This year’s SAS Innovate 2026 was chalk full of amazing content and speakers.  This year's Banking sessions delivered some of the most forward-thinking, real-world demonstrations of AI-powered risk, fraud, and compliance we've seen on the Innovate stage. The sessions spoke for themselves.   Weren’t able to make it to Texas?  No problem.   Didn’t sign up for the free digital pass (expires June 30)?  No worries.   New this year you can find almost all the sessions from SAS Innovate 2026 on YouTube.    Check out some of the highlighted sessions below and presentations in the curated SAS Innovate Banking playlist.    The Sessions That Made an Impact   Accelerating Intelligent Decisioning Development with Python, Containers and Automation Organizations need to deliver analytics faster without compromising reliability. Traditional workflow development is often slow to progress and can increase the risk of errors. This session demonstrates how containerization, Git, and automated pipelines deployed to containerized environments transform the development lifecycle. Attendees will learn how local environments using Python, containers, and database management tools can closely replicate production—enabling rapid testing, consistent data handling, and fewer deployment issues. The session shows how this approach supports faster iteration, stronger version control, and reduced downtime, helping teams deliver SAS solutions with greater speed and confidence in today’s fast-paced environment.   Agentic AI in Financial Services: From Automation to Intelligent Decisions Join SAS and AWS alongside leaders from JP Morgan Chase and Allianz to explore how Agentic AI is transforming decision-making across banking and insurance. This session will highlight how intelligent automation, adaptive learning, and insight generation are driving operational efficiency, risk management, and customer-centric innovation. Hear real-world perspectives on how Financial Services organizations are leveraging Agentic AI to build trust, scale impact, and stay ahead of disruption.     A Tale of Two Languages: SAS and Python in One Analytics World Presented by Hyundai Capital America, this session shows how seamless interoperability between SAS and Python enabled by SAS Viya and SAS Viya Workbench has become a cornerstone of its analytics modernization journey. Attendees will see how integrating both languages enables cross‑skill collaboration, allowing SAS users to extend capabilities with Python libraries while Python users leverage SAS’s enterprise‑grade data management in a unified environment.   Closing the Loop: Meeting SMEs Needs and Bank Business Goals through Smarter AI‑Oriented Decisions Discover how banks shift from static offers to adaptive, capital‑aware decisioning. This demo showcases a closed‑loop approach that uses AI to connect campaign activity, next best offer decisions, and multi‑channel customer responses into a continuously learning system aligned with profitability and regulatory capital goals. The integrated approach delivers attractive offers to customers, with an optimized RAROC‑based pricing by bringing together customer insights, risk assessment, collateral allocation, expected loss, and capital requirements. The result: sharper pricing, smarter offers, and campaigns that consistently meet return‑on‑capital targets, driving sustainable growth for (SMEs) small, medium enterprises.   Finding the Signal in the Noise: AI‑Driven AML Alerts in SAS Rule‑based AML systems generate excessive alerts that slow investigations and obscure true financial crime risk. This demo shows how AI‑driven AML alert prioritization enhances rules with supervised machine learning and explainable AI to score and rank alerts by risk—reducing false positives, focusing investigations, and improving overall program efficiency.   How Consumer Portfolio Services Transformed Decision-Making with SAS Viya At Consumer Portfolio Services, we embarked on a mission to modernize our legacy decisioning platform, a project that became our gateway to the AI world. Migrating from an AS400 rules system to SAS Viya 3.5 Intelligent Decisioning wasn’t just a technical upgrade; it was like converting a manual car into an automatic, enabling our company to move faster, smarter, and more efficiently. This session shares how we engineered a real-time, high-volume decision engine that unites data, analytics, and automation — paving the road for predictive modeling, machine learning, and GenAI-driven decisioning.   Hidden in Plain Sight: How Traffickers Exploit the Financial System via Transactional Data Human traffickers move billions through legitimate financial channels, often blending illicit proceeds into normal commerce. This session exposes how trafficking operations exploit transactional patterns, payment networks, and emerging financial technologies to conceal criminal proceeds. Using real-world case studies, we’ll explore the analytical signatures of trafficking from structuring and funnel accounts to digital payment layering and demonstrate how  analytics can uncover hidden risk behaviors. Attendees will learn how to identify high-risk typologies, apply data-driven detection models, and strengthen multi-disciplinary collaboration between AML, fraud, and human trafficking response teams. With Human trafficking, Sextortion and Pig Butchering Scams all tying together it is crucial to take action.    Lighting the Future of AML with Agentic AI In AML programs, the distance between a compliance officer’s intuition and a deployed scenario and model is often measured in months of back-and-forth, manual coding, and governance bottlenecks. This weakens organizations’ defenses to efficiently and effectively catch suspicious actors. What if you could responsibly develop and optimize your detection strategies at the speed of thought? In this session, join the award-winning Consortix-AURORA hackathon team for a superdemo of their innovative and trustworthy Agentic AI solution for AML, which is designed to seamlessly bridge the gap between domain experts and technical teams, execute detection strategies in a governed way and outpace criminals.   