Both: AI will eliminate many existing jobs/tasks while creating new ones and transforming far more, with the net long-term effect historically favoring creation—if adaptation, education, and policy keep pace.
Displacement Is Real and Already Happening
AI (especially generative models, automation, and robotics) excels at routine, predictable, data-heavy, or pattern-based work. Clear impacts include:
•Administrative, clerical, and basic cognitive tasks: Data entry, simple report writing, basic coding/boilerplate, bookkeeping, scheduling, and first-line customer support/chat. Tools already handle large volumes of this.
•Certain creative and knowledge work: Drafting marketing copy, basic graphic design, translation, summarization, and parts of legal/document review or junior analysis. Output quality varies, but volume and speed displace headcount in high-volume settings.
•Physical/repetitive roles: Manufacturing, warehouse sorting, and potentially driving/logistics as autonomy improves (though physical world complexity slows this).
•Scale estimates from sources like Goldman Sachs, McKinsey, and the World Economic Forum have projected hundreds of millions of roles globally affected or transformed over a decade, with concentrated pain in mid-skill white-collar and some blue-collar work. Entry-level and routine jobs face the highest near-term risk.
Displacement is uneven: faster in high-adoption sectors/countries, slower where human judgment, relationships, physical dexterity in unstructured environments, or regulation dominate. It often hits tasks first, then roles, and creates transition friction, wage pressure in exposed occupations, and inequality if skills don't transfer.
Creation and Transformation Are Also Substantial
Technology shifts rarely produce pure net job loss over decades. AI follows the pattern of prior waves (mechanization, computers, internet):
•Direct new roles: AI engineers, data scientists, prompt/context engineers, AI trainers/evaluators, safety/alignment specialists, MLOps, robotics technicians, AI ethicists/compliance experts, and specialists who integrate AI into domain workflows (healthcare AI, legal AI tooling, etc.). Demand here has grown rapidly.
•Augmented and higher-productivity work: Most jobs will change rather than vanish. Radiologists use AI for image screening but still diagnose and communicate; programmers use copilots to ship more; teachers/personalize with AI tools; scientists accelerate discovery. Productivity gains can expand output, lower costs, and grow demand—creating roles in management, complex problem-solving, sales, care, and oversight of AI systems.
•New industries and demand: Entire categories emerge around AI infrastructure, personalized education/entertainment, AI-enabled R&D (drug discovery, materials), creative tools that lower barriers (more creators), and services that become newly economical. Historical parallels: ATMs reduced some teller tasks but banking jobs grew with more branches/services; spreadsheets didn't eliminate accountants.
•Forecasts (WEF Future of Jobs and similar) often project net job growth alongside churn, with growth in healthcare, green tech, education, STEM, and human-centric services offsetting losses. AI can also reduce drudgery, potentially improving job quality in remaining roles.
Net effect depends on speed of adoption vs. creation of complementary work. Short/medium term: visible losses and disruption in specific sectors. Longer term (decades): evidence from economic history and productivity-driven growth leans toward more opportunities overall, plus higher living standards, though distribution matters enormously.
What Determines the Outcome
•Complement vs. substitute: Jobs heavy on social intelligence, complex physical skills, novel creativity, ethics, or accountability are more complementary. Pure prediction/optimization tasks are more substitutable.
•Human adaptation: Reskilling, continuous learning, and shifting toward AI-leveraging skills (problem framing, verification, domain expertise + tools, interpersonal work). Education systems and employers lag here.
•Policy and institutions: Retraining programs, portable benefits, R&D incentives, competition policy (to avoid concentrated gains), and managing transition costs. Extreme scenarios (rapid AGI-like capabilities) raise harder questions about labor share and may need new social contracts; current narrow/ generative AI does not.
•Productivity and demand: If AI boosts output broadly and costs fall, new demand appears (as with past tech). Bottlenecks in energy, chips, regulation, or trust can slow both destruction and creation.
•Measurement issues: Official stats lag; "jobs" include tasks within roles; gig/new forms of work appear differently.
Practical Takeaway
AI is a general-purpose technology like electricity or computing—expect large structural change, not simple "end of work." Individuals: build scarce complementary skills (judgment, taste, relationships, systems thinking), learn to use AI tools deeply in your domain, and treat continuous upskilling as normal. Organizations: redesign workflows around human+AI teams rather than pure replacement. Society: focus on reducing friction of transitions rather than halting the tech.
Predictions vary and timelines are uncertain; capabilities, adoption, and second-order effects (new business models) will clarify over the next 5–15 years. The evidence so far shows simultaneous destruction, creation, and—most commonly—transformation of work. The societies and people who capture the upside will be those who invest in complementary human capital and flexible institutions.