed.

If you are 22 years old and you have spent your formative learning years outsourcing your thinking to an AI tool, you may be accumulating cognitive debt in real time. You never built the base layer of mental models that AI is supposed to amplify. Your “signal” is thin not because you are inherently less capable but because you never had the generative struggle that produces durable expertise.

If you are 50 years old, you spent thirty years building that base layer before AI existed. The cognitive patterns that give you pattern-matching ability, cross-domain analogical thinking, and domain-specific intuition were forged in an environment where there was no shortcut. You accumulated mental equity. AI now lets you deploy that equity across a much larger surface area.

This is the paradox at the heart of the age-in-AI story: the very people Silicon Valley spent two decades edging out of the workforce are the only people who built the cognitive infrastructure AI actually rewards.

The Six Traits of the Persona That Wins

If you want a clinical description of the human being who is about to dominate the AI-first future — whether you are trying to become that person, hire that person, or build an organization of them — the traits are identifiable and largely measurable.

One: A long and varied career history, including at least one major pivot. Cross-domain experience is the single best predictor of associative thinking, which is the single best way to prompt AI effectively. A marketer who became a data scientist, a physician who became a product manager, a lawyer who learned to code — these people have wider context windows than any single-domain specialist, no matter how brilliant. The World Economic Forum’s Future of Jobs Report 2025 projects that 39 percent of core workplace skills will shift by 2030, and the workers who handle that transition best are the ones who have already handled a pivot before.

Two: A strong intuition for what “good” looks like in their domain. This is the piece AI cannot supply and cannot replace. When the model produces a financial model, a legal argument, a marketing brief, a piece of code — someone has to know whether it’s any good. That knowledge is earned through production, failure, and correction over years. It is the quality control system of the AI-augmented organization, and it resides only in humans who have done the work long enough to have taste.

Three: Pattern recognition across cycles. The person who lived through the dot-com bust, the 2008 financial crisis, the mobile transition, the cloud transition, and the crypto bubble has a pattern-recognition library for technology cycles that a 26-year-old simply does not have access to. This shows up in AI deployment decisions. The experienced worker is less likely to fall for vendor hype, more likely to demand a measurable business case, and more likely to recognize when an AI initiative is actually solving a real problem versus generating impressive-looking theater.

Four: Associative creative thinking. Palantir CEO Alex Karp has argued publicly that AI will reward two types of workers — the vocationally skilled and the unconventionally-minded. The second group skews older not because older people are inherently more creative, but because creativity as an applied skill requires a large stock of patterns, references, and analogies to combine. You cannot be a creative generalist at 24 because you have not yet accumulated enough to generalize from. Neurodivergence and cross-domain career breadth are both amplifiers of this trait, and both are disproportionately present in the over-40 workforce.

Five: The ability to articulate tacit knowledge. Michael Polanyi famously observed that “we can know more than we can tell” — the tacit knowledge problem. The new twist in the AI-first era is that the ability to tell what you know is suddenly enormously valuable, because tacit knowledge you can articulate can be encoded into AI workflows, prompts, evaluation rubrics, and agent instructions that then scale indefinitely. The experienced worker who can explain why they do what they do becomes an expertise multiplier inside an organization. Their individual throughput is no longer the ceiling; their articulated judgment, once embedded in systems, operates at machine scale.

Six: Comfort with being wrong. This one is counterintuitive but turns out to be decisive. People who have been in the workforce long enough to have made major professional mistakes — shipped the wrong product, hired the wrong executive, invested in the wrong technology — tend to develop a specific kind of intellectual humility that is extremely useful when working with AI systems. They do not over-trust the model, because they have been burned by false confidence before. They instinctively verify, cross-check, and look for the failure mode. Early-career workers who have never experienced a major professional reversal are significantly more likely to accept an AI output at face value, with consequences that show up downstream in bad decisions.

None of these traits is perfectly correlated with chronological age. There are 28-year-olds who have all six and 55-year-olds who have none. But at the population level, these traits accumulate with time-in-the-workforce, and the AI-first economy systematically rewards them in a way the previous economy did not.

The Uncomfortable Implications for Organizations

For anyone running an organization right now, the strategic implications of this shift are significant and underappreciated.

