At a glance: The real threat from AI is interchangeability, becoming replaceable within a job that still exists.
Dror Poleg, author and analyst of the nonlinear economy, argues that AI is a medium, not a tool.
A medium packages what you are and broadcasts it at scale. That distinction matters for every experienced professional building an independent practice, because it determines whether AI amplifies your distinctive value or exposes how replaceable you have become.
The economic force driving AI adoption is the push to make every worker interchangeable, dissolving individual bargaining power while preserving the work itself.
For the right professional the opposite happens, and expertise actually appreciates as AI spreads. Independent professionals who control their own AI usage and build practice around the career capital and discernment that cannot be codified are structurally positioned to resist the interchangeability force.
Those who compete on procedures and routine cognitive tasks are sliding toward commodity pricing, regardless of how many years they have been doing the work.
Ask yourself: what kind of practice am I building, and which side of an increasingly unequal income distribution within my profession does it put me on?
Every Tuesday, I write about what I'm learning from deep research and 90+ podcast conversations
How expertise develops, how independent practices sustain themselves, and what the AI era means for people who've spent decades getting good at something.
If that sounds like your kind of reading,
The Wisepreneur lands in your inbox each Tuesday
Dror Poleg is an author and analyst who studies what happens when the dynamics of show business start governing every industry. His recent conversation with Jacob Shapiro on The Jacob Shapiro Podcast, the episode "Warp Speed for Everything" (6 May 2026), covered AI, the future of work, China, cities, and the war in the Middle East.
One argument stood out: AI is better understood as a medium than a tool, and that distinction reshapes every assumption about how experienced professionals should position their practices.
I think Poleg is right about the direction. Where I think he needs extending is on the question of what to do about it, and on why, for the right professional, expertise appreciates with AI rather than eroding.
This matters particularly for professionals over 50 who have built deep career capital over years of practice and are now working independently. His analysis is aimed at investors and tech analysts. The implications for someone building a practice on their own terms are different, and in some ways more encouraging.
What changes when AI is a medium, not a tool
Poleg's central distinction:
- A tool extends what you do
- A medium packages what you are
Recorded music did not help singers sing better. It took a performance, contained it, and broadcast it to a million listeners simultaneously. The number of singers in the world stayed roughly constant after recorded music appeared. What changed was the income distribution: a small number of stars, and a very large number of performers earning little from their music.
AI does the same thing, but for cognitive work. It can take your thinking, your discernment, your approach to problems, and distribute it at scale. People can interact with a version of your expertise without you being present. This is already happening with AI assistants trained on specific bodies of knowledge, with automated advisory services, with AI-generated content that mimics individual voices.
The question this raises for any experienced professional is simple: when the medium packages and broadcasts your professional identity, what exactly gets broadcast?
If what you have built over years is genuine understanding (the capacity to read a situation, diagnose what is actually happening, and construct a response that accounts for context that no dataset captures), then the medium amplifies something irreplaceable.
If what you have built is a set of procedures that look like expertise but could be codified and replicated by anyone with access to the same information, then the medium exposes your replaceability.
The practical test: imagine AI packaging your professional identity and making it available to a thousand potential clients simultaneously. Does the substance of what gets broadcast rest on discernment that took years of real engagement to build? Or does it rest on knowledge and procedures that a well-prompted AI could reproduce next Tuesday?
The music producer Rick Rubin describes the same thing from inside the studio.
"Stack a guitar part enough times", he says, and "you hear guitar, but you don't hear someone playing guitar. It becomes more generic."
Play it once, so you can hear the fingers on the strings, and the personality comes through. AI is a machine for producing the wall of guitars: competent, layered, generic output at limitless scale. What it cannot produce is the single performance in which a person's accumulated discernment is audible.
In a market about to be flooded with the first, the second is the whole of your value.
What doesn't get broadcast
What AI can't access isn't just tacit knowledge in the Polanyi sense. It's something more specific. The most valuable part of what experienced professionals know was never stored as information. It developed as a capacity for reading situations — a sensitivity, calibrated through thousands of professional encounters, to what's actually happening beneath the surface presentation of a problem.
You can't extract that capacity by asking someone to articulate what they know, because it doesn't exist in articulable form. It operates before language. Asking an expert to describe their diagnostic process is like asking a skilled driver to describe what their hands do through a complex manoeuvre.
