Will AI Replace Your Job? What the Data Says

On February 26, 2026, Jack Dorsey sent a letter to Block shareholders that read more like an obituary for traditional headcount. The company behind Square and Cash App was cutting 4,000 employees, nearly 40% of its workforce, and replacing their functions with what Dorsey called “intelligence tools.” The stock jumped 24% in a single after-hours trading session. Wall Street didn’t mourn the lost jobs. It celebrated the savings.

Block had just posted $2.87 billion in quarterly gross profit, up 24% year over year. This wasn’t a struggling company making desperate cuts. It was a profitable one betting that AI could do the work of thousands. Dorsey predicted most companies would follow within the year.

AI job displacement is the replacement of human work tasks by artificial intelligence systems, and new research shows it is hitting white-collar, desk-based roles far faster than anyone expected while physical, in-person jobs remain largely untouched.

Two weeks after Block’s announcement, Dell confirmed it had eliminated 11,000 positions in fiscal 2026, spending $569 million on severance to pivot toward AI server infrastructure. Meta began floating plans to cut 20% of its 79,000-person workforce to offset $135 billion in AI spending. The pattern was impossible to ignore: the companies investing the most in AI were also shedding the most people. So the question millions of workers are now asking is direct: will AI replace my job? The honest answer is more complicated than the headlines suggest.

Last updated: March 2026

What jobs will AI replace in 2026?

The most comprehensive answer arrived on March 5, 2026, when Anthropic published a study called “Labor market impacts of AI” that introduced a metric called “observed exposure.” Unlike previous research that measured what AI could theoretically do, this study tracked what AI is actually doing in workplaces right now.

The findings were blunt. Computer programmers topped the list at 75% task coverage. Customer service representatives, data entry keyers, and medical records specialists followed close behind. Marketing research analysts came in at 39% observed exposure.

But the study’s most important finding wasn’t about which jobs sit in the crosshairs. It was the gap between theoretical capability and actual adoption. For computer and math workers, AI tools could theoretically handle 94% of their tasks. In practice, they’re handling 33%. That 61-point gap means the technology is advancing faster than companies can implement it.

On March 17, Axios partnered with The Action Network to build a tool that translates this data into something visceral: betting-style odds on your job being disrupted. Computer programmers sit at 45%, customer service reps at 42%, data entry workers at 40%. The Washington Post launched a similar interactive tool the same week, letting users look up their specific occupation’s vulnerability score.

The pattern across all three analyses is consistent. Desk-based, digital, repetitive work faces the highest risk. If your job involves processing information that follows predictable patterns, whether that’s writing code from a spec, answering common support tickets, or pulling numbers into a spreadsheet, the timeline for AI doing it faster and cheaper just got shorter.

Which startup roles are most vulnerable?

Customer support is the first role most founders will rethink. Klarna reported that its AI assistant now handles the equivalent work of 700 full-time agents, resolving issues in under two minutes compared to the 11 minutes a human needed. Block’s 4,000-person cut hit support teams hard. For early-stage startups spending $15,000 to $30,000 monthly on a support team, the math has changed overnight.

Junior developers face real pressure. GitHub data shows developers using Copilot complete tasks 55% faster. When one developer with AI tools matches the output of two without them, hiring plans shrink. The Anthropic study confirmed that software engineering tasks carry among the highest observed AI exposure of any profession.

Marketing research and content generation roles are shifting fast. AI can now produce keyword research, competitive analysis, first drafts, and performance reports that took junior marketers days to compile. The 39% observed exposure for marketing research analysts in the Anthropic data reflects what many startup founders already know: one marketer with AI tools does the work that previously required three.

Data entry and administrative support is the clearest case for automation. At 40% observed exposure and climbing, these roles are dissolving into AI workflows across companies of every size. Accounting tasks are following the same path. AI now handles invoice processing, expense categorization, and bank reconciliation faster and more accurately than humans. Pilot, one of the largest startup bookkeeping services, has woven AI into nearly every step of its workflow.

