The Worth Gap: Who You Are When AI Does the Work

A founder I have known for years sent me a message at midnight last winter. He had just shipped the best week of product work of his life, three features he would have billed a small team a quarter to build, done solo with a model open in one window. And he wrote: “I have never felt less necessary.” That line stayed with me, because I have felt the exact same thing, and because it is the quietest crisis in building right now and almost nobody names it correctly.

We talk about this moment as a jobs story. Through the middle of 2026, companies tied roughly 90,000 layoffs to AI, and the surveys read like a slow alarm. ADP’s People at Work 2026, covering about 39,000 workers across 36 countries, found that only 22 percent feel confident their job is safe from AI elimination. Gallup’s 2026 read put 18 percent of US employees expecting their role to be gone in five years, rising to 23 percent inside companies that have already deployed AI tools. One survey of hiring managers had 70 percent saying their people are afraid of becoming, in the exact word, useless.

That word is the tell. Useless is not an economic term. It is an identity term. The real story under the jobs story is that a lot of us quietly fused the question “what am I worth” to the question “what can I do,” and AI just pried those two apart with a crowbar. The doing got cheap. The sense of worth did not move. The space between them is what I call the worth gap, and if you are building anything right now, closing it is more urgent than any tool you will adopt this year.

The work got cheap and your identity did not get the memo

For most of human history, the value of what you could do and your standing as a person tracked each other so closely that nobody bothered to separate them. Doing was scarce. If you could write clean code, translate a legal contract, draw a brand, or close a hard account, that capability was rare, it took years to earn, and the market and your own self-image both priced it high. The fusion was invisible because the two numbers were always the same number.

AI broke the link in a way earlier tools did not. A spreadsheet made accountants faster. It did not make the output of accounting feel free. What is different now is that the finished artifact, the thing you used to point at and say “I made that,” can be produced to a decent standard at almost no cost and almost no time. Commodity document translation rates have fallen 40 to 60 percent since 2020. A UK Society of Authors survey found more than a third of translators had lost work to generative models and 43 percent had seen their income drop. GitHub’s developer data has AI writing well past 40 percent of code in active workflows, and it writes exactly the part a junior engineer used to cut their teeth on. The craft did not get worse. It got abundant. And abundance is the one thing that identity built on scarcity cannot survive.

So you get the midnight message. Output through the roof, sense of necessity through the floor. People read that feeling as a signal that they are being replaced. Usually it is not that. It is the worth gap opening up, and it opens whether or not your job is actually at risk, because it was never really about the job.

I want to be precise about what changed, because the precision is the whole fix. It is not that AI does your work. It is that AI does the visible, produce-able part of your work, the part you could point at, and that part is exactly where you learned to store your sense of being good. Think about how you were praised, from your first job onward. Nobody handed you recognition for a decision you did not make or a result that showed up three quarters later. They praised the thing on the screen. The shipped feature. The clean draft. The deal on the board. You were trained, over years, to feel worth at the moment of production, and that training worked so well you stopped noticing it was training. Then the moment of production got automated, and the feeling had nowhere to land.

This is why the strongest builders feel it worst. If your whole career was mediocre output, losing the output costs you little. If you were genuinely excellent at the craft, you built a tall, load-bearing identity on top of it, and that is the identity the abundance knocks out from under you. The better you were at the doing, the larger your worth gap. I have seen brilliant engineers and writers get quieter and stranger over the last two years, and almost none of them would say out loud that the thing bothering them is that the work they were proudest of is now something a stranger can get for the price of a subscription.

What the worth gap is

Picture your sense of value as a bar. For your whole career, that bar was filled, top to bottom, with the thing you do: the code, the copy, the deck, the deal. Doing was worth, and worth was doing. One solid column.

Then the cost of the doing collapses. The bottom of the bar, the raw production, is now cheap enough that a model does it on demand. If your worth was the whole column and the bottom of the column just lost its price, you feel the drop as a loss of self, not a gain in productivity. The top of the bar, where judgment and choice and consequence live, is still enormously valuable. But most of us never anchored our worth up there. We anchored it to the doing, because the doing is what we could see.

The gap is not a feeling to wait out. It is an unclaimed space, and it gets claimed one of two ways. Either you move your anchor up into it on purpose, or the feeling of uselessness moves in and takes the lease. I have watched both happen, in other founders and in myself, and the version where you wait for the feeling to pass is the version where it does not.