The pyramid is upside down. If McKinsey’s observation is right — that senior people with AI agents produce several times the output of juniors at modest pay premiums — then the old leverage model of one senior directing ten juniors is economically irrational. The new model is one senior directing ten AI agents, with one or two junior humans in apprenticeship roles focused on learning rather than production. This is a dramatic reduction in headcount and a dramatic increase in experienced-worker concentration.

The entry-level pipeline is breaking. The Stanford data shows this in real time: 22-to-25-year-olds are being hired at dramatically lower rates in AI-exposed fields, while 30-plus workers in the same fields are being hired at higher rates than before. This is the early stage of a crisis that organizations should be actively managing. If you do not hire anyone for their first job, where do your 35-year-olds come from in ten years? Forward-thinking organizations are already moving to an apprenticeship model — small numbers of juniors brought in explicitly to learn from seniors rather than to produce — but most organizations are simply cutting junior headcount and hoping the problem goes away. It will not.

Age-biased hiring practices are now a direct competitive disadvantage. The ProPublica investigations of the late 2010s documented systematic age discrimination in tech — the IBM case alone involved roughly 20,000 workers over 40 pushed out in a five-year window. Every one of those workers, had they been retained and equipped with modern AI tools, would now be producing at levels their younger replacements cannot match. Companies that maintained those biases are now structurally short on the most valuable asset in the AI economy: articulated senior judgment. Companies that did not are ahead without realizing why.

Retention of experienced talent has become the single most important HR metric. When each senior employee with AI tools produces the output of what used to require a team, their departure is not a routine turnover event — it is the loss of a major production asset. The compensation differential required to retain that person is trivial compared to the output gap their replacement will exhibit for years. Organizations that continue to treat experienced employees as cost centers to be managed down are systematically destroying their own capacity.

The “culture fit” signals are getting expensive. The office perks, the hoodie-and-ping-pong aesthetic, the implicit signal that the company is built around 28-year-olds — these are actively repellent to exactly the cohort whose contribution has become most valuable. Companies that have not noticed this yet will notice when their AI-augmented competitors, staffed with 45-year-old ex-consultants who work remotely from their third-bedroom home offices, begin eating their lunch.

What This Means If You Are the Persona

If you recognize yourself in this article — if you are a worker with a long career, scar tissue from previous technology transitions, and a growing sense that the industry has been trying to tell you you are obsolete for a decade — the honest message is this: the game has just changed in your favor, and almost nobody has noticed yet.

The premium on your experience has never been higher. The tools to deploy it have never been more accessible. The $20-a-month AI subscription in your pocket is the first time in the history of the knowledge economy that an individual contributor has had access to leverage previously reserved for people running teams. If you can build the habit of daily use — and the PwC survey makes clear that daily use is where the real productivity and compensation gains live — you become the rare combination of deep signal and broad amplification.

A few practical moves.

Treat AI fluency as a first-priority skill investment, regardless of your domain. The data on the compensation premium for daily users is unambiguous. Do not delegate this to younger colleagues or assistants. The person who does the prompting gets the leverage.

Begin articulating your tacit knowledge in writing. Everything you “just know” about how your industry works is now an asset that can be encoded into an AI system. Write it down. Build a document, a Notion page, a personal knowledge base that captures the rules-of-thumb, the red flags, the pattern libraries you have internalized over your career. This is the raw material for every prompt you will ever write, and it becomes more valuable the longer it gets.

Look for organizations explicitly building for the post-pyramid model. The signals are there if you know how to read them. Companies that emphasize human-plus-agent workflows, that are reducing junior headcount without expanding senior headcount proportionally, that talk about “orchestration” rather than “execution” — those are the companies that understand what is happening. They will pay for your experience rather than discount it.

Refuse to frame yourself as an anachronism. The most common mistake experienced workers make right now is internalizing the Silicon Valley narrative that youth equals technological fluency. It was wrong when it was popular and it is wronger now. Your experience is not a liability to be hidden; it is the single most valuable input the AI-first economy rewards.

The Signal and the Amplifier

Here is the metaphor I keep returning to.

