Firms that invest in AI-skilled staff achieve measurably better tax planning and modest gains in after-tax returns; the benefits concentrate in complex organizations and where tax teams have strategic influence.
The use of artificial intelligence in decision-making: evidence from the effectiveness of corporate tax strategies
Firms with greater AI-skilled employee intensity are associated with more effective tax planning and modestly higher after-tax returns, with stronger effects in operationally complex firms and when the tax function has higher organizational status.
Abstract We examine whether information processing constraints limit managers’ ability to effectively integrate tax planning and core business strategies (i.e., effective tax planning). We propose that artificial intelligence (AI) tools, such as machine learning, can mitigate these constraints by providing enhanced predictive information for key business decisions (e.g., customer demand, supply chain), thereby reducing processing costs. Using a recently developed firm-year measure of investment in AI-related human capital for a broad sample of U.S. nontechnology firms between 2010 and 2018, we find that AI investment is positively associated with tax effectiveness. This effect is concentrated among more complex firms and those where the tax function holds a higher status. Consistent with AI reducing information processing costs, we find that it improves tax effectiveness by enhancing internal information quality and internal capital management. We provide novel evidence that processing constraints hinder effective tax planning and show that AI can mitigate these constraints.
Summary
Main Finding
Investment in AI-related human capital is positively associated with firms' tax effectiveness (the ability to integrate tax planning with core business strategy). This relationship is strongest for more complex firms and for firms where the tax function has higher status. Evidence suggests the effect operates via improved internal information quality and better internal capital management, consistent with AI reducing information-processing constraints.
Key Points
- Sample & scope: Broad sample of U.S. non-technology firms, firm-year observations from 2010–2018.
- Treatment variable: A recently developed firm-year measure of investment in AI-related human capital.
- Primary result: Higher AI investment → greater tax effectiveness.
- Heterogeneity: Effects concentrated in (a) more complex firms and (b) firms where the tax function holds higher organizational status.
- Mechanisms: Internal information quality and internal capital management improvements mediate the AI → tax effectiveness relationship.
- Interpretation: Supports the hypothesis that information-processing constraints limit managers’ ability to coordinate tax planning with business decisions, and that AI mitigates those constraints.
Data & Methods
- Data: Firm-year panel of non-technology U.S. firms from 2010–2018, using a novel measure of AI-related human capital investment at the firm-year level.
- Empirical approach: Regression analyses relating AI investment to an established measure of tax effectiveness, with tests for heterogeneity (firm complexity, tax function status) and mediation (information quality, capital management).
- Identification/causality: Results are presented as associations consistent with the information-processing mechanism; the abstract does not claim a randomized or natural-experiment identification strategy.
- Robustness: The paper reports concentrated effects across firm types and mechanism tests that are consistent with reduced processing costs driving the results.
Implications for AI Economics
- Conceptual: Provides empirical support for the view that AI reduces managerial information-processing constraints, extending benefits beyond operational prediction (demand, supply) to strategic financial outcomes (tax planning).
- Firm strategy: AI human-capital investments can yield tax-related efficiency gains, especially in complex firms or where tax is strategically integrated into decision-making.
- Organizational design: The status and integration of the tax function matter—organizational arrangements that elevate tax expertise amplify AI’s benefits.
- Public finance & policy: Widespread AI adoption may alter corporate effective tax outcomes, with potential implications for tax revenue forecasting and regulation.
- Future research directions: Causal identification of AI effects on tax outcomes; cross-country comparisons; disaggregating types of AI investments; interaction with automation, governance, and external tax enforcement.
Assessment
Paper Typequasi_experimental
Evidence Strengthmedium — Multiple robustness checks, a long-window changes design, and an IV are used, and results are consistent across specifications, which supports a plausible causal interpretation; however, the study remains archival and observational, the AI measure is an imperfect proxy (captures internal AI-skilled human capital but misses third-party/outsourced AI), the IV and identification assumptions may be contestable, and residual confounding or reverse causality cannot be fully ruled out.
