Generative AI forces a rethink of professional expertise and firm design: existing theories of work and organizations need extension to capture new, evolving interdependencies between professionals and generative machine learning, the paper argues.

Main Finding

The paper argues that existing theoretical approaches to professional work and professional service firms are insufficient for the “generative machine learning” (generative ML) era. It proposes a new, synthesis-style conceptual approach that foregrounds emerging and future interdependencies between professionals and generative ML systems, requiring both extensions and reimagining of theories about expertise, tasks, firms, and organizational change.

Key Points

• Generative ML is qualitatively distinct from earlier digital and algorithmic technologies (e.g., by being co-creative, probabilistic, capable of synthesizing diverse knowledge, and able to produce communicative artifacts), and those distinctions matter for theorizing professional work.
• The relationship between professionals and technology should be reframed as evolving interdependencies rather than simple substitution or augmentation; professionals and models form hybrid work systems with shifting boundaries of authority and expertise.
• Core theoretical objects that need revision include:
  • Expertise: from individual possession of tacit knowledge to distributed, hybrid expertise negotiated between humans and models.
  • Work and tasks: task bundles will be reconfigured (new tasks created, some tasks automated, others enhanced), altering task decomposition and coordination needs.
  • Organizations and firms: governance, knowledge management, incentive structures, and firm boundaries will shift as firms adopt generative ML (e.g., new coordination of human-AI teams; implications for outsourcing and platformization).
• Theorizing change must account for emergent, path-dependent dynamics (learning effects, model updates, data feedback loops) rather than static comparative statics.
• Ethical, legal, and epistemic consequences (accountability, trust, validation of outputs) are integral to how professions evolve with generative ML.

Data & Methods

• The paper is conceptual/theoretical. Methods used include:
  • Synthesis of existing literatures on digital technologies, algorithmic systems, and AI.
  • Conceptual analysis highlighting the distinctive technical and organizational properties of generative ML and their theoretical implications.
  • Development of an agenda and theoretical lenses for future empirical work.
• No original empirical dataset or quantitative analysis is reported; the contribution is primarily theoretical framing and literature-informed argumentation.

Implications for AI Economics

• Labor demand and skill composition:
  • Expect shifts toward hybrid skill sets (human judgment, prompt design, oversight, model evaluation) and away from some routine cognitive tasks; both complementarities and partial substitution effects are likely.
  • Potential for rapid re-skilling needs within professional services; differential adoption could create heterogeneity in labor outcomes across firms and occupations.
• Productivity and value capture:
  • Generative ML may raise productivity in many professional activities (drafting, research, design), but value capture will depend on how firms coordinate human-AI teams and appropriate gains (firm boundaries, IP, platforms).
  • Network and scale effects in model/data can amplify winner-take-all tendencies across firms and platforms, with implications for market structure.
• Organization of production:
  • New modularization of work (human roles focused on oversight, validation, creative synthesis) and possible redefinition of firm boundaries (insourcing AI capabilities vs. using AI platforms).
  • Changes in pricing models for professional services (bundled offerings, subscription/platform models, pay-per-output with model-generated content) and fee structures tied to monitoring/assurance services.
• Investment and capital composition:
  • Increased returns to data and AI-capital could shift investment away from labor towards model-building and data infrastructure; capital-labor substitution will vary by task and firm.
• Market competition and entry:
  • Lower marginal cost of producing knowledge artifacts may lower barriers to entry for some services but raise them where large training datasets and compute are required.
• Regulation, trust, and externalities:
  • Economic models must incorporate costs of verification, liability, and trust-building; these frictions influence adoption, pricing, and welfare outcomes.
• Research and policy agenda:
  • Empirical work needed on task-level complementarities, firm-level adoption strategies, distributional effects within professions, returns to retraining, and regulatory impacts.
  • Policy considerations include workforce transition support, standards for model transparency and auditability in professional services, and measures to prevent excessive concentration.

If you want, I can (a) map these implications to specific professional sectors (e.g., law, accounting, medicine, consultancy), (b) propose testable hypotheses and empirical designs to study the interdependencies, or (c) produce a short literature map of key empirical papers relevant to each implication.