For most of human history, knowledge workers were production units limited by the hours in a day. The smartest, most experienced professionals were constrained by the same 24-hour cap as the least experienced. You could be the most brilliant litigator in New York, but you could only be in so many depositions per week. You could be the sharpest strategist at a consulting firm, but you could only write so many decks. The glass ceiling on human output was not cognitive — it was physical.

AI has removed that cap.

What you now produce is a function not of your typing speed but of the quality of the signal you generate — and signal quality is built over decades. The experienced worker’s signal is rich, contextual, cross-referenced, analogical, and specific. The inexperienced worker’s signal is shallow, narrow, and derivative. Before AI, both produced roughly similar volumes, because both were limited by human bandwidth. After AI, the rich-signal worker’s output scales up hundreds of times while the thin-signal worker’s output scales up by volume but not by value. One becomes a hub in the network. The other produces high-volume noise.

This is the inversion. This is the shift almost nobody in the tech press is willing to say clearly yet.

The AI-first future does not belong to the young. It does not belong to the fastest typist, the most energetic all-nighter, the freshest graduate with the cleanest LinkedIn. It belongs to the person who has seen enough, done enough, been wrong enough, and built enough mental equity to know what to point the machine at.

It belongs to the master, not the apprentice.

The 50-year-old with a messy career, a well-organized prompt library, and twenty-five years of hard-won judgment is about to run circles around the 25-year-old prodigy. The data is already showing it. The only question is how long it takes the rest of the economy to catch up to what is right there in the payroll records.

Sources and References

  • Brynjolfsson, E., Chandar, B., & Chen, R. (2025). Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence. Stanford University working paper, August 2025.
  • PwC. 2025 Global Workforce Hopes and Fears Survey. Released November 12, 2025. Survey of 49,843 workers across 48 countries.
  • McKinsey & Company. Redesigning technology workforce for the agentic AI era. November 2025.
  • McKinsey & Company. Superagency in the Workplace: Empowering people to unlock AI’s full potential at work. January 2025.
  • McKinsey & Company. The State of AI in 2025: Agents, innovation, and transformation. 2025 survey of 1,993 respondents across 105 countries.
  • McKinsey & Company. The agentic organization: Contours of the next paradigm for the AI era. September 2025.
  • Kosmyna, N., et al. (2025). Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task. MIT Media Lab, arXiv:2506.08872.
  • Stack Overflow. Domain expertise still wanted: the latest trends in AI-assisted knowledge for developers. March 2026 pulse survey.
  • World Economic Forum. Future of Jobs Report 2025.
  • Microsoft Research. New Future of Work Report 2025. December 2025.
  • National CIO Review. The Evolution of Expertise in an AI-Driven Enterprise. 2026.
  • Axios. “AI is already taking jobs away from entry-level workers.” August 26, 2025.
  • Fortune. “First-of-its-kind Stanford study says AI is starting to have a ‘significant and disproportionate impact’ on entry-level workers in the U.S.” August 26, 2025.
  • TIME. “Who’s Losing Jobs to AI? New Stanford Analysis Breaks It Down.” August 2025.
  • CBS News. “New study sheds light on what kinds of workers are losing jobs to AI.” August 28, 2025.
  • ProPublica. Investigation into IBM age discrimination, 2018.

This article is part of an ongoing series on AI as an amplifier of human capability. Related pieces: “AI Is a 100X Multiplier — But 100 Times Zero Is Still Zero” and “The Age of the Master: How AI Is About to Destroy Silicon Valley’s Obsession with Youth.”

Brian Curry is a Kansas City–based data scientist, AI researcher, and founder of Vector1 Research, an independent lab advancing marketing economics, causal inference, and cognitive AI systems. He has led data and analytics initiatives at enterprises from $4B to $125B, including Koch Industries, Tractor Supply, Vail Resorts, Hallmark, Garmin, and AT&T. He is the creator of MeaningFlow (semantic content modeling), Papilon (causal inference and complex systems), PyCausalSim (causal discovery through simulation), and the Memory-Node Encapsulation (MNE) architecture for cognitive AI.

There is a quiet revolution happening in the labor market that almost nobody is talking about honestly.

