Older companies struggle to attract AI talent: a one-standard-deviation increase in firm age is linked to a 5.2% lower share of AI workers, though R&D spending and modern infrastructure can reverse the trend; diverse boards and slow decision-making in mature firms may further slow AI integration, while firms that do integrate AI talent see gains in valuation and efficiency.

Main Finding

Older firms are less likely to employ AI-skilled workers: a one standard-deviation increase in firm age is associated with a 5.2% reduction in the share of AI workers. This negative effect is economically meaningful but can be mitigated by investments in R&D and capital expenditure. Board composition (higher shares of female and minority directors) is associated with lower AI workforce shares in older firms, possibly via slower consensus-building or greater risk aversion. Firms that overcome age-related barriers and integrate AI employees show improved firm performance (higher Tobin’s q and better profitability measures).

Key Points

• Principal result: Firm age negatively predicts the internal AI workforce share; a 1 SD rise in age → −5.2% in AI-worker share.
• Moderating factors:
  • R&D intensity and capital expenditures attenuate the negative effect of firm age on AI hiring.
  • Board diversity (female and minority directors) in older firms is linked to lower AI workforce shares; interpreted as mixed governance effects (better oversight but potentially slower decision processes for complex technological change).
• Performance link: Firms that increase AI-worker share (including the portion attributable to age-related variation) realize statistically significant gains in market valuation (Tobin’s q) and operational profitability (EBIT/Total Assets, ROA).
• Theoretical framing: Contrasts the “legacy challenge” (structural inertia, legacy IT, cultural resistance) with the “legacy advantage” (accumulated data, financial resources, networks).
• Caveat: The AI workforce measure captures internal AI-skilled employees (from resumes and job-posting text) and may understate firms’ use of external AI vendors or outsourced solutions.

Data & Methods

• Data source: Novel firm-level AI workforce measure from Babina et al. (2024), constructed via textual/NLP analysis of resumes and job postings to identify AI-related skills and roles. Sample covers U.S. firms (details of years/sampling in the original dataset).
• Outcome: Share of AI workers (stock of AI-skilled employees and hiring demand).
• Key explanatory variable: Firm age (corporate life-cycle perspective).
• Empirical approach:
  • Cross-sectional analyses with controls for firm characteristics.
  • Long-differences regressions following Babina et al. (2024) to examine changes over time.
  • Robustness checks: propensity score matching and entropy balancing to reduce confounding and to validate that observed differences are linked to firm age rather than covariates.
  • Interaction terms to test moderating roles of R&D intensity, capital expenditures, and board composition.
• Identification and limitations:
  • Authors use balancing and difference specifications to bolster causal interpretation, but potential endogeneity remains (e.g., unobserved factors that affect both firm age-related trajectories and AI adoption choices).
  • The AI-worker measure focuses on internal hiring signals and may not capture external contracting, platform usage, or other non-hiring modes of AI adoption.

Implications for AI Economics

• Diffusion dynamics: Firm age is an important determinant of AI diffusion within the economy — older firms may slow aggregate adoption unless they invest in complementary assets (R&D, IT modernization).
• Productivity and concentration: If older, large incumbents struggle to build internal AI capability, adoption patterns across firm cohorts could affect productivity growth, market structure, and industry concentration. Conversely, older firms that invest can generate sizable value gains, reinforcing superstar-firm dynamics.
• Complementarities matter: The findings emphasize that human capital (AI workers) interacts with tangible/intangible investments (capex, R&D). Models of AI-driven productivity should include complementarities between workforce skills and firm-level investments.
• Governance and organizational design: Board composition affects technological decisions in nuanced ways. Policy and governance research should consider how diversity, decision speed, and risk preferences jointly shape technology adoption.
• Policy levers:
  • Targeted incentives (tax credits, matching grants) for legacy firms to modernize IT and boost R&D could accelerate internal AI capability building.
  • Support for upskilling and internal retraining programs in mature firms may be more effective than generic labor-market interventions.
• Research directions:
  • Better causal identification (instrumental variables, policy shocks, or quasi-experiments) to separate firm-age effects from selection/omitted variables.
  • Extend measurement to capture external AI use (vendors, cloud services, AI-as-a-service) and hybrid sourcing strategies.
  • Industry- and size-specific analyses: do the legacy effects differ across sectors with varying regulation, data availability, or capital intensity?
  • Micro mechanisms: study how legacy IT architecture, hiring practices, and managerial incentives mediate the age–AI adoption link.
• Practical takeaway for managers: To avoid being left behind, mature firms should pair hiring of AI talent with explicit investments in R&D, infrastructure, and governance processes that speed decision-making about technology.