Global banks increase AI investment amid governance gaps

By Divya Shah

Key findings

• Nearly all surveyed banks reported using AI in credit and risk decision-making, with generative AI adoption reaching 82%.
• Only a small proportion of banks had fully operational AI governance frameworks, highlighting a gap between adoption and oversight.
• Most banks expected AI to take on greater responsibility for autonomous credit and risk decisions by 2031

Why this matters: Banks are moving beyond basic AI applications towards using generative and agentic AI in credit decisions, lending, treasury and risk management. This creates opportunities for FinTech providers offering AI-powered banking solutions.

Artificial intelligence (AI) adoption has increased rapidly across the banking sector, with financial institutions using the technology to support credit, treasury and enterprise risk decisions. However, gaps in governance, data quality and regulatory oversight continue to limit its wider use in autonomous decision-making.

According to a new study by FT Longitude, SAS and Microsoft, 82% of banks have adopted generative AI (genAI), while 41% are using agentic AI in their decision-making processes. When traditional predictive AI technologies, including machine learning (ML), are also considered, 299 of the 300 banks surveyed reported using AI for credit and risk decisions.

The research surveyed 300 credit, treasury and risk leaders at banks across 31 countries in June and July 2026. It found that AI adoption had moved beyond initial experimentation, with 59% of respondents reporting that the technology was already embedded in their decision-making processes. This included 44% with partial adoption and 15% with widespread implementation.

AI adoption was also becoming more common across different banking functions. In credit risk management, an average of 63% of banks had partially or widely integrated AI into their main processes. In asset-liability management and risk steering, the figure stood at 55%.

AI was increasingly influencing decisions throughout the lending process, including credit origination, customer management, collections and recoveries.

More than half of the surveyed banks reported that AI was already shaping or making decisions across these credit-related activities. Its use in capital management and portfolio composition remained lower, although more than four in ten institutions were using the technology in these areas.

The findings highlighted how banks were moving beyond using AI for basic automation towards applying it in more complex financial decisions, including credit assessments, risk management and capital allocation.

Despite the rapid adoption of AI, the research identified significant gaps in governance and risk management.

Only 18% of banks had fully operational AI governance processes in place, while 26% lacked a dedicated AI model risk policy. Just 19% had real-time monitoring systems to identify changes or performance issues in AI models.

The findings also showed that 42% of banks did not have clearly defined mechanisms, or ‘kill switches’, to deactivate AI models when they crossed predefined risk limits. Meanwhile, only 36% continuously retrained or updated their AI models.

These gaps highlighted the challenge financial institutions faced in balancing faster AI adoption with appropriate oversight, accountability and risk controls.

The study also identified internal resistance, data quality and regulatory concerns as major barriers to wider AI adoption.

Around 71% of respondents said internal resistance had led their banks to restrict AI use in automated decision-making. Concerns about data quality were also significant, with 79% identifying a lack of trust in data as a primary or contributing factor in limiting autonomous AI applications.

In addition, 77% cited regulatory and compliance risks as a main or contributing reason for restricting the use of autonomous AI.

As a result, banks remained cautious about allowing AI systems to make decisions independently. Only 10% of respondents said their institutions currently allowed AI to make autonomous credit or risk decisions at scale. Overall, 70% of banks placed moderate or significant restrictions on autonomous AI decision-making.

However, expectations for the future were more positive. Around 76% of respondents expected AI to make autonomous credit and risk decisions at scale by 2031, while 25% anticipated this development within the next two years.

Banks also expected AI to deliver greater value across lending operations. Credit-decision optimisation was identified by 75% of respondents as one of the most important expected benefits over the next two years. A further 58% anticipated improvements through the automation and optimisation of decision strategies.

Alongside the global growth in banking AI adoption, Singapore continued to strengthen its position as a financial technology and innovation hub.

According to the Financial Services State of the Nation Survey 2026, conducted in November 2025, 64% of financial institutions in Singapore were actively deploying AI across key business functions. Another 35% were piloting or researching AI, while none reported having no plans to adopt the technology.

To support wider adoption of emerging financial technologies, the Monetary Authority of Singapore (MAS) launched the Future of Finance Institute (FFI).

The initiative aimed to work with financial institutions and industry participants to move financial innovation beyond experimentation and towards broader implementation.

Its planned resources included a repository of validated use cases, deployment guides, solution providers, implementation toolkits and industry sandboxes. These facilities would allow financial institutions to test new technologies in controlled environments before wider deployment.

FFI’s initial focus was on AI in financial services. It aimed to bring institutions together to share practical implementation experience, develop good practices for agentic AI and strengthen risk management around AI applications.

The institute also planned to expand into other areas of financial innovation, including tokenisation.

As banks increasingly integrate AI into lending, treasury and risk management, the findings highlighted the importance of building strong governance frameworks alongside technological capabilities. For financial institutions, the next stage of AI adoption will depend not only on deploying more advanced systems but also on ensuring that these technologies remain transparent, reliable and accountable.