The layoffs happening across enterprise technology right now are not cyclical. White-collar headcount has been thinning for two years, and in the last few quarters companies have stopped calling the cuts restructuring. They are saying AI out loud, and the roles they are naming go well past tech: customer service, junior analysis, mid-level coding, sales operations, compliance review, and a widening slice of what used to be stable career tracks in law, finance, and middle management. The market is repricing what reliable task execution is worth, and the price is falling fast.
A job was always a bundle of tasks sold at a premium. The premium reflected the cost of finding, training, and coordinating a reliable human to execute those tasks. AI has not destroyed work. It has destroyed the premium on generic task execution. Once you see it that way, the layoff cycle stops looking like bad news and starts looking like a preview of how the next decade is going to allocate value.
This piece is an argument about where that value is going and what to do about it, whether you are the person being repriced, the leader watching your org chart melt, or the capital trying to figure out where the next decade’s returns actually live.
The Third Compression
This is the third time in two hundred years that execution has been repriced, and by a wide margin the most aggressive of the three.
The modern school system was adapted by 19th-century industrialists, who first began turning craft into task. The requirement was simple: Factories needed workers who could arrive on time, sit in rows, follow written instructions, move through standardized procedures, and produce reliable output at a predictable cost. The school that prepared them used the same architecture: bells, grades, age-cohort progression, assessment by compliance to a rubric. The purpose was not to produce thinkers or problem-identifiers. The purpose was to produce reliable task-doers and, at the time, that was the right purpose. Industrial production rewarded the output enormously, and the standard of living that came out the other side is part of why we are still running roughly the same curriculum two centuries later.
The second compression came with offshoring. From the 1980s through the 2000s, manufacturing moved first, then back-office white-collar work: IT services, accounting, customer support, legal research. The pattern was consistent. If a job could be clearly documented, it could be done somewhere cheaper. The roles that stayed onshore were the ones that involved judgment, relationship, and original problem definition. The roles that left were the ones that involved execution of someone else’s thinking.
Offshoring should have been the warning. It wasn’t, for a reason worth naming. The jobs that moved were the ones that looked commodity from the start: assembly work, data entry, call centers, low-level coding. The jobs that stayed onshore were harder to export for a specific reason. They ran on accumulated experience applied to repeatable problems: the senior lawyer who had drafted a thousand contracts, the senior analyst who had built a thousand models, the senior consultant who had produced a thousand decks. The apparent moat was twenty years of pattern recognition, and the curriculum doubled down on routing people into exactly those careers. Offshoring appeared to confirm the thesis: commodity execution goes overseas, experienced execution stays home.
AI is that same compression run a third time, without the geographic arbitrage, and it arrives pre-loaded with the pattern recognition that used to require twenty years. Where offshoring could push the cost of a given task down by an order of magnitude, AI is pushing it toward zero. The senior knowledge worker whose value was accumulated experience applied to standard problems, the exact product the system was optimized to produce, is the role now being repriced.
What Died: The Task-Doer, Not the Worker
Two things keep getting conflated in the headlines. “A job” is a transaction. A firm pays a person to reliably complete a bundle of tasks at a known cost. “Work” is the underlying creation of value for a customer. AI came for the margin, not the work.
The task-doer, the role whose economic value came from executing known procedures at a predictable cost, is the role that is vanishing. The roles that survive are the ones that own something. A customer relationship. A piece of judgment. A brand. An audience. Equity. A distribution channel. A craft. None of these can be bundled and sold at a predictable cost, which is exactly why the machine cannot eat them.
If you do not own something, you are being priced. That is true whether you sit in an enterprise sales seat, a junior analyst seat, or a customer success pod. The moment a reliable machine can do your bundle at a tenth of the cost, the price of your bundle collapses toward the machine’s. This is not a forecast, it’s competitive necessity.
The instinct most people have in this moment is to protect the old ladder: upskill, learn prompting, become “AI-augmented,” keep climbing. That is fine advice, and it will buy some people another decade. It does not answer the structural question. The ladder itself is being priced down. Building the ladder is quickly becoming a better bet than climbing it.