There is a piece of psychology worth holding here, because it reframes the whole problem. Self-determination theory, the Deci and Ryan body of work that has held up for decades, names competence as one of three basic human needs, alongside autonomy and relatedness. And competence in that research is not defined as being the best, or being irreplaceable. It is defined as the lived experience of growth, of meeting a hard-enough challenge and producing an effect in the world. Read that twice if you build for a living. The need was never to be the only one who can do the thing. The need is to feel yourself getting better at a thing that matters. AI does not remove that need and it cannot satisfy it for you. What it does is move the altitude at which you get to feel it.

Hold that definition against the uselessness that 70 percent of those managers reported their people feeling. Useless is what competence failure feels like from the inside. But look closely and it is not a loss of competence at all. The person is often more productive than ever. What they have lost is the setting where they got to experience their competence, the hard-enough challenge at the moment of production. The machine made that particular challenge easy, so the feeling of effectiveness evaporated even though the effectiveness went up. That is the worth gap stated in the language of the research: your competence need is intact, but it is pointed at a challenge that is no longer hard, so it goes unmet while you wonder what is wrong with you.

The gap also explains two things that look contradictory in the 2026 data and are not. People are anxious about AI replacing them, and people are quietly refusing to use AI even when their employer is paying for it. If worth sits in the doing, both reactions make perfect sense at once. The tool threatens the doing, so you fear it, and the tool takes the doing, so you resist handing the doing over. Fear and resistance are the same reflex wearing two faces. Neither one closes the gap. They just keep you standing where it opened.

The two ways people get the worth gap wrong

When the floor drops, almost everyone reaches for one of two bad moves, and both come from the same mistake of leaving the anchor where it was.

The first is clinging. You keep doing the cheap part by hand to feel the old feeling. You write the boilerplate yourself because shipping it felt like being good at your job. You quietly route around the tool, not because the output is better, but because using it costs you the hit of competence you were trained to need. This is not rare and it is not only the stubborn. A 2026 GCheck survey of 1,500 US workers found 63 percent had lied about their AI skills and 81 percent were limiting AI at work even as their employers paid for it. Read charitably, a lot of that is not Luddism. It is people protecting the only place they know how to feel worth. Clinging feels like integrity from the inside. From the outside it looks like a senior person spending their scarcest hours on work a model finishes in seconds.

The second is drift. You hand everything to the tool, your output goes up, and your sense of authorship goes to zero. You start to feel like a front end for a machine, pasting and shipping work you could not fully defend if someone pushed on it. That hollow, fraudulent feeling is real, and it has a measured shape: 40 percent of workers who expect AI to hit their role say they have already watched it do part of their job. Drift is what the midnight message was. Not “the tool is bad,” but “I cannot find myself in the output anymore.”

Notice what the grid says that the usual advice misses. “Just embrace AI” pushes you to the right, from clinging to drift, and drift is not better. “Protect your craft” pushes you down, from re-anchored toward clinging. The only move that actually helps is vertical. You have to lift where your worth sits, and that is the part no tool and no pep talk does for you.

Where worth actually went

The value did not vanish. It moved up the same way it always moves when a layer gets commoditized: from the doing to the deciding. Three places absorbed it, and they are the three places to re-anchor.

The first is judgment, which is choosing what is worth doing and knowing when the work is actually good. When a model can generate ten plausible versions of anything in a minute, the scarce act is picking the right one and killing the other nine. That is taste and it is yours, not the tool’s. I have written separately about how to build taste when AI makes everything, because it is the muscle the current moment rewards most and starves fastest.

The second is outcomes, which is owning a result in the real world rather than a deliverable on a screen. A translated document is an artifact. A deal that closed because a human read the room is an outcome. Here is the most instructive receipt of the year. Klarna told the market its AI chatbot was doing the work of 700 support agents and cut headcount from around 5,000 toward 3,800. The efficiency looked clean until customer satisfaction fell, because the bot could not hold the emotionally charged disputes and the complex cases. The CEO reversed, saying plainly that from a brand standpoint there will always be a human, and the company rehired into a hybrid model. The machine took the easy 90 percent. The worth lived in the hard 10 percent, the part where a person is accountable for how it lands. Duolingo ran the same arc, announcing an AI-first stance and walking it back within a week. One survey now has 55 percent of companies regretting AI-driven cuts. The pattern is not that AI is weak. It is that worth concentrated exactly where accountability does.

