Mature service firms favor incremental, efficiency-oriented AI rollouts, yet those adoptions do not boost returns on assets or equity in the observed sample.

Main Finding

Mature Finnish service firms are more likely to adopt AI (positive, significant association), but AI adoption—measured as frequency of AI-related terms in annual reports—is not associated with higher short-term financial performance (ROA or ROE) in the sample.

Key Points

• Research questions: (1) Does firm maturity influence AI adoption? (2) Does AI adoption affect firm financial performance?
• Primary empirical result: firm maturity (RE/TA) is strongly positively correlated with AI adoption (r = 0.65, p < 0.01). In OLS regression, maturity predicts AI adoption (β = 0.59, t = 4.30); model R2 = 0.48.
• AI adoption showed no meaningful relationship with ROA or ROE in the observed period.
• Interpretation offered by authors: mature firms tend to pursue incremental, efficiency-driven AI integration (control, optimisation) rather than disruptive, growth-seeking AI deployments typical of younger firms.
• Novelty claim: linking organisational life-cycle (firm maturity) as an antecedent to AI adoption in service firms; re-examines AI → firm performance relationship in Finnish services.

Data & Methods

• Sample: 12 publicly listed Finnish service firms from ORBIS; panel period 2021–2023 (n reported as 36 observations).
• AI adoption measure: content analysis of annual reports — frequency-based score 0–3 based on occurrences of terms (artificial intelligence, AI, machine learning, deep learning): 0 = none, 1 = 1–5 mentions, 2 = 6–15, 3 = >16 mentions. Six firms explicitly referenced AI; six did not.
• Firm maturity: retained earnings / total assets (RE/TA).
• Controls: firm age (years since incorporation), firm size (ln net annual sales).
• Financial performance outcomes: ROA (net income / total assets) and ROE (net income / equity).
• Analysis: correlation matrix and OLS regressions (no reported instrumental variables, difference-in-differences, or causal identification strategy).

Implications for AI Economics

• Diffusion and adoption dynamics: Evidence that incumbency/maturity can increase AI adoption (likely incremental, efficiency-oriented uses). This challenges a simple view that only young/growth firms are primary AI adopters.
• Returns to AI are not automatic or immediate: Lack of short-term ROA/ROE gains suggests financial returns depend on (a) type of AI use (efficiency vs. growth), (b) complementary investments (digital infrastructure, skills), and (c) longer time horizons.
• Measurement issues matter: Using disclosure frequency as an adoption proxy captures strategic signalling and discourse as much as actual investment or capability; economic studies should triangulate with spending, project counts, or usage metrics.
• Policy and managerial takeaways:
  • Policymakers aiming to boost productivity through AI should consider supporting complementarities (training, integration, standards) not just AI uptake.
  • Managers in mature firms may prioritize low-risk efficiency AI projects; expectations on short-term ROI should be tempered and evaluated against long-run strategic goals.
• Research recommendations: larger samples, richer adoption measures (investment, deployment intensity, task-level automation), longer horizons, and causal designs (IVs, experiments, panel fixed effects) to disentangle selection, reverse causality, and heterogeneous returns across firm types and AI applications.

Limitations to note (from the study): small sample of listed service firms, short timeframe, reliance on disclosure frequency for AI adoption, and absence of causal identification — so generalise cautiously.