Banks expect to hand decisions to AI – even if they can’t explain them

Banks are racing toward a future where AI no longer merely informs key credit, treasury and risk decisions - it makes them autonomously. Yet AI governance is not keeping pace with adoption, according to new AI risk research by FT Longitude, data and AI leader SAS, and its strategic partner Microsoft.

Even as more than 3 in 4 (76%) banks expect autonomous risk and credit decisions at scale by 2031, the study found:

• Only 18% have fully operational, embedded AI governance processes today.
• A mere 19% have real-time AI model drift monitoring.
• 42% lack defined "kill switches" to deactivate AI models that cross preset boundaries.
• Just 36% continuously retrain or update AI models.

The report, Why AI at Scale Needs Governance at Speed, is based on a global survey of 300 banking leaders across 31 countries. Together, the findings underscore a widening gap between banks' AI ambitions and the guardrails and controls needed to deploy AI agents and other autonomous AI models and tools safely and responsibly.

"These risks are not theoretical. When AI systems get decisions wrong, there are real consequences for a bank's customers and its community, as well as for its financial performance and reputation," said Anselmo Marmonti, Vice President of Risk, Fraud and Compliance Solutions at SAS. "Rapid AI adoption alone is not enough. Sustained leadership depends on focused business processes and strong governance combined with trustworthy AI at scale. Together these deliver the business outcomes, resilience, compliance and accountability that modern banking requires."

And while bankers surveyed are keen to quickly embrace AI, regulators are also urging caution. "Ambition should match its capacity," Marilin Pikaro, Director of Innovation, Conduct and Consumers at the European Banking Authority, told FT Longitude. She warned that even compelling AI use cases should not advance faster than an institution's ability to manage the attendant risks.

Adoption leaps, governance lags

82% of banking leaders say their organizations are already using generative AI for decision making, while 41% report using agentic AI. Combined with traditional, predictive AI (such as machine learning), AI adoption has reached near universal levels, with 299 of the 300 banks surveyed using AI in their credit and risk decision making.

Despite this momentum, governance lags significantly. As noted, 82% of banks say they do not have fully operational and embedded governance processes for AI models. And staggeringly, more than one-quarter (26%) of surveyed institutions lack a dedicated AI model risk policy.

Bankers do recognize the importance of good governance. Aurelio Maccario, Head of Group Risk Management at UniCredit, a multinational banking group based in Italy, noted in the report: "Governance means that you have done your homework, that you have created the boundaries, that you have set the rules, and then within those things you are ready to run."

Banks that succeed will be those that advance technology innovation and governance together, enabling faster decisions without sacrificing transparency, oversight or regulatory confidence.

Risky black boxes?

Yet, the survey also revealed an alarming tension between performance and transparency. Nearly two-thirds (63%) of banking leaders said they would be willing to use a black-box model - one with little or no decision-making transparency -- if it offered better performance, returns or liquidity than an explainable alternative.

And many banks today appear to be succumbing to this temptation and thus exposing themselves to potential compliance issues. A full 92% of survey respondents reported that at least one large language model within their institutions already was developed or operates outside formal governance frameworks.

"While inefficient governance might drive some banks to look for workarounds and shortcuts, there is real peril here. The quick performance boost from a black-box AI model might seem appealing in the short-term, but its lack of visibility, repeatability and explainability spells trouble from a regulatory and risk perspective," said Tyler Pichach, Global Head of Payments, AI Strategy and Go-to-Market at Microsoft. "As AI becomes more autonomous and takes on more consequential tasks, governance has to evolve with it. The opportunity is to put the right guardrails, monitoring and human oversight in place so banks can move faster with AI while maintaining the trust and accountability financial services demands."

Autonomous AI decisions are getting closer

While AI is influencing day-to-day operations across banks, few institutions currently allow it to make decisions independently. Only 10% of respondents say AI makes autonomous credit or risk decisions at scale today. However, future expectations are dramatically different: while 76% expect AI to be making autonomous credit and risk decisions at scale within five years, one quarter (25%) expect this shift within two.

The survey found AI already plays a substantial role in banking operations:

• 59% of banks have partially or widely embedded AI in their decision-making processes.
• AI influences departmental decisions at least monthly in 53% of organizations.
• AI is partially or widely embedded in 63% of credit risk processes and 55% of asset and liability management (ALM) and risk steering processes.

Canadian digital bank Tangerine, which began using AI to feed into lending decisions in 2023, has seen this progression while balancing people and technology. In the report, Nishant Taneja, Tangerine's Head of Credit Risk, said: "In fields such as retail credit portfolios, more than 90% are AI-based decisions with very little human-in-the-loop methods. When it comes to high-risk portfolios, we have intentionally kept the human in the loop, until we have more confidence in the technology."

Learn more at RiskMinds

To learn more about the FT Longitude research and SAS' extensive risk management offerings, visit the SAS booth at RiskMinds International 2026, the world's top risk management event for financial institutions, November 16-19 in London. SAS is a major sponsor of the conference, its executives will moderate a panel of chief risk officers and lead a technical session on credit risk, and it will provide networking opportunities for customers and partners.

About the research

The study was conducted by FT Longitude in June and July 2026 and surveyed 300 banking leaders across 31 countries spanning North America, Europe, Asia-Pacific, the Middle East, Africa and Latin America. Twenty percent of respondents were C-suite executives, and 66% represented banks with more than $50 billion in assets.