Spanish blue‑chip firms that signal stronger AI adoption report intangible assets more visibly and positively, especially in finance and telecom; the effect appears driven by narrative disclosure rather than higher R&D intensity.

Main Finding

There is a robust, positive association between firms’ AI adoption intensity (AIAI) and the visibility of their intangible assets (IAVS) in external reporting for IBEX 35 firms over 2019–2024. AI-intensive firms disclose intangibles more frequently and with higher narrative quality; this effect is stronger in digitally mature sectors (finance, telecom) and weaker in traditional sectors (construction, utilities). Sentiment/topic analysis shows AI-related intangible narratives are predominantly positive (brand, sustainability, talent). R&D intensity did not predict intangible visibility, suggesting a shift toward narrative-driven valuation.

Key Points

• Sample and scope: 35 IBEX firms, balanced panel 2019–2024 (210 firm-year observations).
• New measures: AI Adoption Intensity (AIAI) and Intangible Asset Visibility Score (IAVS) — both disclosure-based indices developed for this study.
• Main empirical result: AIAI positively and statistically significantly correlates with IAVS; sectoral heterogeneity (finance and telecom lead).
• NLP results: AI-enhanced disclosures framed positively; topic models emphasize brand value, sustainability, and human capital.
• R&D intensity is not a significant predictor of intangible-asset visibility in the models.
• Interpretation: AI functions both as a meta-capability (internal value creation) and as a signaling device (external narrative), so higher AI disclosure co-evolves with richer presentation of intangibles.
• Limitations highlighted: reliance on disclosure proxies (may understate internal capabilities), observational design (no strong causal claims), and Spanish/IBEX 35 focus (external validity concerns).
• Practical recommendation: firms should align AI strategy with reporting frameworks; regulators should update disclosure standards (notably under CSRD and AI Act) to reflect AI’s role in intangible capital formation.
• Social/ethical note: Transparent AI communication can support trust and responsible governance; risk of “AI washing” should be monitored.

Data & Methods

• Design: Multi-method study combining panel regressions, NLP (topic modeling + sentiment analysis), and qualitative case profiles (three illustrative IBEX 35 firms).
• Sample: All firms listed on IBEX 35 as of 31 Dec 2024, observed yearly 2019–2024 (210 observations), grouped into sectors (finance, telecoms, energy/utilities, construction/industrial).
• Sources: Audited annual reports and ESG reports (primary), investor presentations and AI reports (secondary), media coverage (tertiary; used only to corroborate).
• Index construction:
  • AIAI: disclosure-based index measuring depth/breadth of AI embeddedness across business functions, governance and human capital (codes for pilot vs embedded deployment).
  • IAVS: composite measure of frequency and substantive quality of intangible-asset disclosures and recognition in financial statements.
• Coding protocol: firm-year corpora read by coders; hierarchy favored audited reports/ESG documents; media-only claims limited to low AIAI scores or excluded from IAVS upgrades.
• Econometric approach: panel data regressions testing AIAI → IAVS, controlling for overall report verbosity and other firm controls; alternative specs tested (narrative-only index, intangible-assets ratio).
• Textual methods: NLP for semantic framing, topic modeling to identify dominant themes, sentiment analysis to assess emotional valence of AI-intangibles narratives.
• Qualitative component: three case profiles to illustrate mechanisms (how AI strategy, governance and reporting co-evolve).

Implications for AI Economics

• Valuation and asset measurement:
  • AI increases the visibility of intangibles, meaning markets may update beliefs about future returns when firms disclose AI integration — but disclosure-driven visibility can create valuation premia that reflect narrative quality as much as underlying productivity.
  • Traditional accounting/value models that rely on tangible metrics may underweight AI-enabled intangible formation; integrating AIAI-like measures could improve firm valuation accuracy.
• Pricing and market efficiency:
  • Sectoral heterogeneity implies cross-sectional mispricing risk: markets may systematically undervalue AI-enabled intangibles in traditional sectors unless disclosures catch up.
  • Positive framing and signaling effects raise the prospect of “AI washing” driving short-term price effects absent commensurate economic gains — important for event studies and asset-pricing tests.
• Corporate strategy and resource allocation:
  • AI as a meta-capability suggests complementarity between AI investments and investments in human, structural and relational capital — economic returns to AI may depend on these bundles.
  • The apparent weak link between R&D intensity and disclosure visibility suggests firms may extract valuation benefits via narrative/communication strategies as much as via R&D output; models of investment choice should account for disclosure strategy.
• Policy and regulation:
  • Disclosure standards (CSRD, AI Act) should be refined to standardize reporting on AI capabilities and their links to intangibles so investors can distinguish substantive adoption from signaling.
  • Better disclosure templates and verification could reduce information asymmetry and mitigate reputational/market risks from overstated AI claims.
• Research opportunities:
  • Causal identification: natural experiments or instrumented approaches to separate genuine productivity effects of AI from disclosure/signaling effects.
  • International and cross-market comparisons to test generalizability beyond IBEX 35.
  • Incorporate AIAI/IAVS into asset-pricing, corporate finance, and productivity studies to quantify how AI-enabled intangibles affect cost of capital, investment returns and market liquidity.

If you want, I can (a) extract the exact variable definitions and regression specifications from the paper appendices, (b) produce a one-page slide-ready summary, or (c) map how to incorporate AIAI/IAVS into an empirical asset-pricing model. Which would be most useful?