It is the inversion of one of the most durable myths in the knowledge economy — the belief that the young, by virtue of being young, are better positioned to use new technology than the old. For forty years this was Silicon Valley’s operating theology. Hire the 22-year-old with a computer science degree and a willingness to sleep under their desk. Age out the 50-year-old with two mortgages and a family. Call it “culture fit” and “grit” and “digital nativism” and a dozen other euphemisms for what it actually was — a preference for cheap, interchangeable throughput over expensive, accumulated wisdom.

The AI-first future is tearing this framework apart, and the data is no longer ambiguous.

In August 2025, a team of Stanford economists led by Erik Brynjolfsson — arguably the most prominent labor economist working on AI — published a working paper that should have been front-page news for a week. Using payroll data from ADP covering roughly 25 million American workers, they found that employment for workers aged 22 to 25 in the most AI-exposed occupations — software development, customer service, entry-level analytics — had fallen by 13 percent since late 2022. In the same occupations, employment among workers aged 30 and older had grown by 6 to 9 percent. Brynjolfsson put it bluntly in the Wall Street Journal: “There’s a clear, evident change when you specifically look at young workers who are highly exposed to AI.” His colleague Bharat Chandar called it “the canary in the coal mine.”

This is not a story about displacement. This is a story about redistribution — of output, of value, of economic leverage — toward people with the one thing AI cannot manufacture at any price.

Context.

The future of work in an AI-first world belongs to the people with the largest, deepest, messiest, most associative mental models of how the world actually operates. And those people, for the most part, are not 25 years old. They have careers that zigzagged through three industries. They have watched at least two technology cycles rise and collapse. They have been wrong enough times to recognize the shape of being wrong. And when they finally sit down with a state-of-the-art AI tool, they do not ask it to write a function — they ask it to help them architect a system, and they know within three seconds whether its answer is any good.

This is the persona that will win the AI-first future. Let me describe who they are, why the data supports it, and what this means for how organizations should be thinking about talent right now.

The Collapsing Value of Book Knowledge

Start with why the Stanford data looks the way it does.

Brynjolfsson offered the cleanest explanation of the mechanism in an interview with CBS MoneyWatch: large language models are trained on books, articles, and written material scraped from the internet — “that’s the kind of book learning that a lot of people get at universities before they enter the job market, so there is a lot of overlap between these LLMs and the knowledge young people have.”

This is worth sitting with for a moment.

The entire value proposition of hiring a recent college graduate in a knowledge-work role has historically rested on the assumption that their book learning was scarce. A computer science graduate from a top-ten school had, in their head, a reasonably comprehensive model of data structures, algorithms, system design patterns, and the syntax of several languages. That model was expensive to produce, took four years and six figures of tuition to construct, and was therefore worth paying a starting salary of $85,000 to rent access to.

Large language models have reduced the marginal cost of that knowledge to roughly zero.

Anyone with a $20-a-month Claude or ChatGPT subscription now has on-demand access to a synthesized version of every textbook ever written, every Stack Overflow answer ever posted, and every technical blog post ever indexed. The junior developer’s core product — “I know things from books” — has been commoditized overnight. What has not been commoditized is everything the junior developer does not yet have: pattern recognition across failed projects, intuition about which stakeholders will kill an initiative six months in, a felt sense of when a vendor is lying, the ability to recognize that the problem the customer described is not actually the problem the customer has.

None of this is in any textbook. None of it can be scraped from the open web. It lives in the heads of people who have been doing the work for twenty or thirty years, and its value — precisely because AI has commoditized everything else — has just gone vertical.

What McKinsey Is Telling Its Clients (Quietly)

The McKinsey consulting machine, which has a financial interest in being on the right side of this shift, has begun publishing the quiet part out loud.

In November 2025, McKinsey released an analysis titled Redesigning technology workforce for the agentic AI era that reads less like management consulting and more like an obituary for the conventional tech org chart. The key passage, which almost nobody has written about:

“In our experience, with agents as tools, expert technologists can be several times more productive than less experienced peers, and the pay differential between them is modest relative to the output gap. Companies that continue to hire for volume rather than expertise risk inflating costs without increasing impact.”

Read that again.

