Trust and apparent accuracy determine whether AI helps finance teams: where both are high, firms make faster, more data-driven strategic pivots and improve risk management; but misplaced trust, cognitive biases and low AI literacy can create systemic risks and blunt productivity gains.

Main Finding

Perceived trust in AI tools and the apparent accuracy of AI-generated insights are key drivers of finance professionals’ willingness to use AI and their confidence in AI-assisted decisions. When trust and accuracy are high, human–AI collaboration improves organizational agility—allowing faster, data-driven strategic pivots and better risk management—while risks such as cognitive bias, overreliance on algorithmic suggestions, and AI illiteracy limit full realization of benefits.

Key Points

• Focus: human–AI collaboration in financial strategy evaluated along three dimensions — trust, accuracy, and organizational agility.
• Positive effects:
  • Higher perceived accuracy of AI outputs increases decision confidence and perceived utility for forecasting and risk management.
  • Greater trust in AI correlates with greater willingness to adopt AI tools and incorporate AI recommendations into decisions.
  • Effective human–AI collaboration supports faster response to market changes and more agile strategic adjustments.
• Cautions and barriers:
  • Cognitive biases and inappropriate trust (both overtrust and distrust) distort decision outcomes.
  • Excessive reliance on algorithmic suggestions can erode human judgment and create systemic risks.
  • AI illiteracy (lack of understanding of AI capabilities/limits) impedes adoption and appropriate use.
• Normative emphasis: advocates for open, responsible, human-centered AI design and organizational practices to maximize strategic benefits while mitigating harms.

Data & Methods

• Research design: mixed-methods study using structured questionnaires, semi-structured interviews, and performance data analysis.
• Sample: finance professionals across multinational financial institutions (paper summarizes results at the institutional and perceptual levels).
• Measured constructs: perceptions of AI trust, perceived accuracy of AI outputs, willingness to use AI tools, confidence in decision-making, and organizational agility indicators; supplemented with operational performance/risk metrics where available.
• Analysis approach: quantitative analysis to estimate relationships between trust/accuracy and adoption/confidence/agility, augmented by qualitative interview evidence to explain mechanisms and barriers.
• Limitations (noted or implied): potential measurement reliance on perceptions rather than purely objective performance, sample heterogeneity across institutions, and likely correlational rather than strictly causal identification.

Implications for AI Economics

• Adoption and productivity:
  • Perceptions (trust & perceived accuracy) are central frictions in AI adoption within finance; efforts that raise perceived and demonstrable accuracy (e.g., explainability, transparent validation) will increase uptake and productivity gains.
  • Human–AI collaboration is more likely to augment rather than replace skilled finance workers, implying reallocation of tasks toward higher-value judgment and oversight.
• Labor and skills:
  • Demand will increase for AI-literate finance professionals (analytics, model oversight, interpretability skills), shifting returns toward complementary human capital.
  • Risk of deskilling or overreliance suggests a role for continuous training and certification to preserve human judgment.
• Organizational design and capital allocation:
  • Firms that successfully integrate trustworthy, accurate AI can achieve faster strategic pivots — implying potential competitive advantages and higher returns to organizational capital that embeds AI capabilities.
  • Investment cases should account for governance, training, and monitoring costs alongside model development.
• Policy and regulation:
  • Findings support policy emphasis on transparency, auditability, and standards for model validation in finance to reduce systemic risks from misplaced trust or opaque algorithms.
  • Regulation should balance innovation incentives with safeguards against algorithmic overreach and market-stability risks.
• Research directions:
  • Quantify the welfare and productivity gains from improved trust/accuracy (causal estimates of output, risk-adjusted returns).
  • Study heterogeneity across tasks (routine vs. judgmental), institution size, and regulatory regimes.
  • Explore dynamic effects: how trust evolves with experience, feedback loops between model errors and organizational learning, and long-term labor-market impacts.

Practical takeaways: prioritize explainability and validated accuracy, invest in AI literacy and governance, design AI to augment human judgment, and measure both perceptual and objective performance to guide adoption and policy.