Auditors who report using voice‑data analytics are associated with more accurate going‑concern opinions among listed Nigerian manufacturers, the study finds; the evidence is observational and based on self‑reported technology use and firm reports.

Main Finding

Voice data analytics (VDA) is positively associated with improved audit opinion accuracy for listed manufacturing firms in Nigeria. The study concludes that audio-derived evidence (e.g., recorded calls, management representations) analyzed with deep-learning–based VDA can help auditors reduce opinion misclassification and strengthen going-concern assessments.

Key Points

• The paper responds to concerns that AI might threaten the auditing profession by demonstrating a complementary role for AI-based voice analytics in improving audit quality.
• Theoretical framing uses sociomateriality: audit outcomes (social) and analytics technology (material) are inseparable and co-constitutive.
• Voice Data Analytics (VDA) leverages speech-to-text, tonal/pitch and phonological-discrepancy analysis, and embodied conversational agents to extract deception signals, sentiment, and other cues from recorded oral communications (conference calls, customer service calls, management interviews).
• Audit opinion accuracy (AOA) is operationalized in the going-concern context: auditors’ going-concern opinions are compared with financial-distress measures (Altman Z-Score). Type I/II misclassifications correspond to opinion inaccuracies; a binary indicator records whether opinion accuracy is confirmed.
• Empirical result: VDA usage (measured via firm-level primary data from external auditors) is positively associated with accurate audit opinions (secondary data from audited annual reports).
• Policy/recommendation highlights: firms should integrate voice-generating platforms with internal audit systems (leveraging Nigeria’s Voice Data Retention Policy) and the Financial Reporting Council of Nigeria should institutionalize VDA in audit practice.

Data & Methods

• Research design: mixed methods.
  • Primary data: collected at firm level from external auditors to measure the presence/usage of voice data analytics.
  • Secondary data: extracted from audited annual reports of listed manufacturing firms to determine audit opinions and to compute financial-distress measures.
• Audit opinion accuracy measure: compared auditor going-concern opinions against Altman’s Z-Score to classify firm distress; constructed a dummy variable (1 = opinion accuracy confirmed; 0 = otherwise).
• Analytical approach: statistical association testing between VDA (from primary survey/firm reports) and audit opinion accuracy (from secondary data). (Specifics on sample size, time period, questionnaire items, and econometric model are not included in the provided excerpt.)

Implications for AI Economics

• Complementarity and labor effects: VDA exemplifies AI augmenting professional judgment rather than substituting it — auditors using VDA can be more accurate, implying upskilling demand and a shift in task composition toward higher-value judgment and oversight activities.
• Information economics & capital markets: more accurate going-concern opinions reduce information asymmetry and investor uncertainty, potentially lowering firms’ cost of capital and improving capital allocation efficiency.
• Valuation of data assets: voice recordings and conversational data become valuable economic inputs for audits and corporate governance, increasing firms’ incentives to collect, retain, and monetize high-quality audio data (subject to privacy and compliance constraints).
• Regulatory and governance externalities: institutionalizing VDA (as the authors recommend) would require investment in data-retention infrastructure, standardized data formats, model validation procedures, and accountability frameworks to prevent misuse, bias, and privacy violations — all factors with economic costs and benefits that regulators must weigh.
• Market structure and competition: widespread adoption of VDA may raise audit quality floor, affecting competitive dynamics among audit firms (firms with superior analytics capabilities could capture market share or command premium fees).
• Risks and caveats affecting economic outcomes:
  • Data privacy, consent, and legal restrictions could constrain usable voice data and raise compliance costs.
  • Model bias, adversarial manipulation of recorded interactions, or overreliance on automated signals could create new types of audit failures.
  • Upfront investments in AI capability and data pipelines may favor larger firms, potentially increasing concentration in analytics-enabled auditing services.
• Research and policy opportunities: cost–benefit studies on VDA adoption, macro-level assessments of VDA-driven changes in audit market quality and capital costs, and regulatory impact analyses of mandated voice-data retention/analytics standards.