Enterprise generative and agentic AI tools are linked to measurable productivity and workflow gains across firms, but current evidence offers no consensus on whether these technologies produce net job losses or wage changes; policymakers should treat labor-impact claims as provisional and prioritize targeted reskilling and rigorous field evaluations.

Main Finding

Enterprise generative, multimodal, and agentic AI technologies are reshaping firm organization and labor markets in mixed ways: they raise productivity and enable new tasks, roles, and business models while simultaneously automating tasks that can cause job displacement. Net effects on employment, wages, and unemployment are ambiguous and context-dependent — driven by task complementarities/substitutabilities, firm adoption, upskilling/reskilling responses, and regulatory/market conditions. Policy and firm-level interventions (reskilling, active labor-market supports, privacy/monitoring safeguards) are therefore central to realizing net job-creation and equitable wage outcomes.

Key Points

• Scope and core claim

  • The paper is a systematic review focused on enterprise deployments of generative, multimodal, and agentic AI (including digital twins, computer vision, IoT-enabled robotics, and multi-agent systems) and their effects on job loss and creation.
  • Authors find that these AI systems simultaneously enable task automation (risking displacement) and task augmentation (creating higher-skill roles), so outcomes vary across sectors, occupations, and firm strategies.

• Mechanisms through which AI affects jobs

  • Task substitution: agentic and robotic systems perform multistep tasks (e.g., IoT-driven factory workflows, computer-vision inspections) replacing routine manual and some cognitive tasks.
  • Task augmentation: generative and multimodal systems augment human productivity (decision support, design, knowledge management), creating demand for new complementary skills.
  • Reallocation & organizational redesign: AI-driven workforce scheduling, team-formation, and performance monitoring change job design, rotation, and promotion pathways.
  • Talent pipeline effects: HR automation (screening, recruitment, training personalization) can speed hiring and influence skill development, potentially easing re-employment but also intensifying monitoring.
  • Productivity → labor demand ambiguity: productivity gains can expand output and labor demand (job creation), but cost reductions and scale economies may reduce employment in affected tasks.

• Empirical/analytical conclusions highlighted

  • Effects on wages, unemployment, and job quality are heterogeneous: some studies predict wage increases (when AI complements skills) while others predict wage pressure (when AI substitutes tasks).
  • Generative/multimodal AI broadens the set of tasks that can be automated (including creative and interpretive tasks), increasing uncertainty about net job impacts.
  • Effective mitigation of displacement requires upskilling/reskilling, organizational adaptation, and policy interventions.

• Limitations noted by authors

  • Review limited to publications from 2024–2025, with 51 papers analyzed — short time window and early literature.
  • Rapidly evolving technology and firm practices imply high uncertainty and need for more causal empirical work.

Data & Methods

• Search strategy

  • Databases: ProQuest, Scopus, Web of Science.
  • Search terms: "enterprise generative/multimodal/agentic artificial intelligence" + "job loss/creation".
  • Timeframe: publications from 2024–2025.
  • Initial hits: 364 sources; final corpus analyzed: 51 papers.

• Review protocol & tools

  • PRISMA framework used for reporting and screening; Shiny app used for screening workflow visualization.
  • Bibliometric and visualization: Dimensions for mapping, VOSviewer for layout algorithms.
  • Evidence-synthesis, screening, and quality tools employed (examples): Abstrackr, AMSTAR, AXIS, CADIMA, CASP, DistillerSR, Eppi-Reviewer, JBI SUMARI, MMAT, Nested Knowledge, PICO Portal, Rayyan, ROBIS, SRDR+, SWIFT-Active Screener, Systematic Review Accelerator.
  • Three main thematic clusters identified:
  • Agentic/robotic industrial technologies and autonomous decision systems for organizational performance.
  • Generative/multimodal AI-augmented HR decision algorithms (recruitment, monitoring, upskilling).
  • Digital twin / enterprise operations and forecasts of job replacement and layoffs.

• Evidence base characteristics

  • Mix of conceptual, modeling, and early empirical studies; emphasis on case studies and technological demonstrations in manufacturing, HR, and supply-chain contexts.
  • Heavy reliance on recent (2024–2025) literature — the field is nascent with few long-run causal estimates.

Implications for AI Economics

• For modeling and empirical research

  • Task-level, occupation-task mapping matters: researchers should adopt granular task-based frameworks (not only occupation-level aggregates) to identify complementarities vs. substitutions.
  • Need for causal identification: exploit firm-level adoption, staggered rollouts, and instrumental variables to estimate heterogeneous employment/wage effects.
  • Incorporate multimodality and agentic behavior: models should allow AI agents to perform multistep, multimodal tasks and to coordinate across agents (network/multi-agent models), not just single-task automation.
  • Measure non-wage outcomes: job quality, skill polarization, worker surveillance, turnover, and re-employment trajectories should be integrated into impact assessments.
  • Data priorities: matched employer–employee microdata, task survey data, firm adoption records, digital-trace/telemetry from enterprise AI systems, and administrative earnings/employment panels.

• For policy and labor-market design

  • Active labor-market policies matter: public investment in rapid reskilling, portable credentials, and targeted transition support will affect net employment outcomes.
  • Social insurance & redistribution: short-run displacement risk argues for unemployment insurance, wage insurance pilots, and subsidized hiring/training to smooth transitions.
  • Regulation of workplace monitoring: HR/AI monitoring tools can affect worker privacy, bargaining, and behavior; regulators should set transparency, fairness, and data-governance standards.
  • Promote AI complementarity: incentives (tax credits, co-funding) for uses of AI that clearly complement human skills and expand labor demand can steer adoption toward job-creating paths.
  • Regional and sectoral policies: because impacts vary by sector and region, targeted interventions are more effective than one-size-fits-all measures.

• For firms and labor economics interpretation

  • Short-run vs long-run tradeoffs: firms may realize immediate cost-savings and process improvements, but long-run industry dynamics (new products, expanded markets) can create jobs; economists should capture dynamic general-equilibrium effects.
  • Heterogeneous firm responses: small/medium enterprises (SMEs) may experience different constraints and benefits from agentic AI than large firms — implications for market structure and wage dispersion.
  • Role of HR automation: automated hiring and skill assessment reshapes labor supply matching — research should evaluate impacts on hiring frictions, sorting, and discrimination.

• Policy-research agenda

  • Evaluate large-scale reskilling program effectiveness tied to AI adoption.
  • Study distributional impacts across skill groups, age cohorts, gender, and regions.
  • Develop methods to value AI-induced productivity gains vis-à-vis displaced labor costs.

Summary caveat: conclusions are conditioned on a limited and very recent literature (2024–2025). Stronger causal evidence and broader time coverage are needed to pin down net employment and wage effects of enterprise generative, multimodal, and agentic AI.