The description is real, but incomplete — because the most valuable part of the performance isn't happening at the verbal layer.
AI as a tool versus AI as a medium. The shift reframes the question from whether AI replaces a task to what the medium broadcasts.
The interchangeability drive and why it matters more than job replacement
Poleg's sharpest observation: capital does not want to replace you. Capital wants to make you interchangeable.
The entire history of work follows this pattern.
- The artisan becomes the production line worker.
- The creative professional becomes an office worker on a different kind of production line.
- The freelancer becomes a profile on a matching platform where, if the client does not like you, twenty others are one click away.
Each step preserves the work and dissolves the worker's bargaining power.
He uses Hollywood as the cleanest illustration. Early studios refused to print actors' names on posters, a deliberate strategy to keep talent interchangeable. When performers like Mary Pickford became recognisable, they gained pricing power. The studios paid up because stars reduced risk. But as soon as they paid, studio executives started searching for ways to drive that pricing power back down. The logical conclusion is Marvel and Disney, where the intellectual property is the star and the actors playing Spider-Man are replaceable. The value sits in the asset, not the person.
This pattern is now playing out across every profession. AI accelerates it. When your firm's AI system can perform 80% of what you do, your bargaining position changes even if your job title stays the same. You are still employed, but you are more interchangeable than you were last year. And the decisions about how AI gets used in your workflow are being made by someone else, someone with a shorter time horizon and different objectives from yours.
When someone else controls how AI is used in your work, the rational decision from their perspective can erode the very expertise that makes you valuable. Independent professionals who control their own AI usage are structurally protected from this.
Recent research from MIT Sloan (Michael Caosun and Sinan Aral, 2026) formalised this problem. They proved mathematically that when someone other than the worker controls AI usage intensity, the worker's long-term skill level drops below what they would have chosen for themselves. Under illustrative parameters, the gap was 14%. That 14% is the mathematical expression of interchangeability: the employer optimises for output, and the worker's distinctive discernment erodes over time.
The same research proved something else: a professional who controls their own AI usage never falls into this trap.
The alignment of decision-maker and worker is the structural guarantee. This is the independent professional's advantage: a mathematical theorem, not a motivational claim.
Who controls your AI use. When the employer sets AI intensity, skill erodes and you become interchangeable; when you control it, you stay structurally protected
The income distribution question your profession is not asking
Poleg reframes the entire AI employment debate. The number of singers did not change after recorded music.
What changed was who got paid, how much, and how stable that pay was. We do not have to imagine how that looks, because streaming already shows it.
Li Jin's analysis of the creator economy found that on Spotify the top 1.4% of artists take about 90% of the royalties, while the other 98.6% earn an average of around $36 a quarter.
Will Page, Spotify's former chief economist, found that listeners spend 90% of their time on less than 2% of the songs. Cheap production and cheap distribution were supposed to spread the rewards more widely. Instead the hits at the head became more dominant than ever.
Poleg argues this dynamic now governs every profession. If 30 years ago a good lawyer earned five times more than an average lawyer, today it can be 30 times more, and soon it might be 500 times more. The same dynamic applies to consultants, advisors, and every other knowledge professional. The shape that streaming made visible is now forming inside the professions.
Three forces are compounding to make this worse
First, automating entry-level tasks destroys the apprenticeship mechanisms through which deep expertise develops in the next generation.
Second, automating within workflows removes the incidental learning that builds capability beyond the specific task being automated (a financial analyst who manually processes data develops pattern recognition for anomalies; automate the processing and that recognition never forms).
Third, individual professionals whose AI usage is controlled by employers experience gradual skill erosion within their own careers.
The result: the pipeline producing new deep expertise is narrowing at the same time that demand for such expertise is increasing.
Professionals who already hold deep career capital, the rare skills, the relationships, and the body of work that took years to build, are holding an appreciating asset in a market with constrained supply. That is what a moat looks like in this market: not a job title, but capability the medium cannot reproduce.
Professionals who depend on skills that can be codified, taught quickly, or replicated by AI are holding a depreciating asset in a market that is flooding with substitutes.
