Programmers Before Plumbers. AI Knows Who to Fire.

37%.

That's how many companies plan to replace workers with AI by the end of 2026. Not "explore AI." Not "experiment with AI." Replace. Workers. With AI. One in three companies looked at their payroll, looked at ChatGPT, and chose ChatGPT.

Think that won't include you? Think again.

Here's what nobody's talking about: every job AI kills creates a gap. And gaps are where builders make money.

I've spent the last year building AI tools for the construction industry. Not gonna lie — what I've learned has nothing to do with construction and everything to do with where the economy is actually heading. I was wrong about which jobs would go first. Dead wrong. I assumed physical industries were vulnerable. The opposite happened. The pattern is so clear it's almost offensive that more people aren't seeing it.

Let me show you the pattern.

The Numbers Are Brutal. Read Them Anyway.

39% of companies already laid off workers due to AI — in 2025 alone. Not 2030. Not "sometime in the future." Last year. Done. Gone.

Here's what that actually means: if you work at a company with 100 people, statistically 39 of your peers at competing firms already lost their jobs to a language model. Their desks are empty. Their Slack accounts are deactivated.

58% expect even more layoffs in 2026. Dead serious — more than half of companies are planning another round. The first wave wasn't enough.

Anthropic — the company that builds Claude — published research that Fortune summarized in one devastating phrase: "a Great Recession for white-collar workers." Not blue-collar. Not manual labor. White-collar. The people who thought their degrees protected them. Wrong.

I'll be blunt: your MBA is not a moat. Your "10 years of experience" is not a moat. If your experience consists of doing the same pattern-based task for a decade, you just gave the machine a better training set. That's not a resume. That's a training corpus.

Here's where it gets interesting. These aren't random layoffs. There's a pattern in who gets cut, and that pattern is your roadmap.

Who's already being replaced? Customer support triage. Basic content production. Data entry and analysis. Scheduling. Recruitment screening. Internal reporting. Notice anything? Every single one of these is trained, standardized cognitive labor. Rules in, output out. The truth is simple: if your job has a manual, your job has an expiration date.

Who's safe? Electricians. Plumbers. Nurses. Therapists. Creative directors.

But actually — why does the Washington Post report that web designers and secretaries face more risk than janitors? Because a janitor operates in physical chaos. Every building is different. Every mess is different. Every broken toilet is a unique diagnostic puzzle. A web designer? They arrange components according to conventions. Conventions are patterns. Patterns are what AI eats for breakfast.

I noticed something while reading these reports that no analyst seems to highlight: the correlation between "job requires sitting at a desk" and "job is at risk" is almost 1:1. If your work lives entirely inside a screen, you're competing directly with software. Software doesn't negotiate salary. Software doesn't take sick days. Game over.

Why White-Collar Goes First (And Why That's Counterintuitive)

We spent 40 years telling kids: "Work with your brain, not your hands." The economy just flipped that script. Turns out, we had it backwards the entire time.

Here's the thing — we weren't just wrong. We were exactly backwards. The hierarchy we built — thinkers above makers, degrees above trades, keyboards above wrenches — AI just inverted the whole thing overnight. Gone.

The key insight is this: AI is essentially a pattern-matching engine running at inhuman scale. The more rules-based your job is, the more automatable it becomes. This is not speculation. This is math.

Programming? Pattern-matching over syntax and logic. Copywriting? Pattern-matching over tone and structure. Data analysis? Pattern-matching over numbers and trends. Customer support? Pattern-matching over complaint categories and resolution scripts. Not one of these requires hands.

Here's what that actually means for individuals: if you can describe your job in a flowchart, an AI can do your job from that flowchart. Wake up.

A plumber diagnosing a leak in a 50-year-old building? That's judgment under uncertainty. The pipe isn't where the blueprint says it is. The wall has been modified three times. There's water damage that could be from the roof or the bathroom two floors up. The plumber has to think in meat-space, with hands, eyes, and decades of intuition about how water behaves in old structures.

No language model on earth can do that. Not GPT-5. Not Claude. Not whatever comes next year. Honestly, I was wrong about this — I originally thought AI would commoditize physical trades by enabling less-skilled workers to perform at expert level. The opposite happened. AI made the diagnosis harder to fake, not easier. The trades got more valuable, not less.

