SAS experts map AI's next phase in banking

The era of AI experimentation in banking is ending. By 2026, artificial intelligence will no longer sit at the margins as pilots or proofs of concept. Instead, it will be embedded deeply into decision-making, operations and customer interactions, bringing new efficiencies, new revenue opportunities and new forms of risk, according to SAS, a provider of advanced analytics, data management and AI software to financial institutions.

That is the central message behind a new set of 13 expert predictions recently released by SAS, which outline the breakthroughs, blind spots and breaking points banks are likely to face over the next two years.

SAS draws on decades of work across risk, fraud, compliance and decisioning to frame what it calls a "banker's dozen" of industry-defining shifts. Collectively, the predictions suggest that banking is moving from model-driven intelligence to proof-driven intelligence, where trust, transparency and governance become as important as speed and automation.

Alex Kwiatkowski, director of global financial services at SAS, argues that trust itself is being redefined. As AI systems make decisions in milliseconds—approving credit, flagging fraud or resolving customer inquiries—banks can no longer rely on assurances alone.

"In 2026, trust will morph from a promise to a performance metric," Kwiatkowski said. He predicts that institutions will be expected to demonstrate verifiable transparency across every AI-driven prediction and interaction, shifting from "don't worry, trust us" to "here's the evidence."

Agentic AI moves from pilots to production

One of the most consequential shifts SAS anticipates is the move from AI-assisted workflows to agentic AI operating at scale. Rather than simply supporting employees, semi-autonomous agents will increasingly manage customer requests, orchestrate internal workflows and make governed decisions across the enterprise.

Diana Rothfuss, global solutions strategy director for risk, fraud and compliance at SAS, said 2026 will mark a turning point as banks industrialize AI rather than experiment with it. She pointed to forecasts showing financial services firms spending tens of billions of dollars on AI in the coming years, with decisioning and operational use cases driving the largest returns.

The challenge, however, is that agentic AI introduces new risks alongside new efficiencies. Adam Neiberg, global banking senior marketing manager at SAS, warned that banks will inherit the fallout of autonomous AI agents acting on behalf of customers, particularly as agentic commerce expands.

Disputes could surge when AI agents make purchases customers never explicitly approved. Fraud teams will also face new threats as criminals learn to hijack or mimic legitimate agents. As a result, banks may need to authenticate not just people, but also the AI agents acting in their name.

Data integrity becomes a strategic battleground

Another theme running through the predictions is data contamination. As banks increasingly rely on generative AI and synthetic data to accelerate model development, SAS experts warn that subtle errors and biases could quietly seep into core systems.

Ian Holmes, director and global lead for enterprise fraud solutions at SAS, predicted that banks will respond by creating "data purity vaults," which are controlled environments designed to protect golden source data from contamination. Unlike traditional data quality issues, AI-driven errors can spread rapidly and appear realistic, making them difficult to detect once embedded in decision pipelines.

At the same time, generative AI is expected to unlock long-overlooked value from unstructured data. Terisa Roberts, global director for risk modeling and decisioning at SAS, noted that more than 80 percent of enterprise data exists in formats such as text and images. Knowledge agents powered by large language models and retrieval-augmented generation could allow banks to extract insight from that data at scale, reshaping both strategic planning and risk management.

Fraud, compliance and financial crime pressures intensify

Fraud and financial crime loom large in SAS's outlook, particularly as AI lowers the cost and scale of attacks. Romance scams, for example, are expected to surge as fraudsters automate emotional manipulation using large language models.

Stu Bradley, senior vice president of risk, fraud and compliance solutions at SAS, said banks may increasingly be forced into the role of "emotional firewalls," combining behavioral analytics with AI-driven monitoring to detect exploitation patterns before monetary losses occur.

At the same time, investment pressures are likely to trigger a shakeup in financial crime technology. Beth Herron, Americas lead for banking compliance solutions at SAS, argued that many legacy AML and fraud platforms struggle to incorporate advanced AI. As those limitations become more visible, banks are expected to accelerate adoption of cloud-native, AI-first solutions that offer real-time, explainable analytics.

Markets, payments and new revenue models

Beyond risk and compliance, SAS experts foresee significant changes in capital markets and payments. AI-powered quantitative credit strategies are expected to accelerate price discovery in corporate bond markets, pushing investors beyond ratings-centric workflows. Yet Robert Jarrow, an adviser and industry consultant affiliated with SAS, predicts that while bubble-aware risk models should become standard practice, most institutions will fail to adopt them quickly enough.

In payments, Ahmed Drissi, AML lead for Asia-Pacific at SAS, expects regulated stablecoins to move from theory into real banking pilots. With clearer regulatory frameworks emerging in the United States and Europe, banks may begin testing stablecoins for cross-border settlement and treasury use cases. These are early steps toward modernizing international payments infrastructure.

Retail banks are also expected to scale commerce media and financial media network strategies. Cornelia Reitinger, head of advertising business development at SAS, predicts that banks operationalizing these models could see meaningful gains in non-interest income as advertisers tap into verified financial data.

Climate, quantum and what banks aren't ready for

The final set of predictions looks further ahead. Climate risk stress testing is expected to become more integrated into everyday risk management, driven by regulatory scrutiny and real-world climate events.

And while quantum computing remains nascent, Julie Muckleroy, global banking strategist at SAS, said 2026 could mark the first production deployments of hybrid quantum-classical models, particularly in risk and fraud optimization.

Together, the predictions underscore a common message: AI's impact on banking is accelerating faster than many institutions' ability to govern it. Drawing on its long history in analytics and decision science, SAS frames 2026 as a year that will separate banks that have industrialized intelligence from those still struggling to manage its consequences.

For banking leaders, the question is no longer whether AI will transform the industry but whether they are ready for what comes next.

Notice to readers: These are archived articles. Contact information, links and other details may be out of date. We regret any inconvenience.