Something unusual happened in early 2025. Across the United States, workers in offices and creative studios began receiving unexpected news. Not budget cuts. Not restructuring. Something else entirely.

Artificial intelligence had arrived, and it brought pink slips with it. Consulting firm Challenger, Gray & Christmas tracked at least 55,000 layoffs tied directly to AI rollouts in 2025 alone. Nationwide, 1.17 million jobs vanished in the largest workforce reduction since the COVID pandemic. Writers, programmers, and analysts found themselves competing against software that worked faster, never slept, and cost a fraction of their salaries.

But amid all the panic about machines replacing humans, a quieter story went mostly unnoticed. While some professions faced extinction-level threats, others remained almost untouched. Microsoft, one of the largest AI developers in the world, conducted research that revealed something surprising about which jobs would survive the coming wave. Some answers were predictable. Others were not.

How Microsoft Mapped AI’s Reach

Before jumping to conclusions about winners and losers, it helps to understand how Microsoft arrived at its findings. Researchers analyzed 200,000 anonymized conversations between users and Microsoft Copilot, the company’s AI assistant formerly known as Bing Copilot. Every query, every request, and every task became a data point.

Next, they mapped these conversations to the O*NET database, a classification system maintained by the Department of Labor that breaks down occupations into specific tasks and activities. By comparing what users asked AI to do against what different jobs require, researchers could estimate how applicable AI might be to each profession.

Workers who spend their days drafting reports, researching topics, or explaining concepts face the highest exposure. Entry-level employees with advanced degrees appear especially vulnerable, as their work often involves exactly the kind of information processing AI handles well.

Marketing professionals and journalists might want to pay attention. So might anyone whose job involves sitting at a computer and typing.

Knowledge Workers in the Crosshairs

Microsoft’s data painted a troubling picture for certain professions. Writers, interpreters, programmers, and historians ranked among the most exposed to AI disruption. Each of these roles involves tasks that AI chatbots perform with increasing skill.

Need a first draft of an article? AI can produce one in seconds. Require translation between languages? AI handles dozens of language pairs without hesitation. Want to debug code or explain a historical event? AI trained on billions of documents can often deliver passable results.

Entry-level workers face particular risk because their roles typically involve more routine tasks. A senior programmer might spend time mentoring juniors, attending meetings, and making judgment calls that require experience. A junior programmer might spend more time writing code that AI could generate instead.

None of this means these jobs will disappear tomorrow. But the pressure is real, and it shows no signs of easing.

Why Some Jobs Resist Automation

For all its power, AI runs into walls. Some tasks resist automation, and understanding why reveals something important about work itself.

Microsoft’s research found that physical tasks showed low AI applicability. “On the flip side, it’s not surprising that physical tasks like performing surgeries or moving objects had less direct AI chatbot applicability.”

An AI chatbot can tell you exactly how to install drywall. It can explain the chemistry behind adhesives and recommend the best tools for the job. What it cannot do is pick up a trowel and spread joint compound. Physical presence remains beyond its reach.

However, the barriers extend beyond mere physicality. Empathy and human connection prove difficult for machines to replicate. When a patient sits nervously in a clinic chair, waiting for a needle, they need more than technical precision. They need someone who notices their anxiety, speaks gently, and adjusts their approach based on subtle cues no sensor can detect.

Trust and accountability also matter. When something goes wrong with hazardous waste removal, someone must answer for it. Liability cannot be assigned to an algorithm. Real humans must take responsibility, and that requirement alone shields certain professions from automation.

Five Careers AI Cannot Touch

Microsoft’s research identified several occupations with almost no AI applicability. Five stood out for what they reveal about the nature of work itself.

Phlebotomists ranked as the least affected profession in the entire study. Drawing blood requires steady hands, calm nerves, and the ability to reassure patients who might flinch or panic. Would you trust a robot to find your vein and try again if it missed? Even AI enthusiasts might hesitate before answering yes. Robotics firms have experimented with ultrasound-guided blood collection machines, but adoption remains low. Patients still prefer a human touch.