Here's the paradox that should reshape your career thinking: the jobs that looked "safe" because they required education are now the most vulnerable, and the jobs that looked "replaceable" because they required physical labor are now the most protected.

Look — I'm not telling you to quit your desk job and become a plumber. I'm telling you to ask yourself, honestly: does your daily work look more like a plumber's or more like a prompt? Be brutal about the answer.

The Gap Map: Where Money Moves When Jobs Die

Every displaced function creates demand for something new. This isn't optimism. This is economics. Let me map it.

You'd think displaced workers create a labor surplus. Not really. They create a demand vacuum. When companies fire their support teams, they don't stop needing support — they just stop getting the human kind. And when customers notice, they pay a premium for the real thing. Which leads to an entirely new market tier.

Customer support gets automated. Basic tickets? Gone. Chatbots handle them. But what about the 15% of tickets that are genuinely complex? The customer who's angry, confused, and has a problem that doesn't fit any template? Companies still need humans for that. But now those humans need to be better — more empathetic, more creative, more authoritative. The floor rises. The pay rises with it. This means the worst customer support job just disappeared. The best one just got a 40% raise. The move is clear: specialize in the hard cases.

Basic content production gets automated. Your SEO blog post about "10 Tips for Better Sleep"? AI writes that in 40 seconds. Dead category. Crashed. But authentic human perspective? Contrarian takes? Writing that makes people feel something? That's premium now. Not despite AI — because of AI. When everyone can produce content, only distinctive content has value. The supply of generic exploded, which results in the demand for genuine skyrocketing. I'll be blunt: if you're a writer and you're worried, you should be. But not because AI writes better — because AI exposed how much "professional writing" was just competent pattern-filling all along. Here's what to do: stop writing what a prompt can write. Write what only you can write.

Data analysis gets automated. AI can crunch numbers faster than any analyst. But can it decide what the numbers mean for your specific business in your specific market with your specific constraints? Can it tell your CEO to ignore the data and trust the qualitative signal from the sales team? No. Never. Data interpretation and decision-making — the messy, contextual, political, human part — that's where the premium moves. The truth is this: the analyst who says "here are the numbers" is dead. The analyst who says "here's what we should do about the numbers, and here's why I'd bet my job on it" just became invaluable.

Recruitment screening gets automated. Resume filtering? Done. AI handles it. But who decides what "culture fit" actually means? Who reads between the lines of a candidate's career story? Who knows that the person with the "wrong" background might be exactly right for this team at this moment? Judgment. Ambiguity. Human. Always human.

See the pattern? AI eats the floor of every function. And the ceiling gets higher and more valuable.

Here's the tension nobody wants to admit: the gap between floor and ceiling is widening so fast that middle-skill workers have nowhere to stand. You're either premium or you're automated. The comfortable middle? Collapsed. There's no middle anymore.

Is your current work closer to the floor or the ceiling? Be honest. Not where you think your work is. Where it actually is. If you don't know, that's your answer.

For Builders: Where to Aim Your Next Project

Look — here's the opportunity most people are missing.

37% of companies plan to replace workers with AI. That means 63% are not. Sixty-three percent. Almost two-thirds of companies still need their humans. But those companies still need to compete with the 37% who are automating.

What do they need? AI-augmented workflows. Not AI replacements. Tools that make their existing workers 3x more productive without firing anyone. The move is obvious once you see it.

That's the SaaS goldmine of 2026-2028. And I'm not speculating — I built a company in one of these gaps.

Think about it. A company that wants to keep its team but stay competitive needs software that integrates AI into their people's workflow. Not software that replaces their people. The emotional and political dynamics inside companies heavily favor this approach. No manager wants to be the one who fired half the team. Every manager wants to be the one whose team suddenly produces 3x output. That's not a feature request. That's human nature.

The key decision for any builder right now: build for the augmentation market, not the replacement market. Dead serious — the replacement market is a race to the bottom. The augmentation market is a race to premium.

We're doing exactly this with RenoClear. It's an AI tool for the construction industry — helps contractors estimate costs, review quotes, measure rooms. It doesn't replace a single contractor. It makes contractors faster, more accurate, and more competitive. The contractors love it because it's their tool, not their replacement. Honestly, the first version was terrible. I assumed contractors wanted automation. They wanted superpowers. Big difference. I was wrong, and the users told me in the first week.