Methods Rigormedium — The paper uses a novel, large-scale résumé-based AI intensity measure, appropriate control variables, entropy balancing, mediation/path analysis, long-window changes, and an IV—showing methodical thoroughness; nevertheless, key challenges remain (measurement error in the AI proxy, potential weak/excludability concerns for the instrument, limits of DEA-based tax-effectiveness measures), which constrain causal claims.
SampleFirm-year panel of U.S.-domiciled public (largely nontechnology) firms from 2010 to 2018; AI intensity measured as the share of employees with AI-related skills from the Cognism résumé dataset (per Babina et al. 2024, covering ~64% of the U.S. workforce in 2018 and 120 million employee-firm-years); tax effectiveness measured using Schwab et al. (2022) DEA-based scores; additional firm-level financial and governance controls used.
Themeshuman_ai_collab productivity
IdentificationObservational firm-year analysis (U.S. public firms, 2010–2018) exploiting cross-firm and over-time variation in a résumé-based firm-level AI-skilled employee intensity (Babina et al. 2024); controls for firm fundamentals; entropy balancing for continuous treatment; weighting and sample restrictions to address measurement precision; long-window changes (difference) specifications to remove time-invariant firm heterogeneity; and an instrumental-variables strategy that instruments changes in firm AI investment with pre-2010 connections to universities with strong AI faculty/research.
GeneralizabilityRestricted to U.S. public firms (largely non-technology firms) over 2010–2018; may not apply to private firms or firms in other countries., AI proxy measures internal AI-skilled human capital and may undercount AI delivered via third-party vendors, consultants, or advisors., Findings pertain to tax planning effectiveness (an accounting/managerial outcome) and may not generalize to other firm-level outcomes (e.g., wages, sectoral productivity) without further evidence., The 2010–2018 window predates the recent rapid diffusion of large language models (post-2018), so results may not map onto newer AI technologies or deployment modes., DEA-based tax-effectiveness measure reflects a specific conceptualization of tax planning and may not capture all facets of tax strategy or regulatory/legal differences across jurisdictions.
Claims (8)
| Claim | Direction | Confidence | Outcome | Details |
|---|---|---|---|---|
| Investment in AI-related human capital is positively associated with tax effectiveness. Fiscal And Macroeconomic | positive | medium | tax effectiveness (effective tax planning) |
0.29
|
| The positive association between AI investment and tax effectiveness is concentrated among more complex firms. Fiscal And Macroeconomic | positive | medium | tax effectiveness (effective tax planning) |
0.29
|
| The positive association between AI investment and tax effectiveness is concentrated among firms where the tax function holds a higher status. Regulatory Compliance | positive | medium | tax effectiveness (effective tax planning) |
0.29
|
| AI improves tax effectiveness by enhancing internal information quality. Regulatory Compliance | positive | medium | internal information quality (mediator) and tax effectiveness |
0.29
|
| AI improves tax effectiveness by enhancing internal capital management. Regulatory Compliance | positive | medium | internal capital management (mediator) and tax effectiveness |
0.29
|
| Information processing constraints hinder managers' ability to effectively integrate tax planning and core business strategies (i.e., processing constraints hinder effective tax planning). Regulatory Compliance | negative | medium | effective tax planning / tax effectiveness |
0.29
|
| AI tools (e.g., machine learning) can mitigate managers' information processing constraints and thereby help integrate tax planning with core business strategy (i.e., AI mitigates constraints to improve effective tax planning). Regulatory Compliance | positive | medium | effective tax planning / tax effectiveness |
0.29
|
| The study uses a recently developed firm-year measure of investment in AI-related human capital, applied to a broad sample of U.S. nontechnology firms between 2010 and 2018. Other | null_result | high | investment in AI-related human capital (independent variable / measure) |
0.48
|