AI & Work
AI is not coming for your job in some distant future. It’s restructuring work right now, task by task. Understanding which tasks are shifting — and which skills are becoming essential — is the difference between being displaced and being prepared.
What’s Actually Happening
Section titled “What’s Actually Happening”AI automates tasks, not entire jobs. McKinsey estimates 57% of U.S. work hours involve tasks that are technically automatable. But most jobs are bundles of tasks — some get automated, the job changes shape, and the person doing it needs different skills.
Here’s where the impact is hitting hardest right now:
- Customer service — 80% of roles are automatable. Klarna replaced 700 agents with AI, then partially reversed course when quality dropped.
- Entry-level software development — 20% employment decline for ages 22-25 since late 2022.
- Administrative support — 46% of admin tasks are already automatable. IBM halted hiring for HR roles it expects AI to absorb.
- Junior analysts, paralegals, content writers — knowledge work execution is the front line.
- Freelance writing, design, translation — commoditized execution work is collapsing on platforms.
The numbers: 55,000 U.S. job cuts in 2025 explicitly cited AI as the reason, up from roughly 5,000 in 2023. In Q1 2026, nearly 50% of 80,000 tech layoffs cited AI.
Still safe (for now): Skilled trades (HVAC, electrical, plumbing), healthcare direct care, complex physical labor. Only 4-6% of construction and maintenance tasks are AI-automatable.
Skills: What’s Dying vs. What’s Emerging
Section titled “Skills: What’s Dying vs. What’s Emerging”| Becoming Obsolete | Becoming Essential |
|---|---|
| Manual data entry and processing | AI literacy (prompting, evaluating output) |
| Routine report writing | Judgment and decision-making with AI output |
| Basic boilerplate code | Human-in-the-loop oversight and QA |
| First-pass document review | Cross-functional workflow design |
| Scripted customer interactions | Relationship-building and trust |
| Template-based design work | Problem framing (knowing what to ask) |
The World Economic Forum projects 39% of workers’ core skills will change or become obsolete by 2030. McKinsey found demand for “AI fluency” in job postings grew 6.8x from 2023 to 2025.
What Average People Must Understand
Section titled “What Average People Must Understand”AI does not think. Large Language Models predict the next word. They are powerful pattern engines, not reasoning beings. They hallucinate — fabricate confident-sounding wrong answers. Always verify.
AI agents are the next wave. These are AI systems that plan and execute multi-step tasks autonomously — not just answering questions, but doing work. Gartner predicts 40% of enterprise apps will have task-specific agents by end of 2026.
Companies are cutting jobs based on AI’s potential, not proven results. Harvard Business Review flagged this: layoffs are outpacing actual AI productivity gains. The fear of being “left behind” is driving premature cuts.
Terms You Should Know
Section titled “Terms You Should Know”| Term | Plain English |
|---|---|
| LLM (Large Language Model) | The AI engine behind ChatGPT, Claude, and Gemini. Trained on massive text data to generate human-like responses. |
| AI Agent | Software that autonomously plans and executes multi-step tasks — not just answering questions. |
| Prompt | The instruction you give an AI. Better prompts produce dramatically better output. |
| Hallucination | When AI confidently generates false information. Common and dangerous if unchecked. |
| Automation vs. Augmentation | Automation replaces you. Augmentation makes you faster. Most AI today is augmentation. |
| Human-in-the-loop | A human reviews and approves AI output before it’s used. The fastest-growing role pattern. |
What You Can Do Today
Section titled “What You Can Do Today”- Use AI tools this week. ChatGPT, Claude, or Gemini — pick one real work task (drafting emails, summarizing documents, researching) and let AI do it. Experience beats theory.
- Take one free course. Google AI Essentials (about 10 hours, free on Coursera) or similar. Not to become an engineer — to understand what AI can and cannot do.
- Classify your own tasks. List every task in your week. Mark each: (A) AI can do this now, (B) AI will do this soon, (C) requires human judgment or relationships. Shift your time toward C.
- Learn to write clear instructions. This is the new baseline professional skill. Specificity, context, and examples produce dramatically better AI output.
- Build a financial buffer. Goldman Sachs estimates unemployment may bump roughly half a percent during the transition. Having savings reduces the pressure to make fear-based decisions.
The bottom line: The data from McKinsey, WEF, IMF, and BCG all converge — AI is restructuring work task by task, not eliminating jobs wholesale. But the people who refuse to learn the tools will lose ground to those who do. The window to adapt is open now.