I think this is the most important question for any experienced professional to sit with. It changes the entire conversation from "will AI take my job?" to "what kind of practice am I building, and which side of this income distribution does it put me on?"
Can't AI just get better?
The more capable AI becomes, the better it gets at producing outputs that look like the products of genuine expertise. That's worth taking seriously. But there's a structural limit on how far this closes the gap that matters.
AI develops by training on the outputs of human intelligence — the language, the recorded decisions, the documented judgments.
Human professional intelligence develops in the opposite direction: years of embodied, consequential practice first, then language as the surface expression of what that practice produced.
These two construction sequences create different internal structures, not just different knowledge stores.
AI trained on the outputs of experienced professional judgment can produce outputs that resemble it. It can't reconstruct the capacity that generated them, because that capacity was never in the outputs to begin with.
Discovery in a world flooded with noise
Poleg is candid about a problem that complicates the picture. AI will flood every market with content, analysis, and advice that is competent enough to pass a quick inspection. Discovery becomes harder even for genuinely capable professionals. The algorithms that mediate discovery feed whatever is already popular, so the already-visible get more visible, and the richer get richer by design. Two equally talented people can end up with radically different economic outcomes because one happened to get algorithmic amplification and the other did not.
This matters for anyone building an independent practice. Your value may be real, but potential clients are wading through more noise than ever to find you.
The way discovery works in a medium-driven economy is through relational connection, reputation, referral, and demonstrated expertise that builds trust before any commercial conversation begins.
Podcasts do this. Sustained writing does this. A clear articulation of who you serve and why your career capital matters to them specifically does this. It is slower than chasing reach, and it is far harder for anyone to commoditise.
David C. Baker, who consults with expertise-based firms on positioning, has a useful test: can you build a targeted campaign reaching your specific audience? If you cannot describe your audience precisely enough to find them, your positioning is too broad. In a noisy market, broad positioning means invisibility among the AI-generated alternatives that are all saying roughly the same thing.
Where Poleg's analysis ends and the independent professional's strategy begins
Poleg predicts the entire economy is heading toward what he provocatively calls "a variation of OnlyFans," where everyone sells emotional attention and personalised value to each other.
He includes himself in this description:
his listeners can get similar analysis elsewhere, but they choose him for the personal connection, the individual voice, the specific way he thinks through problems.
I think he is describing something real, but the framing needs adjusting for experienced independent professionals.
You are operating in what I would call the understanding economy. The relational element matters because trust, discernment, and genuine accountability can only be delivered through a relationship.
A client who needs someone to diagnose what is actually going wrong in their business, or to help them articulate the value they cannot see in their own expertise, needs a person who has spent years developing that capacity.
Karl Fast's research, which is an important influence on my Wisepreneurs approach, puts it plainly: information is cheap, understanding is expensive. That cost reduction is what experienced professionals provide, and it is inseparable from the person delivering it. Poleg's frame is scale and winner-take-most dynamics. His audience is investors.
For an experienced professional building practice, the objective is different. You do not need to become a superstar.
You need a sustainable practice serving the right clients at premium positioning, with enough visibility to maintain your pipeline and enough depth to justify your fees. That requires clarity about what you offer that cannot be replicated, and the discipline to keep building the cognitive capacity that sustains it.
The independent professional's structural advantage in the nonlinear economy is real. When you control your own AI usage, choose which problems to engage with, and build practice around discernment that only years of real engagement can produce, you are positioned where the interchangeability drive cannot reach.
Four questions to sit with
If Poleg's analysis is right (and I think the direction is sound), these are the questions worth taking seriously for anyone building an independent practice around accumulated professional expertise.
- What would AI broadcast if it packaged your professional identity today? Is the substance of what you offer grounded in discernment and career capital that took years to build, or in knowledge and procedures that could be replicated quickly?
- Where is the interchangeability pressure in your profession? Even if your specific role is secure, how is the income distribution within your profession changing? Are you positioned where deep career capital commands its value, or where the market is being flooded with substitutes?
- Who controls how AI is used in your work? If you are employed, someone else is making decisions about your AI workflow with different objectives from yours. If you are independent, you have the structural alignment that the MIT research identified as the protection against skill erosion.
