Artificial intelligence (AI) and machine learning (ML) are being increasingly adopted across financial services and telecommunications (telcos) to improve decision-making and reduce risks.

A new study conducted by Forrester Consulting for credit reporting company Experian looks at the state of ML adoption in these sectors, revealing that approximately two-thirds of respondents are already using ML in live decisioning, while about one-quarter are experimenting with it.

Strong performance gains

The study, which surveyed nearly 1,200 senior decision makers responsible for developing and implementing AI and ML in credit risk in financial services and telcos across 11 countries in Europe, the Middle East and Africa (EMEA) and Asia-Pacific (APAC), found that most are already seeing substantial benefits from using ML.

More than half of the organizations that have adopted ML reported seeing a significant or large improvement in their acceptance rates for all categories of lending since adoption. Most notably, the biggest improvement is for small and medium-sized enterprise (SME) loans, with 88% seeing an improvement in acceptance rates for this segment.

ML is also improving bad debt performance. 65% of respondents reported a significant or large improvement in their bad debt rates since adoption, with credit cards seeing the strongest uplift (86%).

These gains are attributed to the greater predictive accuracy of these ML models, as well as their ability to analyze non-traditional data sources and better identify vulnerable customers. 70% of respondents cited operational efficiency and cost saving, as well as improved risk prediction, as the biggest benefits of ML.

Based on Experian’s experience, ML models typically perform 5-20% better than the usual statistical models used in credit scoring. These systems also make the process faster and more efficient since they can evaluate alternative data variables automatically, test new data sources quickly, and reduce dependency on manual analytics, freeing up resources for strategic tasks.

ML is also helping improve financial access, with 70% of those who are using advanced ML agreeing that the improved accuracy of these models means that they can widen access to credit for consumers who would otherwise be denied credit with traditional scorecards.

Limited resources as top challenge to ML adoption

Despite the clear benefits, several challenges are still hindering ML adoption. The most common barrier, cited by 55% of respondents, is the time and resources required to implement ML.

Adding to this issue is the scarcity of AI talent. LinkedIn’s Jobs on the Rise 2025 report ranks AI and ML roles among the top fastest-growing positions in the US. Reuters estimates that in 2024, IT spending exceeded US$5.1 trillion, with AI spending pushing past US$550 billion. Yet, there was a hiring gap approaching 50% of all AI positions needed.

Following time and resource constraints, the second biggest challenge, cited by 53% of respondents, is turning raw data into credit attributes that make up the individual features of each model.

A lack of in-house expertise was also highlighted by half of respondents, reflecting broader concerns about access to specialized skilled labour. 66% of respondents expect the talent shortage to be a significant or large challenge over the next few years.

Regulatory issues are also a major concern. 75% of respondents agreed that regulatory compliance is limiting their organization’s ability to innovate within credit decisioning, while 70% are holding back on implementing more automated ML credit decisions due to concerns about regulatory backlash. A similar percentage (66%) agreed that their national regulators lack a clear, consistent understanding of how ML models function in practice.

AI, ML implementation among top business priorities

Moving forward, 73% of respondents believe that organizations that adopt ML in credit underwriting will gain a significant long-term competitive advantage.

Their top business priorities for the next one to three years include implementing AI and ML into risk decisions (76%), moving onto a unified data, analytics, decisioning, and fraud platform (75%), and improving the time to decision for customers (74%).

Finally, 79% of respondents believe that in five years’ time, the vast majority of credit decisions will be fully automated.

In the financial services industry, AI and ML are being deployed in an array of use cases that extend beyond credit scoring. HSBC, for example, uses AI to fight financial crime, leveraging a system it developed in collaboration with Google. The platform helps it check about 900 million transactions for signs of financial crime each month, across 40 million customer accounts to identity suspicious activity.

JPMorgan, meanwhile, spends about US$2 billion a year on developing AI technology, and reports that the resulting cost savings now offset its spending. In 2024, the bank launched its proprietary generative AI platform, LLM Suite, which enrolled 200,000 employees within eight months and support them with tasks such as idea generation, and content drafting.

Worldwide spending on AI is forecast to total nearly US$1.5 trillion this year, up 50% year-over-year (YoY), according to Gartner, a business and technology insights company. By 2026, that amount is projected to rise by about 37% to more than US$2 trillion, led in large part by AI being more broadly integrated into products such as smartphones and PCs, as well as infrastructure.

Featured image: Edited by Fintech News Switzerland, based on image by freepik via Freepik