What jobs are safe from AI?

Physical, in-person work has the lowest observed exposure to AI across every dataset published in 2026. Construction workers, electricians, plumbers, nurses, surgeons, and emergency responders exist in a different risk category from anyone sitting at a desk.

But safety from AI isn’t limited to trades. Several white-collar roles are proving resistant for reasons that have nothing to do with physical labor.

Therapists and counselors sit near the bottom of the risk chart. You can build a chatbot that asks “how does that make you feel?” but clinical psychology requires reading what a patient isn’t saying, recognizing patterns across months of sessions, and making judgment calls about medication and treatment that carry real consequences.

Creative directors and strategists make decisions about brand, risk, and taste that require cultural context no model possesses. Individual writing or design tasks can be automated. Deciding what to create, for whom, and why still needs a human who understands what the audience wants before the audience knows it themselves.

Complex B2B salespeople operate through relationships that take months to build. AI can source leads and draft outreach emails. It cannot sit across a table from a skeptical CFO, read the room, and earn enough trust to close a six-figure deal. AI voice agents can handle thousands of routine calls, but the calls that matter most still need a person.

The thread connecting every AI-safe role is accountability. When a building fails inspection, a patient gets the wrong diagnosis, or a business strategy craters, someone needs to own the outcome. AI can advise. It cannot be held responsible. That distinction is the moat.

How do I know if AI will replace my job?

Skip the generic listicles. Three questions will give you a more honest assessment than any online tool.

First: how much of your daily work involves processing structured information that follows predictable rules? The higher your answer, the higher your exposure. If your day consists of emails, data pulls, report formatting, and scheduling, AI tools already do each of those things faster than you can.

Second: does your job require being physically present in an environment that changes unpredictably? Construction sites shift daily. Patients present new symptoms every visit. Restaurant kitchens run on real-time adaptation. If your work demands constant physical improvisation, AI is not coming for it anytime soon.

Third: would your employer trust an AI system to make the decisions you make, given the consequences of getting them wrong? If the answer is no, because the stakes are too high, the judgment too context-dependent, or the liability too serious, your role has a moat that technology alone won’t cross.

The Axios tool lets you look up specific occupations and see their disruption odds. But odds are probabilities, not death sentences. A computer programmer at 45% risk doesn’t mean half of all programmers lose their jobs tomorrow. It means their daily task profile overlaps significantly with current AI capabilities, and that overlap will keep growing.

How founders are responding to AI job replacement

The founders paying closest attention aren’t panicking about AI replacing jobs. They’re restructuring around it.

The playbook emerging across startups in early 2026 goes like this: hire fewer people, pay them more, and arm them with AI tools that multiply their output by 3x to 10x. Block’s 40% cut is the dramatic corporate version. The startup version is building a five-person company that performs like a fifty-person one.

Sam Altman’s prediction about the “one-person billion-dollar company” still sounds extreme, but the direction is undeniable. Founders are building AI agent departments that handle customer support, data analysis, content creation, and basic operations while the humans focus on strategy, sales, and product decisions.

Others are building micro apps with AI coding tools that replace $50,000-a-year SaaS subscriptions, eliminating both the software cost and the person who used to manage it.

The response isn’t just about cutting costs. It’s about concentrating human attention on the work that actually moves the needle: relationships that close deals, strategic bets that open new markets, creative ideas that differentiate a brand. Everything else is increasingly a candidate for automation.

For anyone worried about their own position, the best move isn’t to fight the trend. It’s to become the person who knows how to use AI tools better than anyone else on the team. The Anthropic study’s most reassuring finding was that gap between what AI can do and what companies have actually implemented. Whoever figures out how to close that gap inside an organization becomes the most valuable person in the room, not the most replaceable.

And if AI does take your job? There’s a growing argument that the best response is entrepreneurship. Turning an AI layoff into a business within 12 months is becoming a real playbook, not just motivational advice.