McKinsey is telling its clients — the largest employers in the world — that the fundamental economics of engineering team composition have inverted. The classic pyramid model (one senior architect supervising a platoon of mid-level implementers supervising an army of juniors) was a rational response to the bottleneck of senior labor. Seniors cost three or four times what juniors cost and could not physically type fast enough to produce the volume of code a system required. So you paid the premium for their judgment and rented out the typing to less expensive hands.

AI agents have removed the typing bottleneck. A senior engineer with a well-orchestrated set of agents can now produce, in a day, what used to take a team of five a week. The pay differential between that senior engineer and a junior is — as McKinsey notes — modest relative to the new output gap. The entire pyramid is economically upside down.

The same McKinsey research, drawing on their State of AI in 2025 survey of nearly 2,000 organizations, found that only about 6 percent of companies qualify as “AI high performers” — the subset that has captured more than 5 percent of EBIT from AI and whose leaders report significant value. What separates them is not technology. It is organizational design, and specifically, the presence of senior people who know what to point the AI at. The technology is broadly available. The context to wield it well is not.

McKinsey’s parallel workforce research identifies two emerging archetypes who dominate the new org chart:

M-shaped supervisors: broad generalists fluent in AI, orchestrating agents and hybrid human-AI teams across multiple domains. These are people with wide context windows — exposure to multiple industries, functions, and problem types. They tend to be mid-to-late career.

T-shaped experts: deep specialists who reimagine workflows, handle the exceptions AI cannot handle, and safeguard quality. Their depth was built over decades.

Neither archetype describes a 24-year-old.

The Daily-User Pay Premium Goes to the People Who Already Know Things

The argument that older workers are technophobic Luddites who will be left behind by AI is empirically wrong, and the surveys now bear this out.

PwC’s 2025 Global Workforce Hopes and Fears Survey, released in November 2025 and covering nearly 50,000 workers across 48 countries, found that daily generative AI users are dramatically outperforming infrequent users across every dimension that matters:

  • 92 percent of daily users report tangible productivity gains, versus 58 percent of infrequent users.
  • 58 percent report higher job security, versus 36 percent.
  • 52 percent report higher salaries tied to AI use, versus 32 percent.

The question is which workers are in that high-productivity, high-pay, high-security daily-user cohort. McKinsey’s 2025 Superagency in the Workplace report provides the uncomfortable answer. When respondents self-reported AI expertise, 62 percent of workers aged 35 to 44 claimed high expertise — the highest of any age group. Only 50 percent of 18-to-24-year-old Gen Z workers claimed the same. The “digital native” advantage, if it ever existed, does not survive contact with professional AI use.

The reason is not mysterious. Comfort with consumer technology — swiping, tapping, posting — is not the same skill as orchestrating an AI system to produce professional-grade output. The latter requires knowing what good output looks like in your domain, which requires having produced a lot of output in that domain. You cannot prompt your way to expertise you do not already have. The AI can only amplify the signal you bring to it. A 25-year-old with three years of work experience brings a weak signal. A 50-year-old with twenty-five years across three industries brings a signal you can see from space.

Stack Overflow’s 2026 developer pulse survey added an illuminating data point: experienced developers are significantly more likely than early-career developers to consult technical documentation alongside AI, rather than treating AI as their first and only source. Only 1 percent of developers across all cohorts rely on AI alone. The experienced ones know that AI is a starting point, not an ending point — and they have the reference frameworks to know when it’s wrong.

The Cognitive Debt Problem (And Why Your Experience Is a Hedge Against It)

This is the part that turns a sociological observation into something closer to a biological one.

In June 2025, a team of researchers at the MIT Media Lab led by Nataliya Kosmyna published a working paper titled Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task. The methodology was straightforward: 54 participants were divided into three groups — one writing essays with ChatGPT, one writing with only Google search, and one writing with no tools at all — while EEG sensors measured neural activity across 32 brain regions over multiple sessions.

The results, reported across outlets from TIME to The Guardian, were striking. The ChatGPT group showed the weakest neural connectivity and the lowest brain engagement. The brain-only group showed the strongest, most distributed networks. Perhaps most importantly, over successive sessions, the ChatGPT group got worse — their engagement declined, their originality declined, and by the final essays many participants were simply copying and pasting whatever the model produced. The researchers coined the term “cognitive debt” for what was happening: a feedback loop in which AI dependence progressively erodes the cognitive capacities that make AI use valuable in the first place.