The third is people, which is the part founders forget because it does not feel like output at all. The judgment you use to unblock someone, the standard you hold a team to, the person you grow, none of that shows up as a commit or a shipped feature, and all of it compounds in a way the doing never did. When the doing is cheap, the return on lifting another person is the highest it has ever been, and it is permanently uncopyable. A model can draft the feedback. It cannot be the person whose belief in someone changed what they attempted. I have gotten exactly one thing consistently right across four companies, and it was never a feature I wrote. It was a handful of people I bet on early who went on to do things that dwarfed anything I shipped with my own hands. None of that would survive the subtract-the-tool test as a loss, because the tool was never doing it.

	When doing was scarce	Now that doing is cheap
What you pointed at	the finished artifact	the decision behind it
What was rare	the skill to produce it	the taste to choose it and own the result
Who could do it	a trained few	anyone with a model, for the doing part
Where your worth sat	in the hands	in the judgment, the outcome, the people

The easy ninety percent going free feels like the whole thing going free

There is a specific illusion that makes the worth gap feel worse than it is, and naming it helps. When a model takes over a job, it takes the easy, high-volume part first, because that is what is legible and repeatable. In support, that is the routine ticket. In translation, the boilerplate document. In engineering, the boilerplate function. The hard, rare, accountable part, the escalation that is emotionally charged, the contract where a wrong word is a lawsuit, the architecture call that will haunt you for years, stays with the human. So the machine ends up doing something like 90 percent of the volume and almost none of the weight.

But here is the trap. Because the 90 percent was most of what you did all day, watching it disappear feels like watching your whole value disappear, even though the 10 percent that is left is where your worth actually lived the entire time. You spent years feeling productive because of the volume, and the volume was the cheap part. The Klarna reversal is the cleanest proof I have seen. The company could automate the 700 agents’ worth of routine volume, and did, and the numbers looked wonderful right up until the hard cases, the disputes where a person is upset and accountable judgment is the product, broke the experience and the brand. The worth was concentrated in the weight, not the volume, and you only find that out when you try to remove the humans and keep the weight.

This is the same shape I described in a business context as the capability overhang, the widening distance between what a tool can produce and what an organization can actually absorb and stand behind. At the level of a single person, the overhang shows up as this: the model can produce the artifact, but it cannot absorb the consequence, and it cannot be the one who is accountable when the artifact meets the real world. That non-absorbable weight is not a rounding error. It is the job now.

	Clinging	Drift	Re-anchored
The tell	you do cheap work by hand to feel useful	you ship work you could not defend if pushed	you own the call and the result, not the keystrokes
What it costs	your scarcest hours, spent on the cheap part	your sense of authorship and judgment	short-term comfort, because the new work is hard
What it feels like	diligence and standards	speed, then hollowness	growth, with a stretch of being bad at it
The move	hand the cheap part over on purpose	put your judgment back on top of the output	keep climbing before this rung gets cheap too

The founder trap is sharper than the worker trap

For founders this cuts deeper than it does for employees, because the thing most of us built our identity on is the thing we now most need to give away. I started companies because I could build the thing myself. “I am the person who can just make it” was not a skill for me, it was a self-image, the reason I believed I belonged in the room. AI took the market price of that self-image to near zero at the exact moment the company needed me to stop being the best individual contributor and start being something harder to feel good about.

So I caught myself doing it. Reaching for a task I could have handed off, because doing it refilled a tank the model had drained. It felt like diligence. It was actually me protecting an old source of worth at the company’s expense, which is the most expensive kind of clinging there is, because a founder’s hours are the one input that does not scale. The uncomfortable rule I had to accept is that the skill you are proudest of is usually the first one you should give to the machine, and the reason it is hard is the same reason it is correct. Your identification with it is exactly the thing blinding you to the fact that it is no longer where you add the most.

This connects to a mistake I see founders make with motivation, which I have written about as the ambition gap. When building gets cheap, both drive and self-worth lose their old anchor at the same time, and people confuse the two. The ambition gap is about why you still bother. The worth gap is about who you are while you do. They are cousins, and they show up together, usually around midnight.

