SAS has set out 13 predictions for how artificial intelligence will affect banking in 2026, with a focus on autonomous agents, synthetic data risks, and closer scrutiny of trust and governance.

The forecasts describe banks moving from pilot programmes to wider deployments of systems that take decisions and manage workflows. They also highlight areas where banks may face new disputes, higher fraud pressure and more complex compliance demands.

Alex Kwiatkowski, Director of Global Financial Services at SAS, said banks will face rising expectations around explainability and proof in automated decision-making. “AI has made financial institutions faster, smarter and infinitely more confident – sometimes too confident. From credit scoring to fraud detection to customer service, we’ve trained intelligent systems to decide in milliseconds. But has the industry risked losing sight of its most human principle along the way? Trust must be earned, not assumed.

“This year, trust will morph from a promise to a performance metric as banks shift from model-driven to proof-driven intelligence. Demanding verifiable transparency across every prediction, decision and interaction will become the new standard of intelligence. In other words, don’t trust this prediction – until you can prove it,” said Kwiatkowski.

Agentic systems

One of the central themes in SAS’s outlook is the move of “agentic AI” into daily banking operations. Diana Rothfuss, Global Solutions Strategy Director for Risk, Fraud & Compliance Solutions at SAS, said semiautonomous systems will start taking on more work across organisations.

“2026 will mark the dawn of agentic AI in banking as semiautonomous systems begin to take on meaningful work across the enterprise. The future of intelligent banking will be shaped by AI-driven agents that manage customer requests, orchestrate workflows and make governed, explainable decisions at scale. This shift will fundamentally change how banks design operations and measure the value of AI.

“According to IDC, financial services firms will spend more than £50 billion ($67 billion) on AI by 2028. Production deployments tied to decisioning and operations are poised to see the biggest growth. The industry has matured beyond the proof-of-concept, and the banks that succeed will be those that industrialise their AI to turn pilots into profit and governance into competitive advantage,” said Rothfuss.

SAS also expects knock-on effects from the use of agents outside banks. Adam Neiberg, Global Banking Senior Marketing Manager at SAS, described a rise in disputes where purchases get triggered by autonomous systems rather than customers.

“From call centres to the C-suite, financial institutions will be forced to face the impacts of the rapidly expanding agentic commerce economy. Banks will see a surge in disputes triggered by autonomous AI agents making purchases the customer never approved, and fraud teams will face new risks as criminals learn to hijack or mimic legitimate agents.

“As agentic eCommerce grows, banks must learn to authenticate not only people but also the AI agents acting in their name, adding a new layer of complexity to an already tough financial crimes fight. Frameworks such as agentic tokens, behavioural signatures and dynamic risk scoring represent the first wave of controls banks will need to safeguard their human customers and their bottom line,” said Neiberg.

Data integrity

SAS’s predictions also focus on data quality and the risks attached to synthetic data. Ian Holmes, Director and Global Lead for Enterprise Fraud Solutions at SAS, said banks may struggle to detect contamination in core repositories as generative systems produce increasingly realistic content.

“Banks will confront a new kind of data integrity crisis as generative AI and synthetic data seep into core repositories in ways that are difficult to detect. Unlike the isolated data quality issues of the past, GenAI can introduce errors at scale – and with a level of realism that makes contaminated data extremely hard to surface.

“As financial institutions experiment with synthetic data to accelerate model development, many will unknowingly introduce subtle biases and distortions into credit, fraud and risk decisioning pipelines. To protect critical workflows, banks will begin securing their golden source data in controlled digital vaults and impose stricter governance on how GenAI tools can interact with core data sets,” said Holmes.

Terisa Roberts, Global Director for Risk Modeling, Decisioning and Governance at SAS, said banks will use generative techniques more widely for analysis of unstructured data such as text and images.

“In 2026, generative AI will become for unstructured data what traditional statistics has long been for structured data, giving banks the ability to extract meaning and insight at scale. More than 80% of enterprise data is in unstructured formats like text and images, and this volume is growing 50% to 60% each year.

“Banks will begin adopting knowledge agents powered by large language models and retrieval of augmented generation technology to turn previously underused unstructured data into quick, actionable answers. They will use these new insights to accelerate strategic business decisioning and transform risk management into a more proactive, intelligence-driven discipline,” said Roberts.

Fraud pressure

SAS expects scams to evolve alongside wider access to automation. Stu Bradley, Senior Vice President of Risk, Fraud and Compliance Solutions at SAS, said romance scams will increase as fraudsters scale emotional manipulation with machine-generated interactions.

“Your chances of dating a model have never been higher – a large language model, that is. While AI-powered romance scams already exist, they will surge to record levels as fraudsters weaponise emotional deception at scale. What once required weeks or months of hands-on engagement can now be automated and accelerated with minimal effort.

“As machine-assisted manipulation advances, the line between genuine connection and synthetic seduction will blur further, testing not only fraud defences but human intuition itself. Financial institutions will be pressed to act as emotional firewalls for their customers, combining behavioural analytics and AI-driven monitoring to detect exploitation patterns before the monetary damage is done,” said Bradley.

