On the heels of the Biden Administration’s AI executive order, and amid ongoing debates from lawmakers and business leaders alike on AI regulation, a new Moody’s Analytics study finds that while there is overwhelming desire for AI adoption in compliance, significant barriers hinder progress, including poor internal data quality, a lack of clarity around regulation and a specialist knowledge gap.
The study, which included a survey of more than 550 senior compliance and risk management professionals from 67 countries, including 101 responses from North and South America, assessed perspectives on and uses of AI. It shows that the top three areas where AI is being applied are data analysis and interpretation (63%), risk management (53%) and fraud detection (51%).
Keith Berry, general manager of KYC solutions at Moody’s Analytics, said, “Compliance professionals are convinced that AI will be transformative for their industry, but obstacles remain that could hinder risk management and compliance functions from capitalizing on its potential. The benefits of AI are currently viewed in easy-to-measure quantitative terms. Process efficiencies are a good start to AI adoption, but they are only scratching the surface of the technology’s capabilities.”
Key findings:
- Internal data quality: Only 14% of those surveyed rated their own data as high quality. Resolving data issues is critical to reducing LLM hallucinations and improving the accuracy of AI outputs. Early adopters of different types of AI are more likely to rate their data quality as high (36%), compared to 9% who are not considering the use of AI. There is a clear data maturity gap, with 75% who are not contemplating the adoption of AI considering their data quality to be poor.
- Regulation: The study found 79% of professionals feel that new legislation to regulate the use of AI in compliance is important, while 66% seek greater clarity around any existing AI-related regulations in risk and compliance. Interestingly, North and South American respondents were less concerned about regulatory guidance, compared to global peers. Just over 70% of Americas respondents rated it as important or very important, compared to 90% in APAC and 80% in EMEA.
- LLM conservatism: Despite the rapid growth of LLMs, caution remains in risk and compliance. Only 28% take a positive stance on these models, 25% are actively discouraging or prohibiting their use, and 46% have yet to adopt an LLM policy. Just 41% associate LLM terminology with risk and compliance.
- Use-case understanding: Only a quarter (26%) rated their overall understanding of AI’s relevance to risk management and compliance as high. Compliance professionals are most likely to identify improved efficiency in processes (72%), increased speed of data processing and analysis (72%) and cost savings due to automation or improved decision-making (66%). Fewer currently recognize the potential for more advanced, transformative benefits, such as improved accuracy of results and predictions (51%) and the reduction of false positives (49%).