There is a cold version of this that is worth saying because it keeps founders honest. In a market where buyers are cutting anything they can rebuild cheaply, which I have described as the cut list, the same logic that applies to products applies to your personal contribution inside your own company. If what you do can be produced by a model your own team already pays for, your attachment to doing it does not make it less cuttable, it just makes you the last person to notice. The founders who see this early move themselves up the stack before anyone has to have the conversation. The ones who do not spend a year defending a role the company has already quietly repriced. Re-anchoring is not only a wellbeing move. It is how you stay the person the company cannot do without, which is a different thing from the person who can do the most tasks.

Moving the anchor up a level

You do not close the worth gap by thinking about it. You close it by relocating where you go for the feeling of competence, deliberately, until the new location becomes automatic. The useful picture is a stack. The doing is the bottom rung. Judgment is the middle. Outcomes and the people you grow are the top. The anchor has to climb.

The practical test I use for any task is simple, and I call it the subtract-the-tool question. Imagine the model did the production part perfectly and instantly. What would still be valuable about my contribution? If the honest answer is nothing, that task was never where my worth lived, and clinging to it is nostalgia. If the answer is a real list, choosing the approach, catching what is wrong, owning whether it ships, carrying the consequence, then that list is your actual job now, and it is the thing to get visibly better at. The test sorts your work into the part to hand over without grief and the part to pour yourself into.

There is a trap on the healthy side of the grid worth naming, the rusting corner. If you re-anchor to judgment but stop touching the work entirely, your judgment slowly goes stale, because taste is built from reps and you quit doing reps. The answer is not to produce everything by hand again. It is to keep a deliberate sample of hands-on work, not for output, but to keep the instrument calibrated. I have written about this failure directly in the piece on fallback competence. Re-anchoring up is the move. Abandoning the ladder below you is a different mistake.

Clinging to the work you are proudest of is what keeps you small

Here is the part most people resist, and it is the whole point. The common advice is to find the thing AI cannot do and hide in it. That is backwards. The instinct to defend your proudest skill is not a strategy, it is the worth gap talking, and following it keeps you anchored exactly where you should not be.

The skill you are most identified with is, almost by definition, the one you have already maxed out. It is the comfortable one, the one that gives you the clean hit of competence with no risk to your self-image. Which means it is also the one where you have the least growth left and, now, the least scarcity. Giving it to the machine is not a loss of self. It is the only way to free the hours and the identity to go operate at a level where you are not yet good, which is the only place the feeling of growth actually lives. Competence, remember, is the experience of meeting a hard-enough challenge. You cannot get that from the challenge you solved years ago. You get it one rung up, where it is uncomfortable, where you might be bad for a while.

The founders I watch pull out of the worth gap cleanest are not the ones who found a corner AI cannot reach. They are the ones who let their old identity get commoditized on purpose, grieved it for about a week, and went to find the next hard thing. The ones who stay stuck are the ones still proving they can personally do the thing a model now does for free, and calling it standards.

A two-week practice to move your anchor

This is not a mindset you adopt once. It is a relocation you run until it sticks. Here is the version I have actually used.

For the first week, just watch. Every time you feel the drop, that small flinch of “I am not needed here,” write down the exact task that triggered it. Do not fix anything yet. You are mapping where your worth is still anchored, and the flinch is the map. By the end of the week you will see it clustered, almost always, on production work, the doing you used to be praised for.

In the second week, take that list and run the subtract-the-tool question on each item. Sort every task into two piles: hand it over, or this is where I add value now. Then do two things that will feel wrong and are right. Hand over three things from the first pile that you have been secretly doing by hand to feel useful, and actually let them go. And block real time, on the calendar, for one thing from the second pile that you are not yet good at, the judgment call or the owned outcome or the person you have been meaning to grow. Being bad at that for a while is the feeling of your anchor climbing. Do not confuse the discomfort with failure. It is the opposite.

One honest warning. The first week of this is worse before it is better, because clinging was doing a job, numbing the gap, and when you stop, the gap is right there. That is not a sign the practice is wrong. It is the gap finally getting your attention so you can fill it on purpose instead of letting uselessness fill it for you. If the whole exercise feels like reinvention, that is because it is the same work. I have written about the longer arc of that in the founder operating system, which is where this fits for me: worth is one of the things you have to run deliberately, not leave to default.