The implication for this discussion is pointed.

If you are 22 years old and you have spent your formative learning years outsourcing your thinking to an AI tool, you may be accumulating cognitive debt in real time. You never built the base layer of mental models that AI is supposed to amplify. Your “signal” is thin not because you are inherently less capable but because you never had the generative struggle that produces durable expertise.

If you are 50 years old, you spent thirty years building that base layer before AI existed. The cognitive patterns that give you pattern-matching ability, cross-domain analogical thinking, and domain-specific intuition were forged in an environment where there was no shortcut. You accumulated mental equity. AI now lets you deploy that equity across a much larger surface area.

This is the paradox at the heart of the age-in-AI story: the very people Silicon Valley spent two decades edging out of the workforce are the only people who built the cognitive infrastructure AI actually rewards.

The Six Traits of the Persona That Wins

If you want a clinical description of the human being who is about to dominate the AI-first future — whether you are trying to become that person, hire that person, or build an organization of them — the traits are identifiable and largely measurable.

One: A long and varied career history, including at least one major pivot. Cross-domain experience is the single best predictor of associative thinking, which is the single best way to prompt AI effectively. A marketer who became a data scientist, a physician who became a product manager, a lawyer who learned to code — these people have wider context windows than any single-domain specialist, no matter how brilliant. The World Economic Forum’s Future of Jobs Report 2025 projects that 39 percent of core workplace skills will shift by 2030, and the workers who handle that transition best are the ones who have already handled a pivot before.

Two: A strong intuition for what “good” looks like in their domain. This is the piece AI cannot supply and cannot replace. When the model produces a financial model, a legal argument, a marketing brief, a piece of code — someone has to know whether it’s any good. That knowledge is earned through production, failure, and correction over years. It is the quality control system of the AI-augmented organization, and it resides only in humans who have done the work long enough to have taste.

Three: Pattern recognition across cycles. The person who lived through the dot-com bust, the 2008 financial crisis, the mobile transition, the cloud transition, and the crypto bubble has a pattern-recognition library for technology cycles that a 26-year-old simply does not have access to. This shows up in AI deployment decisions. The experienced worker is less likely to fall for vendor hype, more likely to demand a measurable business case, and more likely to recognize when an AI initiative is actually solving a real problem versus generating impressive-looking theater.

Four: Associative creative thinking. Palantir CEO Alex Karp has argued publicly that AI will reward two types of workers — the vocationally skilled and the unconventionally-minded. The second group skews older not because older people are inherently more creative, but because creativity as an applied skill requires a large stock of patterns, references, and analogies to combine. You cannot be a creative generalist at 24 because you have not yet accumulated enough to generalize from. Neurodivergence and cross-domain career breadth are both amplifiers of this trait, and both are disproportionately present in the over-40 workforce.

Five: The ability to articulate tacit knowledge. Michael Polanyi famously observed that “we can know more than we can tell” — the tacit knowledge problem. The new twist in the AI-first era is that the ability to tell what you know is suddenly enormously valuable, because tacit knowledge you can articulate can be encoded into AI workflows, prompts, evaluation rubrics, and agent instructions that then scale indefinitely. The experienced worker who can explain why they do what they do becomes an expertise multiplier inside an organization. Their individual throughput is no longer the ceiling; their articulated judgment, once embedded in systems, operates at machine scale.

Six: Comfort with being wrong. This one is counterintuitive but turns out to be decisive. People who have been in the workforce long enough to have made major professional mistakes — shipped the wrong product, hired the wrong executive, invested in the wrong technology — tend to develop a specific kind of intellectual humility that is extremely useful when working with AI systems. They do not over-trust the model, because they have been burned by false confidence before. They instinctively verify, cross-check, and look for the failure mode. Early-career workers who have never experienced a major professional reversal are significantly more likely to accept an AI output at face value, with consequences that show up downstream in bad decisions.

None of these traits is perfectly correlated with chronological age. There are 28-year-olds who have all six and 55-year-olds who have none. But at the population level, these traits accumulate with time-in-the-workforce, and the AI-first economy systematically rewards them in a way the previous economy did not.