- How do the right clients find you in a noisier market? Can you describe your audience precisely enough to reach them directly? Is your positioning specific enough to cut through AI-generated noise, or are you competing in a space where everything sounds the same?
These are not questions you need to answer alone. Almost every successful independent professional I have spoken with across 90 episodes of the Wisepreneurs Podcast has had help with their thinking: a coach, a mentor, a strategic partner, someone who asks the questions you cannot ask yourself when you are too close to your own expertise.
Frequently asked questions
How does AI make professionals interchangeable rather than replacing them?
Capital's goal has always been to make workers interchangeable, preserving the work while dissolving the worker's bargaining power. AI accelerates this by performing routine cognitive tasks, reducing the pricing power of professionals whose value rests on codifiable skills. Research from MIT Sloan demonstrates that when employers control AI usage in a worker's workflow, the worker's long-term expertise erodes. Independent professionals who control their own AI usage maintain their distinctive discernment because they balance productivity gains against preserving the expertise those gains depend on.
Why is the income distribution question more important than whether AI takes jobs?
Recorded music did not reduce the number of singers. It changed who got paid and how much. Streaming made the pattern stark: on Spotify the top 1.4% of artists take about 90% of royalties while the rest average around $36 a quarter. AI is bringing the same shape to every profession. The income gap between average and top performers is widening because the pipeline producing new deep expertise is being disrupted while demand for genuine career capital increases. Experienced independent professionals hold an appreciating asset in a market with constrained supply.
What is the augmentation trap and how do independent professionals avoid it?
The augmentation trap is the condition where rational AI adoption erodes the expertise it depends on. Research by Michael Caosun and Sinan Aral at MIT Sloan proves mathematically that the trap only binds when someone other than the worker controls AI usage intensity. When a professional controls their own usage (as independent practitioners do), they never fall into the trap. The alignment of decision-maker and worker provides structural protection against capability erosion.
How can experienced professionals maintain visibility in an AI-saturated market?
In a market flooded with AI-generated content and advice, discovery depends on relational connection, reputation, referral, and demonstrated expertise. Precise positioning (knowing exactly who you serve and articulating why your career capital matters to them specifically) becomes essential. Podcasts, sustained writing, and professional networks that build trust before any commercial conversation begins are the discovery mechanisms that work in what Poleg calls the nonlinear economy.
Why can't AI keep improving until it can replicate experienced professional judgment?
AI capability is advancing fast, and it's worth taking the question seriously. But the two processes produce structurally different results: AI develops by training on the outputs of human intelligence — the language, recorded decisions, documented judgments.
Human professional intelligence develops in the opposite direction, years of embodied practice first, then language as the surface expression of that. Different construction sequences produce different internal structures, not just different knowledge stores. AI can produce outputs that closely resemble experienced professional judgment. It cannot reconstruct the capacity that generated them, because that capacity was never in the outputs to begin with.
References
Dror Poleg: Author and analyst. "Warp Speed for Everything," The Jacob Shapiro Podcast (6 May 2026). Poleg's newsletter and writing at drorpoleg.com, including his essay on AI and the long tail.
Michael Caosun and Sinan Aral: MIT Sloan Management. Research on the augmentation trap: when AI usage erodes expertise. Published 2026.
Li Jin: "Building the Middle Class of the Creator Economy." Source of the Spotify royalty distribution figures: the top 1.4% of artists take about 90% of royalties, while the rest average around $36 per quarter.
Shaaron L. Macrine, Jennifer M. B. Fugate, Asif Abdulali, and Josh Hughes (Eds.): Embodied Intelligence: Multidisciplinary Perspectives on Natural, Artificial, and Hybrid Systems. MIT Press, 2026. The cognitive science grounding for the two sections on why expert knowledge resists extraction (capacity built through practice, not stored as information) and why AI's construction sequence cannot reproduce it.
Will Page: Former chief economist at Spotify and author of Tarzan Economics. Source of the finding that listeners spend 90% of their time on less than 2% of available songs.
David C. Baker: Author of The Business of Expertise (2017). Positioning and pricing framework for expertise-based firms. punctuation.com
Karl Fast: Co-author (with Stephen P. Anderson) of Figure It Out: Getting from Information to Understanding. The "information is cheap, understanding is expensive" formulation that informs the Wisepreneurs approach.