← Papers

Generative AI boosts management accounting by enriching data and sharpening analysis rather than substituting judgment; its benefits appear only when tied into standardized routines, integrated data infrastructures and strong cross‑functional governance, though evidence comes from public disclosures rather than internal causal proof.

Research on the Impact of Generative AI on the Quality of Management Accounting Decisions: Evidence from Manufacturing Firms

Xintian Shen · Fetched April 25, 2026 · Pacific International Journal

semantic_scholar descriptive low evidence 7/10 relevance DOI Source PDF

Based on documentary case analysis of three large Chinese manufacturers, GenAI appears to improve management accounting decision quality not by replacing managerial judgment but by enriching information, augmenting analysis, and delivering benefits only when embedded in data-integrated routines and cross-functional governance.

Generative artificial intelligence (GenAI) is moving from experimentation to organizational infrastructure in accounting and finance. Yet firm-level evidence on how it shapes management accounting decision quality in manufacturing remains limited, especially in contexts where financial, operational, engineering, and service information are tightly intertwined. This paper revises the question in a more analytically cautious way: rather than claiming direct observation of internal decision processes, it examines what can reasonably be inferred from public corporate disclosures about the ways GenAI-related capabilities may influence management accounting decision quality. The study adopts an interpretive multiple-case design and analyzes three major Chinese manufacturing firms - Midea Group, Haier Smart Home, and Dongfang Electric - using official annual and semi-annual reports, corporate disclosures, and recent AI-and-accounting literature. The findings suggest that GenAI does not improve management accounting decision quality primarily by replacing managerial judgment. Instead, its potential effects appear to operate through three linked mechanisms: information enrichment, analytical augmentation, and organizational embedding. First, GenAI-related capabilities broaden the informational basis of management accounting by making operational, service, quality, and ecosystem data more usable in planning and control. Second, they enhance analysis by translating complex data into more interpretable, scenario-sensitive, and action-oriented outputs. Third, these benefits are likely to materialize only when AI capabilities are embedded in standardized routines, integrated data infrastructures, and cross-functional governance arrangements. At the same time, the documentary evidence also indicates important boundary conditions, including data maturity, process integration, governance discipline, and the degree of functional trust between finance and operating units. The paper contributes by sharpening the concept of management accounting decision quality, distinguishing GenAI from broader digital transformation, and offering a cautious process model grounded in documentary case evidence from leading Chinese manufacturers. Because the evidence is drawn primarily from external disclosures rather than direct internal observation, the claims should be read as interpretive analytical inferences rather than as definitive causal proof.

Summary

Main Finding

Public documentary evidence from three major Chinese manufacturers (Midea Group, Haier Smart Home, Dongfang Electric) suggests that generative AI (GenAI) is unlikely to improve management-accounting decision quality by simply replacing managerial judgment. Instead, GenAI appears to have plausible, conditional effects through three linked mechanisms: (1) information enrichment (broader, decision-relevant visibility of operational/service/quality data), (2) analytical augmentation (more interpretable, scenario-sensitive, action‑oriented outputs), and (3) organizational embedding (integration into routines, shared‑finance platforms, and cross‑functional governance). These benefits materialize only when data maturity, process integration, governance discipline, and functional trust are present; otherwise richer outputs can create noise, opacity, or misplaced confidence. The claims are interpretive inferences from public disclosures rather than direct causal evidence.