The Uncomfortable Implications for Organizations

For anyone running an organization right now, the strategic implications of this shift are significant and underappreciated.

The pyramid is upside down. If McKinsey’s observation is right — that senior people with AI agents produce several times the output of juniors at modest pay premiums — then the old leverage model of one senior directing ten juniors is economically irrational. The new model is one senior directing ten AI agents, with one or two junior humans in apprenticeship roles focused on learning rather than production. This is a dramatic reduction in headcount and a dramatic increase in experienced-worker concentration.

The entry-level pipeline is breaking. The Stanford data shows this in real time: 22-to-25-year-olds are being hired at dramatically lower rates in AI-exposed fields, while 30-plus workers in the same fields are being hired at higher rates than before. This is the early stage of a crisis that organizations should be actively managing. If you do not hire anyone for their first job, where do your 35-year-olds come from in ten years? Forward-thinking organizations are already moving to an apprenticeship model — small numbers of juniors brought in explicitly to learn from seniors rather than to produce — but most organizations are simply cutting junior headcount and hoping the problem goes away. It will not.

Age-biased hiring practices are now a direct competitive disadvantage. The ProPublica investigations of the late 2010s documented systematic age discrimination in tech — the IBM case alone involved roughly 20,000 workers over 40 pushed out in a five-year window. Every one of those workers, had they been retained and equipped with modern AI tools, would now be producing at levels their younger replacements cannot match. Companies that maintained those biases are now structurally short on the most valuable asset in the AI economy: articulated senior judgment. Companies that did not are ahead without realizing why.

Retention of experienced talent has become the single most important HR metric. When each senior employee with AI tools produces the output of what used to require a team, their departure is not a routine turnover event — it is the loss of a major production asset. The compensation differential required to retain that person is trivial compared to the output gap their replacement will exhibit for years. Organizations that continue to treat experienced employees as cost centers to be managed down are systematically destroying their own capacity.

The “culture fit” signals are getting expensive. The office perks, the hoodie-and-ping-pong aesthetic, the implicit signal that the company is built around 28-year-olds — these are actively repellent to exactly the cohort whose contribution has become most valuable. Companies that have not noticed this yet will notice when their AI-augmented competitors, staffed with 45-year-old ex-consultants who work remotely from their third-bedroom home offices, begin eating their lunch.

What This Means If You Are the Persona

If you recognize yourself in this article — if you are a worker with a long career, scar tissue from previous technology transitions, and a growing sense that the industry has been trying to tell you you are obsolete for a decade — the honest message is this: the game has just changed in your favor, and almost nobody has noticed yet.

The premium on your experience has never been higher. The tools to deploy it have never been more accessible. The $20-a-month AI subscription in your pocket is the first time in the history of the knowledge economy that an individual contributor has had access to leverage previously reserved for people running teams. If you can build the habit of daily use — and the PwC survey makes clear that daily use is where the real productivity and compensation gains live — you become the rare combination of deep signal and broad amplification.

A few practical moves.

Treat AI fluency as a first-priority skill investment, regardless of your domain. The data on the compensation premium for daily users is unambiguous. Do not delegate this to younger colleagues or assistants. The person who does the prompting gets the leverage.

Begin articulating your tacit knowledge in writing. Everything you “just know” about how your industry works is now an asset that can be encoded into an AI system. Write it down. Build a document, a Notion page, a personal knowledge base that captures the rules-of-thumb, the red flags, the pattern libraries you have internalized over your career. This is the raw material for every prompt you will ever write, and it becomes more valuable the longer it gets.

Look for organizations explicitly building for the post-pyramid model. The signals are there if you know how to read them. Companies that emphasize human-plus-agent workflows, that are reducing junior headcount without expanding senior headcount proportionally, that talk about “orchestration” rather than “execution” — those are the companies that understand what is happening. They will pay for your experience rather than discount it.

Refuse to frame yourself as an anachronism. The most common mistake experienced workers make right now is internalizing the Silicon Valley narrative that youth equals technological fluency. It was wrong when it was popular and it is wronger now. Your experience is not a liability to be hidden; it is the single most valuable input the AI-first economy rewards.