Nursing Assistants occupy a similar position. Hands-on medical care demands more than following protocols. Patients need kindness, rapport, and emotional sensitivity. A nursing assistant might notice when an elderly patient seems confused or when a new mother needs encouragement. These observations arise from human connection, not data analysis.

Hazardous Materials Removal Workers deal with dangerous substances that require physical intervention. AI can advise on proper disposal procedures for industrial solvents or radioactive materials. It can recite safety regulations from memory. But it cannot put on protective gear and remove the threat. More importantly, it cannot be held accountable if procedures fail.

Skilled Manual Laborers such as painters, plasterers, and their helpers, enjoy similar protection. Every job site presents different challenges. Walls have imperfections. Surfaces require judgment calls. Fine motor skills and adaptability keep these roles secure in ways that office work cannot match.

Embalmers might seem like a surprising entry on this list, but the logic holds. Funeral home employees work with grieving families during their darkest moments. Each body presents different challenges that require years of training and fine motor dexterity. Beyond technical skill, embalmers must earn trust from families who are handing over someone they loved. No algorithm can shoulder that responsibility.

Forty Professions That Remain Secure

Beyond these five, Microsoft’s research identified dozens of occupations with low AI applicability. Oral and maxillofacial surgeons, automotive glass installers, ship engineers, and tire repairers all ranked among the safest. So did prosthodontists, highway maintenance workers, cement masons, massage therapists, and surgical assistants.

Common threads connect these roles. Most involve physical labor. Many require specialized equipment or training. Several demand real-time judgment in unpredictable conditions. And almost all require some form of human presence that cannot be delegated to software.

Dishwashers, maids, and housekeeping cleaners also appeared on the list. So did roofers, logging equipment operators, and dredge operators. Machines might eventually take over some of these tasks, but for now, they remain firmly in human hands.

What Microsoft’s Study Actually Says

Before drawing sweeping conclusions, readers should understand what Microsoft’s research actually measured. Researchers calculated AI “applicability,” not job displacement. Their study showed where AI might prove useful, not where jobs would disappear.

Microsoft explicitly warned against misreading the findings. “A job is far more than the collection of tasks that make it up.”

O*NET captures tasks, but it misses judgment, ethics, and interpersonal skills. A report-writing task might appear simple on paper, but the real work involves knowing what to include, how to frame bad news, and when to push back on requests. These subtleties escape algorithmic analysis.

Data came exclusively from Copilot users, which may not reflect how people use other AI systems. Someone might ask ChatGPT for creative writing, but turn to Copilot for research. Different tools serve different purposes, and no single dataset captures the full picture.

Researchers also noted that determining whether conversations occurred in work or leisure settings proved nearly impossible. Someone asking about medieval history might be a professional historian or a curious hobbyist. Context matters, and AI studies often struggle to capture it.

What Machines Cannot Replace

Amid all the data and predictions, one question lingers. What does it mean that machines now write, research, and advise as well as many trained professionals? And what does it say about us that the safest jobs require a human hand, a caring voice, or a steady presence in moments of fear and grief?

Microsoft’s findings point to something deeper than job security. Certain forms of work resist automation because they demand qualities no algorithm can simulate. A phlebotomist calms a nervous patient with a few words. A nursing assistant notices when something feels off before any sensor could. An embalmer honors the dead with care that only another human can offer.

Life on Earth, it turns out, is defined by connection. Our bodies, our emotions, and our ability to trust one another form the bedrock of what makes us irreplaceable. No machine can hold someone’s hand during a difficult moment or take responsibility when something goes wrong.

For readers wondering where they fit in a world of rapid automation, the answer may not lie in learning new software or chasing credentials. Instead, it might rest in asking a different question. What can I offer that only a human can? Empathy, accountability, physical presence, and the willingness to be there when it matters most. In a world racing toward efficiency, these may be the most valuable things we have left to give.

Loading...