Firms only reap AI returns when they both broaden and deepen deployment: Spanish large firms show higher performance when multiple AI applications are widely diffused across the organisation. Skilled AI staff, digital infrastructure and a data-driven culture determine whether AI delivers business value.
Beyond AI Adoption: An Empirical Study on the Antecedents and Performance Outcomes of AI Deployment in Organizations
openalex
correlational
low evidence
7/10 relevance
Summary only
summary available; pdf_status=paywall
Source
PDF
Among 770 large Spanish firms, firm performance is positively associated with the interaction of AI deployment depth and breadth—firms that both implement diverse AI technologies and diffuse them widely perform better, supported by AI-skilled human capital, digital infrastructure, and a data-driven culture.
Despite substantial investments in Artificial Intelligence (AI), evidence on its business value remains mixed. We argue that this heterogeneity stems not from whether firms adopt AI, but from how they deploy it. Drawing on Resource-Based Theory and organizational search theory, we conceptualize AI capability as two-dimensional: AI deployment depth — technological variety of AI implementations — and AI deployment breadth — organizational scope of AI diffusion. Three resources shape these dimensions differentially: AI-skilled human capital drives depth, digital infrastructure drives breadth, and data-driven culture supports both. Using archival microdata from 770 large Spanish firms, staged OLS models show that firm performance is positively associated with the interaction between depth and breadth, consistent with a complementarity logic. We contribute by reconceptualizing AI capability as a deployment configuration and offering a configurational explanation for the AI productivity paradox.
Summary
Main Finding
Firm performance improves when AI deployment is both deep (varied technological implementations) and broad (diffused widely across organizational units). Depth and breadth are complementary: their interaction—rather than adoption alone—explains positive performance effects and helps resolve the AI productivity paradox.
Key Points
- AI capability is two-dimensional:
- Depth: technological variety and sophistication of AI implementations.
- Breadth: organizational scope and diffusion of AI across units/processes.
- Three organizational resources shape these dimensions differently:
- AI-skilled human capital primarily drives depth.
- Digital infrastructure primarily drives breadth.
- Data-driven culture supports both depth and breadth.
- Complementarity logic: depth × breadth interaction is positively associated with firm performance; neither dimension alone is sufficient.
- Reconceptualization: moving from a binary “AI adoption” view to a configurational view of AI deployment explains heterogeneous productivity outcomes.
Data & Methods
- Data: archival microdata from 770 large Spanish firms.
- Empirical approach: staged OLS models testing relationships among resources, AI deployment dimensions (depth, breadth), their interaction, and firm performance.
- Main test: interaction term (depth × breadth) positively associated with firm performance, consistent with complementarity.
- Identification/caveats: results are correlational from observational archival data; further causal or experimental validation is needed.
Implications for AI Economics
- Measurement: Empirical studies should move beyond adoption dummies and separately measure deployment depth and breadth, plus their interaction.
- Investment strategy: Firms should coordinate investments across AI-skilled labor, digital infrastructure, and culture—isolated investments are less likely to yield value.
- Policy and management: Policies that build workforce AI skills, interoperable digital infrastructure, and data-driven norms can increase the returns to AI by enabling complementary deployment configurations.
- Explaining mixed evidence: Heterogeneous business value of AI arises from differences in deployment configuration, not merely adoption prevalence.
- Research agenda: Test causal links, explore dynamics over time, generalize to other countries/firm sizes, and examine optimal sequencing and complementarities among resources.
Assessment
Paper Typecorrelational
Evidence Strengthlow — The paper reports cross-sectional associations from staged OLS models without quasi-experimental variation, instrumental variables, or longitudinal identification; therefore observed relationships may reflect reverse causality or omitted variable bias despite theoretical framing and controls.
Methods Rigormedium — Uses firm-level archival microdata on a reasonably large sample (770 firms) and interaction models to test a theoretically grounded configurational hypothesis; however, reliance on OLS and likely self-reported/aggregate measures, lack of robustness checks described here (e.g., IV, panel methods, or placebo tests), and potential measurement and selection issues limit methodological rigor.
SampleArchival microdata on 770 large Spanish firms (firm-level measures of performance, AI deployment depth and breadth, and firm resources such as AI-skilled human capital, digital infrastructure, and data-driven culture); cross-sectional/staged OLS analysis (timeframe not specified).
Themesproductivity org_design
GeneralizabilitySample restricted to large firms in Spain — limited applicability to SMEs or other countries, Cross-sectional design limits inference to other time periods or dynamic effects, Industry composition may bias results if not representative across sectors, Measures of AI deployment and culture may be context-specific or based on survey/self-report, limiting comparability
Claims (8)
| Claim | Direction | Outcome | Confidence & Evidence | Details |
|---|---|---|---|---|
| Firm performance is positively associated with the interaction between AI deployment depth and AI deployment breadth (consistent with a complementarity logic). Firm Productivity | positive | firm performance |
Reading fidelity
high
Study strength
medium
|
n=770
|
| AI capability can be conceptualized as two-dimensional: AI deployment depth (technological variety of AI implementations) and AI deployment breadth (organizational scope of AI diffusion). Adoption Rate | null_result | conceptualization: AI deployment depth and breadth |
Reading fidelity
high
Study strength
speculative
|
not reported
|
| AI-skilled human capital drives AI deployment depth. Adoption Rate | positive | AI deployment depth |
Reading fidelity
high
Study strength
medium
|
n=770
|
| Digital infrastructure drives AI deployment breadth. Adoption Rate | positive | AI deployment breadth |
Reading fidelity
high
Study strength
medium
|
n=770
|
| A data-driven culture supports both AI deployment depth and breadth. Adoption Rate | positive | AI deployment depth and breadth |
Reading fidelity
high
Study strength
medium
|
n=770
|
| Heterogeneity in the business value of AI stems not from whether firms adopt AI, but from how they deploy it (i.e., the configuration of depth and breadth). Firm Productivity | positive | heterogeneity in AI business value (variation in firm performance associated with deployment configuration) |
Reading fidelity
medium
Study strength
medium
|
n=770
|
| The complementarity between AI deployment depth and breadth offers a configurational explanation for the AI productivity paradox. Organizational Efficiency | null_result | explanation for AI productivity paradox (interpretive/theoretical outcome) |
Reading fidelity
high
Study strength
speculative
|
n=770
|
| The empirical analysis used archival microdata from 770 large Spanish firms and employed staged OLS regression models. Other | null_result | methodological description (data and analytical approach) |
Reading fidelity
high
Study strength
high
|
n=770
|