Going to Market Has Never Been Cheaper
The story getting all the attention is what is being destroyed. The under-told story is what is being built on the other side.
The same forces compressing white-collar employment are compressing the cost of starting a company. Four inputs have moved at the same time.
Software build. A B2B MVP that cost $500,000 to $2 million to ship in 2020 can be built today, by a small team willing to lean on modern tooling, for a fraction of that. Ship cycles that used to run in quarters now run in weeks. Most of what used to be a funded MVP engineering team is now a motivated operator and a weekend.
Admin and back office. Bookkeeping, legal templates, compliance reviews, contract redlines, CRM setup, tax filings. The entire middle layer that used to require an accountant, a paralegal, and a part-time operator is now a handful of monthly subscriptions and a model that does the reading for you.
Customer service and operations. The same function enterprise software companies are now eliminating thousands of seats to replace is available to a solo operator for less than the cost of a cell phone plan.
Marketing and distribution. Content production, creative iteration, SEO research, customer segmentation, lead qualification. All of it used to live inside agencies charging $10,000 to $30,000 a month. Most of it now runs from a kitchen table.
Pick any category. The input-cost curve has bent. This is not a vibe. It is a line on a chart, and the line is going one direction.
Niche Markets Are Suddenly Viable
Here is the math I want you to leave with.
A $10 million total addressable market, captured at 10 percent share, is $1 million in annual revenue. At 60 percent gross margins, that is $600,000 in gross profit. For a single operator or a small team, that is a life-changing business. For a venture fund with $500 million to deploy, it is a rounding error. For the corporate incumbent with a billion-dollar cost base, it is uneconomical to even staff the meeting to discuss it.
Five years ago, that $10 million TAM was untouchable for everyone. Too small for venture. Too unprofitable for incumbents and too expensive to serve with traditional cost structures. The $10 million niche has been an orphan.
The AI cost curve just adopted it.
A *far* from exhaustive list of what that unlocks:
Vertical compliance software for specific regulated trades. Mobile-notary scheduling. Independent insurance-adjuster workflows. Franchise-operator reporting. Each one too small for a big company, too specialized for generic tools, now buildable by two people in a quarter.
Micro-manufacturing of discontinued parts. Replacement components for vintage agricultural, marine, or industrial equipment. AI-assisted CAD, domestic CNC, and Shopify-grade distribution make production runs of 50 to 200 units profitable for the first time in a generation. The customer has been waiting for someone to show up.
Hyper-specific professional services. R&D tax credit work for indie game studios. IP licensing support for YouTube creators with a real catalog. Bookkeeping for Shopify sellers above $2 million GMV. Each of these was too low-volume to staff the traditional way. Each is workable with a small human team and heavy AI leverage.
Occupational sub-niche education. Exam prep for one specific professional license. Training products for a single trade. Certification prep for a specific software platform used by 40,000 people in the country. Production and distribution costs are approaching zero.
Every one of these is a previously unsolved problem getting solved. Not because someone had a new idea. Because the math finally works.
Two Doors, Not One
There are now two doors open for anyone paying attention.
Door A: Build your own. A displaced professional with a thesis and $25,000 can own a real slice of a niche market inside of a year. The ladder climb is being priced down. The build path is open. If you are in the first group of people to walk through that door, you are early, not late.
Door B: Build inside. The same cost curve making solo entrepreneurship viable is providing your best people the ability to build without your infrastructure. If your senior operator can see a $10 million niche and ship a product for $50,000, the only reason they stay is because you have given them a better deal, or partnership, than the one they can now write for themselves. Most companies have not.
This shifts the dynamic for your most self-dependent employees. The old trade was infrastructure dependence, rationalized as loyalty. The new trade is a choice between real partnership and real independence, and you are competing with the latter whether you want to be or not.
The companies that win the next decade will look less like hierarchies and more like platforms for builders. They will make it easier to spin up a new product line, run a small P&L, or launch a sub-brand inside the firm than it is to leave and do it alone.
