AI is often described as a productivity tool. That is true, but incomplete.
A productivity tool helps you do the same work faster. A force multiplier changes what one person can attempt. It compresses the distance between thought and execution.
That compression is why AI feels exciting to some people and threatening to others. It does not merely help people write faster, code faster, analyze faster, or make more things. It changes where the bottleneck lives.
Before AI, the bottleneck was often execution. Could you write the draft? Build the prototype? Produce the report? Turn the idea into something visible?
After AI, execution becomes easier to access. The bottleneck moves upward: can you choose the right problem, ask the right question, judge the answer, and know what good looks like?
"Give me a place to stand, and a lever long enough, and I will move the world." — Archimedes
AI is a force multiplier for judgment. It does not replace judgment. It magnifies it.
Good judgment compounds faster. Bad judgment creates more mistakes at greater speed.
AI Moves the Bottleneck
Every powerful tool changes the bottleneck around it.
The printing press made copying books less scarce. Computers made calculation less scarce. The internet made distribution less scarce. AI is making execution less scarce: the first draft, the prototype, the analysis, the image, the outline.
When something becomes less scarce, status shifts away from merely producing it.
If words are easier to generate, the value moves toward knowing which words are true, useful, and worth saying. If code is easier to generate, the value moves toward knowing what should be built, how it should behave, whether it is maintainable, and whether it belongs in production. If analysis is easier to generate, the value moves toward knowing which question matters, what assumptions are hidden, and what decision the analysis should improve.
This is why AI can feel strange. It rewards people who were not always rewarded by old workplace systems.
The person who can think clearly but could not previously execute fast enough gains leverage. The junior person who learns quickly can ramp up faster. The individual contributor with taste, context, and courage can build more before asking for permission. The domain expert with deep context no longer needs a large team to execute.
AI democratizes execution. But democratized execution does not eliminate hierarchy. It creates a new one.
The new hierarchy is judgment.
The Shift from "Outworking" to Thinking
For years, many professional jobs operated like an exam to be passed. Success was measured by volume, memorization, and visible effort. The programmer who could type code with flawless syntax, deliver feature after feature, or write intimidatingly complex C++ often looked valuable because the system rewarded output that was easy to measure and caught the most attention. In that environment, it was easy to become too attached to work outcomes, using professional intensity as the primary self-identity and sometimes as an excuse to avoid harder personal questions outside work.
Andrew Wilkinson captured the idea neatly:
"Most successful people are just a walking anxiety disorder harnessed for productivity."
That line is funny because it is uncomfortable. A lot of professional intensity is not pure ambition. Sometimes it is anxiety, status fear, and identity protection converted into output.
But AI commoditizes rote execution. The "LeetCode warrior" who treats engineering like a memory test finds that the machine can pass that exam instantly.
When sheer output is no longer a bottleneck, the competitive advantage shifts from the thoughtless doer to the systems thinker. The value moves from the person who can produce a thousand lines of code to the person who knows which hundred lines should exist, how they interact, and whether they solve the actual problem.
The Junior-Senior Compression
This shift compresses the traditional career ladder. Historically, a senior contributor's primary advantage was a massive library of syntax, rules, and execution speed. A junior worker spent years just learning how to translate intent into code or text.
AI changes this timeline. A junior engineer who learns quickly can use AI to draft code and build prototypes at a pace that once took years of experience to match.
This compression forces senior contributors to evolve rapidly. They can no longer rest on the speed of their fingers or their memorization of undocumented rules. To remain valuable, they must move higher up the stack: focusing on judgment, strategy, and domain depth.
Why Resistance Feels Personal
Some resistance to AI is practical and reasonable. The tools make mistakes. They hallucinate. They can produce generic writing, brittle code, shallow analysis, false confidence, and beautiful-looking nonsense.
Those concerns are real.
But not all resistance is really about quality. Sometimes quality is the socially acceptable objection sitting on top of a deeper fear.
AI threatens identity because many people derive meaning, status, and security from being good at the work AI now touches. If you spent years becoming faster, more knowledgeable, or more indispensable inside a particular workflow, it is painful to watch a tool compress part of that workflow into a prompt.
The emotional reaction is understandable. People do not merely lose tasks. They fear losing the story they have been telling themselves about why they matter.
That is why some people want AI to disappear. They want the last few years to have been a strange dream. They want the old rules to return.
But history rarely moves backward just because a new tool makes people uncomfortable.
