Short answer: both—but in different timeframes and for different kinds of work. Most roles will change before they vanish. Historically, general‑purpose technologies (electricity, PCs, the internet) ended up creating more jobs than they destroyed, but the transition was bumpy and unequal. AI looks similar: near‑term disruption with pockets of job loss, longer‑term job growth if businesses, workers, and policymakers adapt.
What AI tends to do
•Automates tasks, not whole occupations. Jobs with many routine, rules‑based tasks face the most pressure; jobs rich in social interaction, physical dexterity, or cross‑disciplinary judgment tend to be complemented.
•Raises productivity. When output per worker rises, firms often expand, which can create new roles—if demand grows and gains are shared.
Where jobs are most at risk (next few years)
•Clerical and back‑office admin (data entry, basic scheduling, form processing)
•Basic customer support and telemarketing
•Routine content creation/translation and simple graphic assets
•Bookkeeping, payroll, and some compliance documentation
•Entry‑level analytical work and rote coding or QA
Where AI is likely to create or expand jobs
•Human‑in‑the‑loop roles: AI product managers, model evaluators, safety/ethics, data stewards, ops, integration engineers
•AI‑enabled specialists: analysts, marketers, designers, software engineers who orchestrate tools rather than do every task by hand
•Trust, risk, and security: cybersecurity, governance, auditing, regulatory compliance
•Services that are hard to automate: healthcare, eldercare, education/tutoring, skilled trades, hospitality, sales, and roles requiring on‑site coordination
•New categories from lower costs and new capabilities (just as the smartphone era created app ecosystems and gig roles)
What will decide the net outcome
•Adoption speed and how work is redesigned (task re‑bundling vs layoffs)
•Whether firms use productivity gains to lower prices or launch new products (demand expansion) vs pure cost‑cutting
•Training and mobility: how fast people can re‑skill and switch occupations
•Policy choices: safety nets, portable benefits, upskilling incentives, and support for job‑creating investment
How to improve your odds personally
•Learn to delegate to AI: prompting, tool selection, verification, and workflow design
•Go up the value chain: problem framing, stakeholder communication, domain expertise, and decision‑making under uncertainty
•Build data and systems literacy: spreadsheets to SQL, basic scripting, API know‑how, and privacy/security fundamentals
•Lean into human strengths: leadership, negotiation, mentoring, creativity, and ethics
•Own outcomes, not tasks: seek roles measured by business results where AI is a lever, not a substitute
•Keep a visible portfolio of projects showing you can combine domain skill + AI tools to deliver results
For employers
•Redesign jobs around tasks; pair automation with upskilling and clear transition paths
•Measure and share productivity gains (bonuses, reduced hours, or growth hiring)
•Invest in change management and governance to avoid “pilot purgatory” and trust failures
Bottom line: Over the next few years, expect visible displacement in routine cognitive work and broad augmentation elsewhere. Over the longer run, AI is likely to create more jobs than it eliminates—but that upside isn’t automatic; it depends on how we redesign work, re‑skill people, and channel the productivity gains. If you share your industry and role, I can map likely impacts and practical next steps for you.