Key Points

Conceptual contribution
- Defines “management accounting decision quality” as accounting-supported judgments that are timely, decision-relevant, sufficiently accurate for the task, internally coherent across functions, intelligible to decision makers, and usable for coordinated action.
- Distinguishes GenAI (content-generating layers: text summaries, explanations, scenario narratives) from broader AI/analytics and general digital transformation.
Mechanisms by which GenAI may affect decision quality
Information enrichment: synthesizes structured and unstructured operational, quality, logistics, procurement, service, and downstream user data into a broader informational perimeter for planning and control.
Analytical augmentation: translates complex data into explanations, exception summaries, and scenario narratives that reduce interpretive friction between data and managerial judgment.
Organizational embedding: conditions benefits—requires integration in shared-finance platforms, ERP/business-finance architectures, standardized routines, and human review/governance.
Boundary conditions and risks
- Data maturity, process integration, governance discipline, and inter‑functional trust are necessary enablers.
- Risks include opacity, overconfidence in persuasive outputs, weak accountability, and potential governance failures when outputs are uncritically used.
Empirical nuance
- Midea: clearest public signals of GenAI use (lighthouse factory with 72 digital/AI solutions, including GenAI for workforce capability and supply‑chain responsiveness) — implies information enrichment and operational impacts with accounting relevance (lead times, working capital).
- Haier: explicit integration (Vertex AI into SmartHQ) and AI roadmap — indicates scenario/insight generation relevant to planning, product mix, lifecycle and service-cost implications.
- Dongfang Electric: stronger evidence of intelligent shared‑finance, scenario pilots, and business‑finance integration but less explicit GenAI labeling — useful for illustrating organizational embedding rather than direct GenAI effects.
Limitations
- Documentary case method relies on public disclosures (2024–2025) and triangulation with literature; cannot observe internal decision dynamics or establish causal estimates.
- Inferences are cautious and only made when supported by documentary signals across cases.

Data & Methods

Research design: interpretive, theory-building multiple-case study (three major Chinese manufacturing firms) chosen by theoretical sampling to maximize variation in GenAI visibility and digital-finance maturity.
Evidence base: official annual and semi-annual reports, lighthouse-factory disclosures, AI roadmap communications, and company materials from 2024–2025; supplemented by relevant academic and professional literature.
Coding and inference procedure:
- Deductive provisional coding frame aligned to the three mechanisms (information enrichment, analytical augmentation, organizational embedding) and boundary conditions.
- Within-case coding preserved case integrity; each coded claim linked to documentary evidence and assigned an evidentiary strength based on explicitness of AI references.
- Triangulation: cross‑document consistency checks within firms, alignment with conceptual literature, and coherence between AI initiatives and operational/financial domains.
- Cross-case inference rule: patterns treated as suggestive only when supported in at least two cases; language intentionally cautious (e.g., “suggests”, “indicates”).
Analytical stance: interpretive reconstruction from public texts — organizational claims treated as evidentiary signals requiring conservative inference (not direct observation).

Implications for AI Economics

Conditional productivity effects
- GenAI’s contribution to economic value within firms likely depends on complementarities with organizational capital (data infrastructures, ERP/shared‑finance, governance routines). Investments in GenAI alone may yield low returns without these complements.
Reallocation and task composition
- Rather than substituting managerial judgment, GenAI augments interpretive tasks — implying demand shifts within accounting: fewer routine-processing hours, more emphasis on validation, cross-functional coordination, and judgment tasks that integrate AI outputs.
Measurement and evaluation challenges
- Standard productivity or ROI metrics may understate GenAI value because benefits accrue through improved coordination, faster interpretation, better forecasts, and reduced cognitive costs that are hard to attribute to single systems.
- Evaluation should track decision-quality indicators (forecast responsiveness, cross‑functional linkage, speed/clarity of variance explanation, procurement/inventory coordination), not only automation or headcount changes.
Investment and diffusion patterns
- Firms with higher data maturity and integrated business-finance platforms (shared finance, ERP coupling) are better positioned to capture GenAI benefits; expect heterogeneity in adoption returns across firms and sectors.
Labor-market and skill implications
- Accounting and finance roles will accentuate skills in AI oversight, scenario interpretation, governance, and cross‑functional communication; upskilling and new governance roles (AI validation, model-risk control) become economically relevant.
Governance and policy
- Opacity and overconfidence risks suggest need for internal governance (model audit trails, human-in-loop validation) and possibly external reporting standards for AI use in managerial decision-making.
- Regulators and standard-setters should consider guidance on disclosure of AI usage in managerial processes to improve market assessment of firm capabilities and risks.
Research directions for AI economics
- Quantitative follow-ups: measure how GenAI-related investments intersect with data/integrative capital to affect measurable outcomes (forecast accuracy, inventory turns, cycle times).
- Causal identification: exploit firm-level experiments or phased rollouts to estimate impact on decision-quality metrics and downstream financial performance.
- Macro-level diffusion: study how complementarities drive heterogeneity in productivity gains across industries and countries.

