AI will reshape Romanian work rather than simply replace it: in the next 24–36 months automation is expected to transform many roles and widen skill and wage polarization instead of causing widespread job loss. Rigid public-sector rules and uneven retraining capacity will slow adjustment and amplify collective-bargaining and strike risks compared with a more flexible private sector.

Main Finding

Over the next 24–36 months in Romania, AI and automation will primarily reconfigure jobs rather than cause wholesale replacement. Expect pronounced task transformation and a rise in “transformed” employees, with skill and income polarization: declines in routinized roles and growth in AI-orchestration, data stewardship, and higher-complexity relational roles. Short-term net employment loss is heterogeneous by sector and firm digital-intensity; in highly standardizable entities workforce reductions of roughly 10–30% are plausible, offset partially by new roles and redeployments.

Key Points

• Nature of change

  • Dominant scenario = task transformation, not total occupational automation (aligns with OECD/ILO/World Bank findings).
  • Heterogeneous impact by sector, gender, and education level; more disruption in routinized back-office, scripted retail, tele-sales, low-complexity BPO.
  • Growth in roles: AI/RPA orchestration, operational data analysts, algorithmic-risk assessors, data stewards, quality-control, customer-success and consultative sales.

• Labour-market and wage effects

  • Compression expected in the lower tail of the wage distribution (high-automatability roles).
  • Premiums for hybrid competences (human decision-making + AI skills); rising skill-based wage dispersion.
  • Pressure on real wages unless productivity gains are shared via bargaining or policy instruments.

• (Re)training and credentials

  • Emphasis on short, demonstrable credentials (micro-credentials, 6–12 weeks) targeted to applied tasks.
  • Public funding avenues: ESF+, ANOFM active labour-market programs; pilots are already underway.
  • Public-sector curricula need focus on AI ethics/compliance, algorithmic decision evaluation, exception management, and citizen communication.

• Collective bargaining and industrial relations

  • Law No. 367/2022 grants broad scope for social partners to negotiate digitalization/AI topics (e.g., protection, retraining, algorithmic transparency, joint committees, transition calendars).
  • Risk factors for conflicts: lack of prior information/consultation, opaque redeployment criteria, wage erosion.
  • De‑risking: joint technology–social-dialogue committees; inclusion of reconversion/protection clauses in CBAs; accelerated conciliation/mediation/arbitration per statutory sequence.
  • Public sector: structured procedural stages for collective disputes (information → consultation → conciliation → mediation → arbitration); procedural compliance critical.

• Public vs private differences

  • Private: faster adoption in firms with standardizable processes; short-term job transformations and selective reductions.
  • Public: reductions expected in front/back-office routine components, with redeployments to inspection, auditing, data governance; pace constrained by budget/IT investment.
  • Remuneration in public sector should be recalibrated to reflect certified digital competencies and performance-based, auditable indicators (resolution time, service quality, error reduction).

Data & Methods

• Research design: multiaxial analytical framework across four dimensions: A) Number of employees (stock and flows: eliminated / transformed / retained); B) (Re)training and occupational mobility (transition pathways, skill validation); C) Wages (level, structure, dispersion, skill premiums); D) Collective conflicts/strikes (triggers, resolution stages, frequency).
• Data sources: OECD, ILO, World Bank, European Commission studies; Romanian administrative sources — Ministry of Labour and Social Solidarity (MLSS), National Institute of Statistics (NIS), National Employment Agency (ANOFM), Ministry of Finance; sectoral shares used to proxy exposure (services ~42%, industry ~30%, agriculture ~9%, construction ~8%).
• Indicators: employment stocks and flows, re-employment time, micro-credential validation rates, wage percentiles/volatility (inflation-adjusted), counts and procedural metrics for collective disputes.
• Limitations: heuristic assumptions about automation, sensitivity to data quality, actual pace of tech adoption, and local collective-bargaining configurations. Conclusions emphasize short-term (24–36 months) horizon and require dynamic recalibration as empirical uptake evolves.

Implications for AI Economics

• Labour-demand modelling: incorporate task-level transformation and role polarization rather than binary job destruction. Models should separately estimate declines in routinized tasks and growth in hybrid-AI orchestration occupations.
• Distributional outcomes: expect increased wage inequality driven by skill-premia for hybrid competencies and compression at automatable lower-wage roles — requires integration of bargaining and institutional factors into distributional models.
• Adjustment costs & transition dynamics: policy and bargaining institutions (CBAs, joint committees, dispute-resolution mechanisms) materially affect adjustment speed and social costs; including these institutional frictions improves realism of macro/sectoral projections.
• Skills policy design: evidence favors short, stackable micro-credentials closely tied to employer demand; evaluate cost-effectiveness of ESF+/ANOFM-funded programs and track re-employment rates and wage outcomes to measure returns.
• Public-sector constraints: budget-driven adoption lags and procurement/IT choices (e.g., APIs vs. ad-hoc workflows) should be modelled as supply-side frictions that shape timing and magnitude of public-employment reallocation.
• Policy levers to preserve demand and share gains: collective bargaining mandates for training/reconversion, wage-protection clauses, transparency/auditability requirements for algorithmic decisions, and mechanisms to share productivity gains (sectoral or firm-level) are critical to avoid adverse distributional effects.
• Research agenda: collect firm-level adoption timelines, matched employer–employee datasets on re-training uptake and outcomes, and detailed records of CBA provisions on AI to quantify causal effects of institutional responses on employment and wages.