Where else does this model apply? Legal — AI-augmented contract review for solo practitioners. Healthcare — AI-assisted diagnosis summaries for doctors who are drowning in paperwork. Education — AI-powered grading assistance for teachers who have 150 students. Finance — AI-enhanced due diligence for small investment teams. Every single one of these is a seven-figure business waiting to be built.

Every industry has workers who are overwhelmed and under-tooled. Build for them. They'll pay because the alternative — getting automated out of existence — is worse. Here's what to do: pick one. Ship in 30 days. Iterate.

Here's what I tell founders who ask me where to start: never build for the industry you think is biggest. Always build for the industry where you've personally felt the pain. Secondhand market research is worthless next to firsthand frustration.

For Individuals: The Positioning Play

The truth is harsh: stop competing with AI on AI's turf. You will lose. Always.

AI's turf: speed, volume, consistency, pattern-matching, working at 3 AM without complaining, processing 10,000 data points in seconds, generating 50 variations of the same email.

You cannot beat AI at those things. Stop trying. Every minute you spend trying to be faster or more productive at pattern-matching tasks is a minute wasted. Here's what that actually means: "productivity" as we defined it for 50 years — output per hour — is no longer a human metric. Machines won that game. It's over. Find a different game.

Here's what AI cannot do:

Judgment in ambiguity. When the data says one thing but your gut says another, and you have to make a call with incomplete information, under time pressure, with real consequences — that's human territory. AI will give you a probability distribution. You have to decide. And look — deciding wrong and learning from it is still more valuable than never deciding at all. The machine can't do either.

Relationship building. A client doesn't trust your AI. They trust you. The handshake. The dinner. The moment when you said "I don't know, but I'll figure it out" and they believed you. That trust is unautomatable. I noticed something: companies that went all-in on AI customer interaction are quietly rehiring humans for their highest-value accounts. The automation savings weren't worth the relationship damage. Turns out, people buy from people.

Physical-world expertise. Anything that requires navigating the chaos of the real, physical, messy world. Construction. Healthcare. Emergency services. Skilled trades. The world is not a dataset. It's a place where pipes leak, patients cry, and nothing works exactly like the manual says. No model can hold a wrench.

Creative direction. AI can generate 100 images. It cannot tell you which one is right. It cannot feel the cultural moment. It cannot say "this is boring, try something that scares you." Creative direction is taste plus courage plus context. AI has none of those. Not even close.

Cross-domain synthesis. AI is trained on patterns within domains. The most valuable insights come from connecting things that don't obviously connect. A biologist who understands economics. A plumber who understands software. A musician who understands data visualization. These hybrids are impossible to automate because the machine doesn't even know the connection exists until a human makes it. This means the most valuable person in any room is the one who doesn't fit the room.

The safest career move in 2026 is not "learn to code."

It's "learn to do what code can't."

Here's the thing — that sounds abstract. So let me make it concrete. This week, find one decision you made at work that required weighing factors no spreadsheet captures. Politics. Gut feel. Cultural context. Relationship history. That's your moat. Do more of that. Get better at that. Make yourself known for that. Share what you learn.

The Real Shift: Tasks, Not Jobs

Here's the nuance that most headlines miss.

AI doesn't replace jobs. It replaces tasks. A marketing manager's job has maybe 40 distinct tasks. AI might automate 15 of them. The job doesn't disappear — it transforms. The marketing manager now spends zero time on reporting and scheduling, and all their time on strategy and relationships.

Sounds great, right? There's a catch. A brutal one.

If your job was 15 tasks, and AI handles 10 of them, your company doesn't need five of you anymore. They need two. Three are gone. The three who leave are the ones who were only good at the tasks AI now handles. The truth here is uncomfortable: the question isn't "can AI do my job?" It's "can AI do enough of my job that my company needs fewer of me?"

But the bigger, darker catch: entry-level positions are vanishing. Those 10 automated tasks? They used to be how juniors learned the job. The report-writing, the data-pulling, the scheduling — that was training. When AI handles the training tasks, how do new people enter the field? Nobody has an answer yet. That should terrify you.