Limitations to keep in mind: the paper’s conclusions are inferential and documentary; they frame plausible mechanisms and boundary conditions rather than definitive causal estimates.

Assessment

Paper Typedescriptive Evidence Strengthlow — Findings are based on interpretive analysis of public disclosures from three firms rather than on direct measurement or causal identification; inferences are plausible but not directly supported by internal process data, counterfactuals, or exogenous variation. Methods Rigormedium — The study uses a systematic multiple-case, document-based approach with triangulation against literature and explicit analytical framing, which is appropriate for exploratory theory-building; however, small-N selection, reliance on external disclosures, and lack of primary interviews/observations limit internal validity and causal inference. SampleThree major Chinese manufacturing firms (Midea Group, Haier Smart Home, Dongfang Electric), analyzed via official annual and semi-annual reports, corporate disclosures, and recent AI-and-accounting literature; no internal audit logs, employee interviews, or transaction-level data were used. Themesorg_design human_ai_collab Generalizabilitylimited_to_large_chinese_manufacturers, based_on_public_disclosures_not_internal_data, small_sample_of_three_firms, sector_specific_to manufacturing, potentially influenced by Chinese regulatory/corporate governance context, findings are time-bound to recent disclosures and may evolve with technology adoption

Claims (8)

Claim	Direction	Confidence	Outcome	Details
The study adopts an interpretive multiple-case design and analyzes three major Chinese manufacturing firms - Midea Group, Haier Smart Home, and Dongfang Electric - using official annual and semi-annual reports, corporate disclosures, and recent AI-and-accounting literature. Other	null_result	high	study design / sample description	n=3 0.3
Because the evidence is drawn primarily from external disclosures rather than direct internal observation, the claims should be read as interpretive analytical inferences rather than as definitive causal proof. Other	null_result	high	strength/causal status of inferences	n=3 0.3
GenAI does not improve management accounting decision quality primarily by replacing managerial judgment. Decision Quality	negative	high	management accounting decision quality	n=3 0.18
GenAI-related capabilities broaden the informational basis of management accounting by making operational, service, quality, and ecosystem data more usable in planning and control (information enrichment). Decision Quality	positive	high	management accounting decision quality (via information breadth/usability)	n=3 0.18
GenAI-related capabilities enhance analysis by translating complex data into more interpretable, scenario-sensitive, and action-oriented outputs (analytical augmentation). Decision Quality	positive	high	management accounting decision quality (via improved analysis/interpretability)	n=3 0.18
GenAI-related benefits are likely to materialize only when AI capabilities are embedded in standardized routines, integrated data infrastructures, and cross-functional governance arrangements (organizational embedding). Decision Quality	positive	high	realization of GenAI benefits for management accounting decision quality	n=3 0.18
Important boundary conditions include data maturity, process integration, governance discipline, and the degree of functional trust between finance and operating units. Decision Quality	negative	high	constraints on GenAI impact on management accounting decision quality	n=3 0.18
The paper contributes by sharpening the concept of management accounting decision quality, distinguishing GenAI from broader digital transformation, and offering a cautious process model grounded in documentary case evidence from leading Chinese manufacturers. Other	null_result	high	conceptual clarity / theoretical contribution	n=3 0.18