We have seen this script play out before. When computers and spreadsheets entered the banking sector in the late twentieth century, they did not destroy finance, but they did dismantle the traditional job description of the bank clerk. In financial hubs such as London, Mumbai, and New York, ledger clerks who spent their lives hand-writing transaction ledgers were suddenly face-to-face with automated databases and spreadsheets. In many retail banks, this transition led to massive structural shifts—including Voluntary Retirement Schemes (VRS) and early-retirement packages designed to manage the displacement of a generation whose manual skills were rapidly losing their old scarcity value.
The clerks who adapted were those who stopped resisting the computer and learned to direct the software.
Online learning did something similar to education. YouTube, MOOCs, and low-cost digital courses did not eliminate classrooms, but they weakened the monopoly of the traditional classroom model. A motivated learner could suddenly learn coding, design, finance, statistics, or music from world-class teachers without waiting for a local institution, a fixed timetable, or a gatekeeper business built around access to instruction.
The better question is not whether AI should exist. It does. The better question is what kind of person gains from learning to use it well.
The Weakening of Knowledge Silos
AI also changes the political economy of organizations.
Many companies contain small forts of private context: undocumented processes, hidden dependencies, tribal knowledge, opaque reports, and workflows that only one person or team fully understands. Sometimes these silos exist for innocent reasons. Work moves fast, documentation is hard, and context accumulates unevenly.
Sometimes the silos are strategic.
A person who controls information can control decisions. A team that makes every handoff complicated can protect its importance. A manager who owns a maze can look more valuable than someone who removes the maze.
AI weakens this kind of moat.
It can summarize documents, draft process notes, explain unfamiliar systems, compare alternatives, and help people ask questions they were previously too junior, too new, or too intimidated to ask. It can help the person doing the work make their thinking visible. It can help a small team move without waiting for every old gatekeeper.
This does not mean AI magically destroys politics. Organizations will always have incentives, status games, and coordination problems.
But AI makes it harder to defend a position built only on friction.
If your value came from judgment, taste, trust, and useful context, AI can multiply you. If your value came from being the bottleneck, AI may expose you.
The Flattening of the Management Layer
This erosion of moats has a direct corporate consequence: it flattens organizations.
Many middle managers exist primarily to coordinate execution—to translate high-level strategies into specific tasks, monitor the progress of those tasks, and pass status updates back up the chain. They are information routing stations.
As AI compresses the time between idea and execution, the coordination tax drops. The need for intermediate translators shrinks. Middle managers increasingly have to choose: either move back down the stack to become hands-on builders who can leverage the tool directly, or move up the stack into high-stakes strategic judgment. The era of the pure coordinator is in decline.
Judge the Output, Not the Tool
A new kind of dismissal is becoming common online: "This was written by AI."
You see it on Reddit, forums, social feeds, and comment sections. Someone reads a post and tries to discredit it by pointing at the suspected tool. The implication is that if AI helped produce the piece, the message is somehow automatically weaker.
That is a shallow test.
The better questions are simple:
- Is it true?
- Is it useful?
- Is it clear?
- Does it help someone think, decide, or act better?
The story matters more than the instrument used to tell it.
We do not ask whether a moving novel was typed on a mechanical typewriter, a laptop, or a phone. We ask whether it moved us. We do not judge a photograph only by the camera. We judge the image, the composition, the timing, and the eye behind it.
AI makes authorship messier, but it does not make quality obsolete. If someone uses AI to produce empty, generic, lifeless writing, the problem is not merely the tool. The problem is weak thinking, weak taste, weak editing, or weak intent.
If someone uses AI to express a true story, sharpen an idea, or communicate something valuable, dismissing the result because a tool helped shape it is lazy criticism.
Use AI to Become More Yourself
Nassim Nicholas Taleb has a sharp line in The Bed of Procrustes:
"Another definition of modernity: conversations can be more and more completely reconstructed with clips from other conversations taking place at the same time on the planet."
That line feels even more important in the age of AI.
The danger is not only that AI will make people lazy. The danger is that it will make people interchangeable. If everyone uses the same tools to ask the same questions, summarize the same material, and imitate the same tone, then the world gets more output but less authorship.
That is the wrong way to use leverage.
Use AI to express yourself more uniquely, not less. Use it to compress the mechanical parts of execution so you can spend more time on the parts that are actually yours: the question, the taste, the story, the lived context, the strange connection, the point of view.