Beth Herron, Americas Lead for Banking Compliance Solutions at SAS, predicted disruption among financial crime technology providers as firms attempt to modernise rules-based platforms.

“The anti-financial crime compliance market will undergo a major shakeup in the year ahead as vendors struggle to embed advanced AI into their offerings. Recent divestitures underscore the scale of reinvestment required to modernise dated, rules-based platforms, leaving many banks with tools that can’t keep pace with evolving fraud and money-laundering threats. As the difficulties of bolting AI onto legacy platforms come to light, financial crime technologies built natively on AI platforms will shine brightest.

“In 2026, financial institutions will accelerate adoption of cloud-native, AI-driven AML and fraud solutions that can surface complex patterns. Our latest survey of ACAMS members shows that most institutions already see AI as essential for AML modernisation, and banks that migrate toward explainable, real-time analytics will gain significant compliance and risk advantages,” said Herron.

Markets and payments

In capital markets, SAS expects changes in how credit risk gets analysed and priced. Stas Melnikov, Head of Quantitative Research and Risk Data Solutions at SAS, said quantitative credit strategies will increase, with models absorbing alternative data and forward-looking indicators more quickly.

“The growth of quantitative credit strategies will accelerate price discovery in corporate bond markets, catalysed by AI-assisted models that rapidly incorporate new information, alternative data and forward-looking credit indicators. Active fixed income teams will move beyond ratings-centric workflows and adopt flexible, ML-driven modeling and decisioning infrastructure that translate diverse signals into trading decisions.

“Strong data governance and rigorous model risk management will be the necessary ingredients for this process and technology evolution. Additionally, innovation in credit rating risk modeling will help investors reduce losses and capture opportunities,” said Melnikov.

Robert Jarrow, Advisor and Industry Consultant, Quantitative Research and Risk Data Solutions at SAS, said some institutions will move toward “bubble-aware” models in pricing and stress testing, but adoption will not become universal.

“In 2026, leading banks and asset managers will start embedding bubble-aware models into pricing, ALM and stress testing. These models will explicitly break down the market price of assets into their fundamental drivers, while also examining risk premiums and transient bubble components. Bubble-aware models help firms recognise factors that cause asset prices to rise sharply and unsustainably. And while these models should become part of all banks’ standard practice in 2026, I fear – and predict – they will not,” said Jarrow.

On payments, Ahmed Drissi, Anti-Money Laundering (AML) Lead for Asia-Pacific at SAS, said stablecoins will enter regulated pilots for cross-border settlement and treasury use cases.

“Imagine a US-EU corporate corridor that settles in minutes rather than days. We aren’t there yet, but the year ahead will see regulated stablecoins move into real banking pilots. With clearer frameworks in the US and EU, banks will begin testing stablecoins for cross-border settlement and treasury for their inherent benefits: faster fund movement, lower costs and greater transparency. Some banks will also explore tokenised deposits or partnerships with licensed issuers to move money on digital rails with stronger auditability and compliance. These early pilots signal the first meaningful step toward modernising international payments,” said Drissi.

New revenue

SAS also expects retail banks to expand efforts in commerce and advertising. Cornelia Reitinger, Head of Advertising Business Development at SAS, said more banks will formalise “media strategy” plans and report results.

“By the end of 2026, every major retail bank will have a media strategy, whether they call it that or not. Banks that quietly tested the model over the past 12 to 18 months will begin reporting measurable revenue gains as advertisers and brands recognise the power of verified financial data. Institutions that operationalise financial media networks could realistically see a 20% to 30% uplift in noninterest income within two years,” said Reitinger.

Climate modelling

Peter Plochan, EMEA Principal Risk Management Advisor at SAS, said banks will expand climate risk stress testing as regulators and stakeholders demand more detail and stronger governance.

“As the impact of storms, wildfires and droughts on bank portfolios intensifies worldwide, banks face mounting pressure from customers, regulators and shareholders to improve their climate risk management efforts. 2025 saw the first-ever fine for a bank’s noncompliance with climate risk regulations. Therefore, I foresee banks stepping up their climate risk stress testing to close gaps in modeling, governance and infrastructure. Its closer integration with banks’ core business-as-usual risk management frameworks will be essential to effectively respond to increasing pressures.

“AI-driven automation and integration of stress testing processes will be critical enablers not only for addressing the requirements of climate risk but also other emerging scenario analysis use cases, such as the European Central Bank’s recently announced geopolitical risk reverse stress test,” said Plochan.

Quantum outlook

Julie Muckleroy, Global Banking Strategist at SAS, said hybrid quantum-classical computing will shift from pilots to production in some areas of banking, particularly for risk and fraud modelling.

“This will mark the year we see the first impacts that hint at how quantum AI will reshape the banking landscape through the end of the decade. Hybrid quantum-classical computing will move from pilots to production, delivering breakthroughs in risk and fraud – and it will expand the frontier of how banks optimise, simulate and make decisions, especially in areas where classical models degrade. Banks building early experience will see transformative gains in accuracy, agility and performance that deliver an outsized edge over the competition,” said Muckleroy.