Short answer: both—but the balance depends on timeframe, industry, and how quickly organizations redesign work. Near term, AI will displace tasks (and some jobs) that are routine and text-based; over time, it tends to create new roles and expand demand in AI‑complementary work. Whether net jobs go up or down hinges on adoption speed, productivity gains, and how well workers and firms adapt.
How AI changes jobs
•Task automation, not just job automation: Roles get re-bundled as repetitive research, drafting, classification, and data-cleaning tasks shrink.
•Productivity effects: If output per worker rises and demand expands (lower prices, new products), firms hire more. If demand is flat and cost-cutting dominates, headcount can fall.
•Reallocation frictions: Even if net jobs grow, transitions can be painful—skills mismatch and geographic frictions matter.
Roles most at risk (partial)
•Routine cognitive/clerical: data entry, basic bookkeeping, form processing, transcription, simple customer support, first-draft content.
•Narrow “glue code” or boilerplate programming; straightforward translation/localization.
•Middle layers of roles whose value is mainly synthesizing public information without judgment.
Roles likely to grow or gain value
•AI-complementary experts: healthcare clinicians, teachers/trainers, lawyers, engineers, designers, sales—where judgment, trust, and context matter.
•Technical and adjacent: ML/LLM engineering, data engineering, evaluation/red-teaming, model operations (MLOps/LLMOps), privacy/safety/governance, AI auditors.
•Human-in-the-loop work: domain labeling, data curation, quality assurance, workflow orchestration.
•Entrepreneurship and product roles that turn AI capabilities into services for specific niches.
What this means for you
•Think in tasks: List your core tasks; mark which are routine, text-heavy, rule-based (most automatable) vs. interpersonal, creative, or physical (more durable).
•Skill to complement, not compete: double down on domain expertise, problem framing, communication, data reasoning, and tool fluency with AI.
•Become “AI-augmented”: learn to design prompts, build small automations, check model outputs, and document workflows that pair human judgment with AI speed.
•Build proof of value: keep a portfolio of before/after metrics (time saved, error rates, revenue impact) using AI tools.
For employers
•Redesign jobs around workflows, not titles; redeploy time saved to higher-value tasks.
•Invest in training and change management; pair adoption with upskilling and clear QA.
•Track outcomes beyond speed—quality, safety, customer satisfaction, and risk.
For policymakers
•Support rapid reskilling, portable benefits, and wage/transition assistance.
•Encourage responsible deployment (auditing, safety standards) to unlock demand safely.
Bottom line: In the next few years, expect noticeable task displacement and churn; over the longer run, net employment can rise if productivity gains translate into new demand and if we invest in people. Tell me your industry and role, and I’ll map likely impacts and a targeted upskilling plan.