This is where AI opens real possibility. One or two people can now attempt work that previously required a much larger team. A small startup can rent parts of a knowledge workforce on demand: a data scientist for analysis, a lawyer for first-pass issue-spotting, a coder for prototypes, a designer for visual exploration, an editor for clarity, a researcher for background.
Those are not perfect substitutes for accountable professionals in high-stakes contexts. But they are powerful enough to change what a small team can try.
Execution is becoming commoditized. That does not make ideas less valuable. It makes original ideas more executable.
The bigger opportunity may be outside the organization entirely. AI lowers the cost of attempting your own ideas — the side project, the indie product, the business you kept postponing because you could not afford to staff it. You can now prototype, draft, test, and iterate on a real idea with a fraction of the resources it once required. The barrier to taking your first shot has never been lower.
The opportunity is not to become a better copier of other people's conversations. The opportunity is to finally build more of your own. Don't deprive the world of the thing only you can make.
AI Also Exposes Shallow Competence
The pro-AI argument has to be honest about the downside.
AI can produce more work, but more work is not the same as better work. It can create code that looks plausible but fails in production. It can write tests that check the wrong thing. It can generate analysis without understanding the business decision. It can produce confident summaries of material it misunderstood.
This is where the phrase "AI slop" is useful.
Slop is not merely low-quality output. It is low-quality output produced at scale, often by people who cannot tell that it is low quality.
A person who does not understand good software architecture can generate bad code faster. A person who does not understand testing can generate bad tests faster. A person who does not understand the domain can generate polished nonsense faster. A person who does not know what production systems require can confuse a demo with a durable product.
AI gives incompetent people more rope. It gives competent people more leverage.
That difference matters.
The winning skill is not blind adoption. It is calibrated adoption: knowing when to delegate, when to inspect, when to reject, when to slow down, and when to do the work yourself.
The J-Curve of AI Adoption
Early AI adoption can feel worse before it feels better.
At first, the tool adds friction. You have to learn how to prompt, how to provide context, how to review answers, how to manage multiple agents, how to prevent hallucinations, and how to avoid drowning in output. You may spend more time correcting the system than using the result.
That is why people say things like:
- "It is faster for me to do it myself."
- "It would have been quicker to talk to a person."
- "Look at these trivial mistakes the AI is making."
Sometimes they are right in the moment.
But judging a tool only during the learning dip is a mistake. Many technologies have a J-curve. Productivity falls at first because the user is learning a new workflow. Then the workflow starts to compound.
The early adopter does not win because the first version is perfect. The early adopter wins because they build taste while the tool is still awkward.
They learn what to ask. They learn what not to trust. They learn where the tool is useful, where it is dangerous, and where it changes the economics of work.
By the time the tool becomes obvious, their judgment has already compounded.
Moving Up the Value Chain
The practical response to AI is not to worship it, nor is it to resist it. The response is to move higher up the value chain.
For decades, knowledge work offered a comfortable shelter: execution. If you could sit at a desk and write lines of code, build dashboards, or draft documentation — keep the business humming along — you were considered productive. You were completing tasks.
AI is dismantling that shelter. When the mechanical cost of execution collapses, the value moves upstream. The bottleneck is no longer how fast you can produce, but how well you can think.
Moving higher up the value chain requires developing and calibrating a distinct set of intellectual disciplines:
- Problem selection (The "Where" of Leverage): In a world of cheap execution, solving the wrong problem faster is a fast track to irrelevance. True leverage is deciding where to point the machine.
- Taste and calibration (The "What" of Leverage): Knowing what "good" looks like. If you cannot distinguish between a fragile prototype and a robust production system, or between shallow copywriting and a compelling argument, you will print slop at scale.
- Strategic questioning (The "How" of Leverage): Formulating prompts and queries that reveal hidden constraints, tradeoffs, incentives, and underlying assumptions.
- The courage of ownership (The "Who" of Leverage): You can delegate execution to a machine, but you can never delegate accountability. The AI is never responsible for the failure. The human remains the sole owner of the outcome.
AI is ultimately a mirror. If you approach it with shallow questions, it will return generic answers. If you approach it with deep context, unique taste, and sharp intent, it will amplify your voice. The opportunity is not to let the machine think for you, but to use the machine so you can think more for yourself.
The question worth carrying forward is not whether AI can produce more. It is whether you are becoming the kind of person who can aim AI at something worth making.