Not gonna lie — this is the part that keeps me up at night. We're not just disrupting jobs. We're disrupting the ladder. The first three rungs are gone. And nobody's building new ones yet. The entire pipeline is broken.

This is the real crisis nobody's solving yet. And it's an enormous opportunity for builders. Whoever figures out the "new apprenticeship" model — how to train humans in an AI-augmented workplace — builds a billion-dollar company. Here's what to do if that's you: start with one trade, one profession, one vertical. Prove it works small.

Bootcamps taught people to code. What teaches people to think alongside AI? What teaches judgment, not just execution?

If you can answer that question with a product, you're sitting on the next big thing. And honestly? I think the answer won't come from Silicon Valley. It'll come from someone who's actually trained apprentices in the real world — a master electrician, a senior nurse, a construction foreman — someone who knows what "learning by doing" actually means when the "doing" gets automated.

The Two-Year Window

Here's my honest read on timing.

We're in a two-year window — 2026 to 2028 — where the displacement is happening but the new equilibrium hasn't formed yet. This is the chaos period. And chaos is where outsized returns live. Always has been.

The big companies are paralyzed by internal politics. They can't move fast. They're debating AI ethics policies and forming committees and hiring "Chief AI Officers" who don't ship anything. Dead weight.

The small builders? The indie hackers? The people who can ship a tool in a weekend and iterate based on real user feedback? This is your moment. The gap between "AI can do this" and "someone built a product that actually does this well" is enormous right now. That gap is pure margin.

In two years, that gap closes. The big players figure it out. The market consolidates. The easy wins disappear. Window shut.

Look — I've watched this exact pattern play out in construction tech. The window where a solo builder could launch and win lasted about 18 months. Then the incumbents woke up. The indie advantage is real, but it's temporary. Move now or compete with Salesforce later. Your call.

The time to build is now. Not "soon." Now. Not next quarter. Today.

Make Sure Your Work Doesn't Look Like a Prompt

AI doesn't know who to fire. It's a tool. It has no opinions, no agenda, no malice.

Companies know who to fire. And they fire the people whose work looks like something a prompt can produce.

Predictable output. Standardized format. No judgment calls. No relationship value. No physical-world engagement. No creative risk. If that describes your daily work, you're not safe. Your degree doesn't save you. Your experience doesn't save you. Your seniority doesn't save you. Nothing saves you except becoming irreplaceable.

Here's the thing — I'm not trying to scare you. I'm trying to move you. Fear without action is just anxiety. Fear with a plan is strategy.

But here's the reversal that matters: if you're reading this, you're already ahead. Most people are in denial. They're hoping their company won't automate. They're trusting their managers to protect them. They're waiting. Waiting is the most dangerous thing you can do right now.

You're not waiting. You're reading about the disruption and thinking about how to position yourself. That puts you in a different category entirely.

So here's your action plan. Three things. Do them this week. Dead serious.

One. Audit your daily tasks. List every task you do in a week. Mark which ones AI could do at 80% quality. The marked tasks are your vulnerability surface. Start shifting your time toward the unmarked ones. This means cutting what's comfortable and leaning into what's hard.

Two. Pick a gap from the gap map above. Customer escalation, authentic content, data interpretation, creative direction — any of them. Start building expertise there. Not "thinking about it." Building. Ship something. Write something. Make a decision that requires judgment and publish the reasoning. That's how you become visible.

Three. If you're a builder, identify one industry where workers are overwhelmed and under-tooled. Talk to five people in that industry this week. Find their pain. Build for it. The move is simple: solve one real problem for one real person, then scale.

The disruption is not coming. It's here. 37% of companies. This year. The question isn't whether AI will reshape the labor market. It already has.

The question is whether you're the one getting displaced or the one filling the gaps.

If your work looks like something a prompt can produce — bookmark this. You'll need it sooner than you think. Share it with one person whose job is on the line. Drop a comment with the one task you're going to stop doing this week. That's not engagement bait. That's accountability.

The people who act on this will look back in two years and say it was obvious. The people who don't will say nobody warned them. You've been warned.

---

Counterintuitive Engineering builds AI tools for industries that AI can't replace. Follow @CounterIntEng for more.