AI strengthens Islamic banks' capacity to innovate largely through indirect channels, but whether those gains translate into ethical, SDG-aligned outcomes depends on middle managers and Sharia governance; the empirical literature is fragmented and concentrated in Indonesia and the UAE.

Main Finding

A systematic review (PRISMA 2020) of 25 studies (2019–2025) concludes that artificial intelligence capability (AIC) functions as a higher‑order dynamic capability mainly via indirect pathways (innovation, ambidexterity, process/business‑model change). The translation of AIC into dynamic capability (DC) outcomes is strongly shaped by managerial microfoundations—especially middle managers—and by Sharia governance (SG), which can act not only as a constraint but also as an enabling/moderating institutional framework. However, the literature is fragmented: micro‑level managerial processes and explicit links to Sustainable Development Goals (SDGs)—notably responsible finance and financial inclusion—are under‑studied, particularly in emerging Islamic finance markets (e.g., Indonesia, UAE).

Key Points

• AIC as dynamic capability:
  • AIC comprises data infrastructure, AI skills, data‑driven culture, and ethical routines.
  • Empirical evidence generally shows AIC → (via mediators) → innovation/DC, rather than a direct AIC → performance path.
• Role of middle managers:
  • Middle managers serve as microfoundations (sensing, seizing, transforming): translating strategy, brokering resources, acting as “AI product owners” and boundary spanners.
  • Their role is critical but rarely modeled or tested empirically in Islamic banking AIC–DC studies.
• Sharia governance (SG):
  • Traditionally compliance‑centric (SSBs, AAOIFI), SG is increasingly framed as a strategic, normative moderator that can enable ethical AI and align innovation with Sharia principles (no riba, gharar, maysir).
  • SG’s moderating effects vary across institutional contexts (e.g., UAE vs Indonesia).
• Regional & thematic gaps:
  • Most empirical work comes from emerging markets (Indonesia, GCC); institutional heterogeneity matters.
  • SDG alignment (responsible finance, inclusion) is insufficiently integrated into AIC–DC research.
• Methodological patterns:
  • Many empirical studies use cross‑sectional PLS‑SEM; conceptual and ANN/machine‑learning models also appear.
  • Overall sample for the review: 1,391 records screened → 25 included; language limited to English; timeframe 2019–2025.

Data & Methods

• Review protocol: PRISMA 2020 systematic literature review; PICOC used to scope population/intervention/comparison/outcome/context.
• Databases searched: Scopus, Emerald Insight, Taylor & Francis, ScienceDirect (plus others implied); Boolean search example: ("Islamic finance" OR "Sharia compliance") AND ("dynamic capabilities" OR "strategic capabilities") AND ("Islamic bank" OR "Sharia bank").
• Screening: 1,391 records retrieved, filtered to 25 peer‑reviewed studies (2019–2025) plus selected foundational pre‑2019 works (e.g., Hamza 2013).
• Study types in corpus: quantitative (frequent use of PLS‑SEM), conceptual/theoretical, scoping reviews, and AI/ANN technical applications used to derive indices (e.g., risk indices).
• Analytical frame used by reviewed works: predominately Dynamic Capabilities View (DCV), Resource/TOE integrations, governance and ethical frameworks; SG frequently treated as compliance antecedent but sometimes modeled as a moderator or input.

Implications for AI Economics

Practical and research implications relevant to AI economics fall into several categories:

• Measurement and causal inference
  • Incorporate institutional moderators (SG) and managerial microfoundations into causal models estimating returns to AI investments. Ignoring these will bias estimates of AIC’s productivity effects.
  • Move beyond cross‑sectional PLS‑SEM: use panel data, instrumental variables, difference‑in‑differences, or natural experiments to identify causal impacts of AIC on innovation, performance, and inclusion.
• Heterogeneity & externalities
  • Recognize that AI investment returns vary with institutional context (SG strictness/structure, regulatory regimes) and managerial capacity. Comparative cross‑country models should explicitly model these interactions.
  • Model distributional outcomes (financial inclusion, access to services) and potential negative externalities (algorithmic opacity, exclusion risks) to evaluate SDG alignment.
• Microfoundations & organizational frictions
  • Endogenize middle managers’ roles in production‑function/technology adoption models: treat managerial capability as an input that mediates AI productivity.
  • Estimate cost of coordination, compliance, and governance when deploying AI in regulated/ethical contexts (Sharia compliance imposes specific constraints and transaction costs).
• Policy and market structure
  • Regulators and standard‑setters (including SG bodies) affect equilibrium adoption and innovation incentives. AI economics should model how governance (standards, certification, supervisory boards) shifts firm‑level investment decisions and market outcomes.
  • Consider public‑good aspects of algorithmic transparency and trust (e.g., subsidized compliance/verification for Sharia‑compatible AI could increase socially desirable adoption).
• Research directions & recommended methods for economists
  • Empirical: firm‑level panel studies across Islamic vs conventional banks; regression discontinuity or DiD around policy/guidance changes in SG; matched comparisons of banks with different managerial structures.
  • Structural and computational: agent‑based models or equilibrium models capturing heterogeneous firms, regulatory regimes, and diffusion of AI‑enabled products.
  • Welfare and SDG evaluation: cost‑benefit and distributional analyses linking AI deployment to poverty reduction, inclusion, and consumer protection metrics.
  • Interdisciplinary work: combine economics, accounting, and information‑systems measures (e.g., AI spending, governance scores, managerial capability indices).
• Operational & investment takeaways
  • For private investors and banks: expected returns to AI depend on governance alignment and managerial capability—investment without concurrent governance and middle‑manager capability building risks low conversion of AI into social or financial value.
  • For policymakers and SG bodies: proactive SG design (from compliance to enabling governance) can unlock AI benefits for inclusion and responsible finance, but needs clear standards and monitoring metrics.

Limitations of the reviewed literature to note for economists: small number of focused empirical studies in Islamic banking, limited longitudinal evidence, and scarce explicit SDG outcome measurement. Designing studies that quantify how SG and managerial microfoundations change the marginal productivity of AI is a high‑value research agenda for AI economics.