The Signal and the Amplifier

Here is the metaphor I keep returning to.

For most of human history, knowledge workers were production units limited by the hours in a day. The smartest, most experienced professionals were constrained by the same 24-hour cap as the least experienced. You could be the most brilliant litigator in New York, but you could only be in so many depositions per week. You could be the sharpest strategist at a consulting firm, but you could only write so many decks. The glass ceiling on human output was not cognitive — it was physical.

AI has removed that cap.

What you now produce is a function not of your typing speed but of the quality of the signal you generate — and signal quality is built over decades. The experienced worker’s signal is rich, contextual, cross-referenced, analogical, and specific. The inexperienced worker’s signal is shallow, narrow, and derivative. Before AI, both produced roughly similar volumes, because both were limited by human bandwidth. After AI, the rich-signal worker’s output scales up hundreds of times while the thin-signal worker’s output scales up by volume but not by value. One becomes a hub in the network. The other produces high-volume noise.

This is the inversion. This is the shift almost nobody in the tech press is willing to say clearly yet.

The AI-first future does not belong to the young. It does not belong to the fastest typist, the most energetic all-nighter, the freshest graduate with the cleanest LinkedIn. It belongs to the person who has seen enough, done enough, been wrong enough, and built enough mental equity to know what to point the machine at.

It belongs to the master, not the apprentice.

The 50-year-old with a messy career, a well-organized prompt library, and twenty-five years of hard-won judgment is about to run circles around the 25-year-old prodigy. The data is already showing it. The only question is how long it takes the rest of the economy to catch up to what is right there in the payroll records.

Sources and References

  • Brynjolfsson, E., Chandar, B., & Chen, R. (2025). Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence. Stanford University working paper, August 2025.
  • PwC. 2025 Global Workforce Hopes and Fears Survey. Released November 12, 2025. Survey of 49,843 workers across 48 countries.
  • McKinsey & Company. Redesigning technology workforce for the agentic AI era. November 2025.
  • McKinsey & Company. Superagency in the Workplace: Empowering people to unlock AI’s full potential at work. January 2025.
  • McKinsey & Company. The State of AI in 2025: Agents, innovation, and transformation. 2025 survey of 1,993 respondents across 105 countries.
  • McKinsey & Company. The agentic organization: Contours of the next paradigm for the AI era. September 2025.
  • Kosmyna, N., et al. (2025). Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task. MIT Media Lab, arXiv:2506.08872.
  • Stack Overflow. Domain expertise still wanted: the latest trends in AI-assisted knowledge for developers. March 2026 pulse survey.
  • World Economic Forum. Future of Jobs Report 2025.
  • Microsoft Research. New Future of Work Report 2025. December 2025.
  • National CIO Review. The Evolution of Expertise in an AI-Driven Enterprise. 2026.
  • Axios. “AI is already taking jobs away from entry-level workers.” August 26, 2025.
  • Fortune. “First-of-its-kind Stanford study says AI is starting to have a ‘significant and disproportionate impact’ on entry-level workers in the U.S.” August 26, 2025.
  • TIME. “Who’s Losing Jobs to AI? New Stanford Analysis Breaks It Down.” August 2025.
  • CBS News. “New study sheds light on what kinds of workers are losing jobs to AI.” August 28, 2025.
  • ProPublica. Investigation into IBM age discrimination, 2018.

This article is part of an ongoing series on AI as an amplifier of human capability. Related pieces: “AI Is a 100X Multiplier — But 100 Times Zero Is Still Zero” and “The Age of the Master: How AI Is About to Destroy Silicon Valley’s Obsession with Youth.”

Brian Curry is a Kansas City–based data scientist, AI researcher, and founder of Vector1 Research, an independent lab advancing marketing economics, causal inference, and cognitive AI systems. He has led data and analytics initiatives at enterprises from $4B to $125B, including Koch Industries, Tractor Supply, Vail Resorts, Hallmark, Garmin, and AT&T. He is the creator of MeaningFlow (semantic content modeling), Papilon (causal inference and complex systems), PyCausalSim (causal discovery through simulation), and the Memory-Node Encapsulation (MNE) architecture for cognitive AI.

Related research