This shift is bigger than any one company. I have argued elsewhere (see HumanScale) that a more interesting economy is one with more owner-operators, fewer unicorn lottery tickets, and capital that gives a damn about what it builds. The AI cost curve is what makes that argument operational rather than romantic.
Venture math does not fit $1M to $10M ARR durable businesses. However, in this new environment where traditional economies of scale are compressed, those businesses become some of the most durable assets an investor can own. Their moat is speed, specificity, and proximity to the customer, all of which widen as AI commodifies the capabilities once reserved for scale. Direct investment, family-office equity, and community capital are all structured to underwrite that profile. The capital category is opening up at exactly the same moment the builder category is.
A Framework: The Four Levers of Internal Entrepreneurship
Whether you are building solo, leading a company, or allocating capital, this is the test. If all four levers are present, the builder will stay and compound inside the firm. If any one is missing, the builder will leave, and sooner than you think.
Mandate. Most people have been trained to execute assignments, not to identify problems. The first job of a good Mandate is to retrain the muscle, provide your team with real problems, owned end to end, with a real P&L or product attached. Not a project, not a committee seat, not a guidebook. Ownership.
Margin. Real economic upside tied to the outcome. Equity, phantom equity, profit share, or a meaningful revenue split. Salary plus bonus is not margin.
Machinery. AI tools, capital, and distribution access that make a small TAM reachable from inside the firm. If your internal builder has less leverage than a solo founder with a credit card, you have lost before you started.
Measurement. An outcome scoreboard everyone agrees on in advance. Revenue, retention, contribution margin. Not activity. Not hours. Not optics.
If you are a builder, ask for the four. If you cannot get them, leave and build your own thing. The cost to do it has never been lower and will not be this high again.
If you are a leader, design the four into your organization. The companies that do this in 2026 and 2027 will look, by 2030, like they compounded talent while everyone else bled it.
If you are capital, the next decade of returns lives in funding builders at both scopes. Inside companies and outside of them. The LP who understands that the $10 Million Niche is newly investable, and that the best operators to back may sit inside existing companies waiting for the four levers, is going to look prescient in five years and obvious in ten.
The New Curriculum
Execution is no longer the goal, and the margin will flow to building. Which means the skills you need to be building now are different from the ones the system trained you for.
For two centuries the skills that compounded were the ones factories and firms needed: showing up on time, following procedures, executing a defined task reliably. Schools taught them and careers rewarded them. And the twenty years of seniority and education just became the machine’s job to do.
What AI cannot do, and what most of us have never been trained to do well, is the other half of work. Identifying problems nobody has defined yet. Seeing markets the incumbents are missing. Building things from nothing. Owning outcomes rather than tasks. Making judgment calls in situations where the rubric does not apply. Developing taste and building the audiences and relationships that compound on their own.
These are the muscles worth training now, and they are not trained by taking another course. They are trained by owning something small that is yours, shipping it, watching what happens, and adjusting. The first product you build will probably not work. Neither will the second. By the fourth, the muscles are different. By the tenth, you are someone who builds rather than someone who executes, and that person is priced very differently.
None of this shows up on a standardized test. That is a problem for the school system to figure out, and it will. For the young professional with decades of career runway ahead, the shift has to start now, in whatever form it can take. A side project, a small business, or a niche product - as long it’s a thing you own and are responsible for. Start small, but start. The cost of starting has never been lower, and the cost of waiting has never been higher.
Close
The market is done paying a premium for execution. It is starting to pay a premium for building. The job is dead. The builder is not. Whether they build for you, with you, or against you is now an organizational design choice.
If you are a founder or operator trying to install the four levers inside your company before your best people go build on their own, I would like to talk. If you are a capital allocator trying to figure out where the next decade of returns actually lives, happy to discuss. And if you are a builder picking between Door A and Door B, that is the conversation I most want to have. duncan@saorsapartners.com
Subscribe to Conduit of Value for the ongoing thread. This piece is a companion to HumanScale, which made the case for right-sized businesses as a capital thesis. Together they argue for building at a human scale as the default path, not the consolation prize.
One question to leave you with, and I read every reply:
What are you building right now that was not buildable two years ago?