AI-Enabled FinTech and Investment Decision-Making: An Integrative Review of Capabilities, Behavioural Adoption, and Governance (International Journal of Economic Practices and Theories)

Summary

• Global AI investment was projected to approach USD 200 billion by 2025, accounting for 2.5% to 4% of U.S. GDP.p.1
• The top four U.S. robo-advisers managed over USD 76 billion in 2015, growing at a year-on-year rate of 116%.p.5
• The global robo-advisory market was valued at USD 6.61 billion in 2023, with a projected compound annual growth rate of 33.6% from 2025 to 2030.p.5
• Retail investors using robo-advisors increased their exposure to non-correlated asset classes by 32% relative to self-managed portfolios.p.5
• Using a robo-advisor was associated with a 15.8 percentage-point decrease in the likelihood of consulting a human financial adviser.p.5
• Survey findings show 42% of investors hesitate to adopt AI investment solutions due to transparency concerns, while 49% view AI models as black boxes.p.8
• Over 75% of surveyed respondents stated that algorithmic transparency and regulatory approval would increase their willingness to trust AI advisers.p.8

Key insights

1. Field evidence links robo-advisory access to higher equity participation, improved diversification, and better risk-adjusted returns.p.4
2. Tree ensembles and neural networks improve out-of-sample return descriptions over linear models across nearly 30,000 U.S. stocks.p.4
3. Statistical predictive accuracy in machine learning models does not automatically guarantee portfolio gains once market frictions are considered.p.6
4. General-purpose large language models have demonstrated the ability to perform financial statement analysis at levels consistent with human-like judgement.p.4
5. Algorithm aversion remains a substantial obstacle to adoption, with 53% of global survey respondents unlikely to trust AI-based robo-adviser algorithms.p.8
6. Model explainability is critical for user adoption, with 82% of surveyed participants rating explainable recommendations as very or extremely important.p.8
7. Systemic risk can emerge when widespread reliance on shared algorithms and datasets drives market convergence and herding during stress.p.10
8. Evidence favours a human-AI hybrid architecture where AI expands information processing while humans maintain context interpretation and oversight.p.12