AI adoption bolsters corporate resilience in Indonesia — large, digitized manufacturers gain the most, and better regional broadband partly explains the advantage.

Main Finding

AI technology penetration significantly enhances corporate resilience among Indonesian listed firms (2015–2024). The positive effect is robust to alternative resilience and AI measures, high‑dimensional fixed effects, and endogeneity treatments reported by the authors. Regional broadband infrastructure partially mediates the AI → resilience link. Heterogeneity tests show the effect is concentrated in large firms, higher‑digitization firms, and manufacturing firms, and is not significant for SMEs, low‑digitization firms, or non‑manufacturing sectors.

Key Points

• Sample and scope: Panel of firms listed on the Indonesia Stock Exchange, 2015–2024 (2,465 firm‑year observations after screening/winsorizing).
• Baseline result: In a firm & year fixed‑effects regression, the AI penetration coefficient ≈ 0.167 (p<0.01) on the dependent variable “Resilience” (sales growth rate = current year sales − prior year sales / prior year sales).
• AI measure(s):
  • Primary: R&D expenditure / operating revenue (used as a proxy for AI technology penetration).
  • Alternative: ICT investment (capital expenditures on software/hardware/IS scaled by revenue or assets).
• Resilience measure(s):
  • Primary: Sales revenue growth rate (captures market recovery/adaptation).
  • Alternative: Ratio of reserve funds to redundant resources (financial buffer).
• Controls: Firm size (ln assets), leverage, cash ratio, fixed assets share, total turnover/asset, asset growth, ROA; firm and year fixed effects included.
• Robustness checks: alternative dependent and independent variables, high‑dimensional fixed effects, winsorizing, and endogeneity treatment (authors state results remain robust).
• Mechanism: Regional broadband infrastructure serves as a partial mediator—better broadband strengthens the translation of AI penetration into resilience gains.
• Heterogeneity: Positive AI→resilience effect stronger in:
  • Large enterprises vs SMEs
  • Firms with higher digitization levels
  • Manufacturing sector vs non‑manufacturing

Data & Methods

• Dataset: Indonesian listed firms (IDX), 2015–2024; standard data cleaning: remove non‑operational firms, drop observations with key missing values, winsorize continuous vars at 1st/99th percentiles.
• Econometric specification:
  • Main model: Resis_it = α0 + α1 AI_it + α2 Controls_it + δ_i + μ_t + ε_it
  • Fixed effects: firm and year fixed effects to control for time‑invariant heterogeneity and common shocks.
  • Standard errors: robust (reported).
• Variable construction:
  • Resilience = (Current year sales − Previous year sales) / Previous year sales.
  • AI penetration = R&D / operating revenue (alternative: ICT investment).
  • Broadband mediator = regional broadband indicators (e.g., fiber coverage or fixed broadband subscriptions per 100 people).
• Empirical strategy: baseline FE regressions; robustness via variable substitution and high‑dimensional fixed effects; mediation analysis for broadband; endogeneity addressed according to authors (details of instruments/identification approach not provided in the excerpt).

Implications for AI Economics

• For firms (managers/boards):
  • Investing in AI (measured by higher R&D intensity and ICT spending) is associated with greater ability to withstand and recover from shocks. Prioritize AI-driven capabilities for operational flexibility, risk prediction, and resource allocation.
  • Complement AI adoption with stronger digital infrastructure and internal digitization to realize resilience gains—especially important for SMEs that show weaker benefits.
• For policymakers:
  • Broadband and digital infrastructure are critical complementarities that amplify the welfare/resilience benefits of firm‑level AI adoption. Public investment or incentives to expand reliable high‑speed broadband can increase the returns to private AI investment.
  • Targeted support is warranted for SMEs and low‑digitization firms (subsidies, training, adoption support) because they appear less able to convert AI investments into resilience.
• For investors and analysts:
  • R&D intensity and ICT capex can serve as signals of a firm’s resilience posture in emerging markets. Consider AI/digitalization metrics when assessing downside risk and recovery potential.
• For researchers:
  • The paper provides empirical evidence from an emerging market context (Indonesia) supporting a technology→resilience channel mediated by infrastructure. Future work should:
    • Improve causal identification (detailed IVs, natural experiments) and report instrument strategies transparently.
    • Use more direct measures of AI adoption/use (software/modules, AI-enabled product lines, job task reallocation, text‑mined disclosures).
    • Explore dynamic effects, implementation costs, organizational disruption, and welfare trade‑offs—especially for SMEs and labor markets.
    • Examine generalizability across other emerging economies and subnational contexts.

Limitations (noted): - AI proxy (R&D/Revenue) is an imperfect measure of actual AI adoption/use; it conflates general innovation spending with AI‑specific activity. - Sales growth as a resilience metric can be influenced by pricing/one‑off items; authors partially address this with alternative financial buffer measure. - Details of the endogeneity treatment were not provided in the excerpt—interpretation should be cautious until identification strategy is fully reviewed.