A few things make the relocation stick that are easy to skip. Tell someone. The worth gap thrives in private, because admitting that you feel less necessary even while shipping more feels like weakness, and so people carry it alone and it festers. Say it to one person who builds. You will almost always find they have the same midnight message saved in their head. Second, separate your two scoreboards on purpose. Keep one for what got produced, which the tool now inflates and which means less than it used to, and keep a second for the calls you made and the results you own, which is the one to actually watch. If you only track the first, you will keep chasing the feeling of worth in the place it left. Third, give it time measured in months, not days. You spent decades wiring worth to production. Rewiring it to judgment and outcomes and people is not a weekend realization, it is a practice you repeat until the new place is where your attention goes on its own.

And measure the right signal. The sign that your anchor has moved is not that you feel great about your output, because your output is now partly the machine’s and feeling great about it is a bit hollow. The sign is subtler and better: you stop needing to have personally made the thing in order to feel good about the thing. You can watch a model produce in an hour what used to take you a week, and instead of the small grief, you feel the specific satisfaction of having chosen well, caught the one flaw that mattered, and put your name on a result you stand behind. That is competence relocated. That is the gap closed, not papered over.

Questions builders ask about the worth gap

Is the worth gap just impostor syndrome with a new name?

No, and the difference matters for the fix. Impostor syndrome is feeling like a fraud when you are actually competent. The worth gap is the opposite situation: your competence at the doing is real, it just stopped being scarce, so the market and your self-image repriced it and you felt the drop. Impostor syndrome is a distortion to correct. The worth gap is an accurate signal that your anchor is in the wrong place. You fix one by trusting your competence. You fix the other by moving it.

If AI makes my skill cheap, should I stop learning it?

No, but learn it for a different reason. You are no longer learning the skill to be the one who produces the output, because the model produces it. You are learning enough of it to judge the output, catch when it is wrong, and own whether it ships. That is a real and different depth. The question of what to learn when AI knows everything is its own answer, but the short version is: learn the parts that let you decide and defend, not the parts that only let you produce.

I run a team. How do I protect my people from this?

Stop measuring and praising only output, because that is what anchored everyone to the doing in the first place. If the only thing that gets recognized is volume shipped, your team will cling to shipping volume by hand long after it stops being where they add value. Start naming judgment and ownership out loud. Praise the call someone made, the thing they chose not to build, the person they unblocked. You move a team’s anchor the same way you move your own, by changing what visibly counts.

Does this apply to junior people, or just senior ones?

It hits juniors differently and harder, because AI took the exact production work they used to learn on, which is a supply problem for the whole field that I have written about as the seniority debt. For an individual junior, the move is the same but earlier: do not spend years trying to become the fastest producer of a thing a model already produces. Build judgment deliberately, get close to real outcomes fast, and treat the doing as something to understand rather than something to be fastest at.

What if I actually like the doing and do not want to give it up?

Then keep some of it, on purpose, and be clear with yourself about why. There is nothing wrong with doing work by hand because it is satisfying or because it keeps your judgment sharp. The problem is only when the hand-doing is a hiding place, when it is quietly substituting for the harder work the company needs from you. Liking the craft is fine. Letting the craft be the reason you avoid growing is the trap. The test is whether you would still choose it if nobody praised you for it.

How is this different from just moving up into management?

It is not about a title. A manager can be fully anchored to the doing and miserable, and an individual contributor can be anchored to judgment and outcomes and feel great. Re-anchoring is about where you get your sense of value, not what your role is called. Some of the most re-anchored people I know never took a management track. They just stopped tying their worth to being the fastest set of hands and tied it to the quality of their calls and the results they own.

Is any of this real, or will the next model make judgment cheap too?

Some judgment will get cheaper, and the honest answer is that the line keeps moving up. That is not a reason to anchor low. It is the reason to anchor to the act of climbing itself rather than to any one rung. The durable thing is not a specific skill the machine cannot do yet. It is the habit of finding the next hard thing and being willing to be bad at it for a while. People who have that habit have survived every previous wave of commoditization. The rung changes. The climb does not.

What is the single first move if I only do one thing?

Pick the task you are most proud of being personally good at, the one that most feels like you, and hand it to the model this week. Not because the output will be better, but to feel, on purpose, that your worth did not leave with it. That one deliberate act does more to close the gap than any amount of reassurance, because it proves to the only person who needs convincing that you are not the doing. You are the one who decides what the doing is for.

I’m a four-time founder building AI-native companies. I write about the founder mindset, AI opportunities, and the mechanics of building with AI at vikasmalpani.com.