Model Documentation: A Single Source of Truth for Your Risk Models This session introduces SAS Risk Modeling’s new Model Documentation capability, a fully UI‑driven, audit‑ready solution that captures the complete model lifecycle in one collaborative environment. Attendees will see how documenting data, assumptions, KPIs, validation results, and governance steps in a single source of truth improves compliance, accelerates reviews, and reduces manual effort.   Operationalizing AI and Machine Learning with SAS Fraud Decisioning Professional Models This session explores how SAS Fraud Decisioning Professional Models deliver high‑performance risk scores for real‑time fraud detection and how to integrate them directly into decision flows. Attendees will learn, step by step, how to apply model scores and reason codes within business rules to operationalize fraud models, improve detection accuracy and preserve a seamless customer experience.   Pilot to Production: Audit‑Ready Synthetic Data for Finance on SAS Viya In this session, Aldermore and SAS show how to build and govern bank‑grade synthetic data on SAS Viya to overcome limited historical data. Attendees will see how techniques like SMOTE in CAS and SAS Data Maker support privacy‑preserving generation, validated across fidelity, utility, privacy and fairness, with outcomes documented in SAS Model Manager to move from pilot to production.   Real-Time Decisioning, Rapid Request Resolution & Hyper-Personalization at National Bank of Greece National Bank of Greece, with Performance Technologies, implemented an Agentic AI customer service and real‑time analytics program using SAS Viya, SAS Event Stream Processing and Azure OpenAI (GPT‑4 with RAG). The hybrid architecture automated request classification, agent guidance and real‑time micro‑campaigns—demonstrating scalable, governed AI‑to‑action execution across service and marketing.   Reinventing Fraud Rules with Generative AI and SAS Viya Copilot This session shows how SAS Viya Copilot replaces manual, complex fraud rule development by generating SAS code using LLMs with deep awareness of available data, profiles, functions, and macros. Attendees will see how no‑code guidance and proactive tuning recommendations help fight AI‑driven fraud with trusted AI improving prevention rates while reducing customer friction.   Reverse Stress Testing with Agentic AI: Anticipating Risk Before It Strikes This session shows how Agentic AI and SAS Viya enable reverse stress testing by automating complex calculations, simulating extreme edge‑case scenarios, and embedding explainability and trust into stress‑testing pipelines to surface actionable vulnerabilities faster even with limited or incomplete data.   SAS + Intel: Modernizing your AI Workloads for FSI without Breaking the Bank! Financial Services leaders are under pressure to modernize faster, while tightening security, meeting regulatory expectations, and proving ROI on AI and analytics. In this session, SAS and Intel share a practical, executive-level framework for building AI-ready infrastructure for FSI workloads without creating a new wave of technical debt. We’ll explore the strategic tradeoffs banks face– hybrid vs. cloud-first, performance vs. cost, speed vs. governance, and the decisions that most influence time-to-value and long-term resiliency.   Streamlining Model Risk Management for Operational Efficiency While regulatory compliance remains a critical focus across the financial and technology sectors, effective Model Risk Management must translate complex governance requirements into tangible operational efficiencies. This presentation moves beyond regulatory summaries to demonstrate how SAS’ Model Risk Management can reduce time and cost while enhancing the trustworthiness and accountability of enterprise risk activities.   Strengthening AML Programs with Advanced Analytics: Rules Engine and Beyond Presented by Ally Bank, this session examines how advanced analytics and machine learning can strengthen traditional rules‑based AML programs overwhelmed by false positives and growing alert volumes. Through practical demonstrations, attendees will see how enhanced scenario monitoring and data‑driven models improve detection accuracy, adapt to emerging threats, and reduce investigative noise so teams can focus on higher‑risk activity.   Strengthening Operational Resilience with SAS Governance and Compliance Manager As banks and insurers digitize operations, they face an expanding risk landscape driven by financial crime, regulatory pressure, and environmental, geopolitical, and social factors. This session shows how SAS Governance and Compliance Manager on Viya helps organizations strengthen operational resilience, meet rising regulatory demands, and reduce risk exposure through transparent, integrated, AI‑enabled governance with lower total cost of ownership.   The Future of FRAML: One Platform, One View of Financial Crime The Fraud and Anti–Money Laundering (FRAML) landscape is evolving rapidly as fraudsters and financial criminals continuously adapt their tactics faster than financial institutions can respond. At the same time, regulatory expectations continue to intensify, increasing pressure on FIs to detect, prevent, and respond to financial crime more effectively. To keep pace, institutions need a modern FRAML approach that unifies fraud and AML capabilities while eliminating operational and intelligence silos. By deploying FRAML capabilities on a single, cloud-native architecture, financial institutions can manage the entire analytics lifecycle using a common set of tools and data. Leveraging advanced AI and machine learning, a holistic FRAML solution enables organizations to connect insights across fraud and AML and proactively mitigate emerging financial crime risks.   Transforming Compliance: SAS GenAI Copilot for Automated Model Governance See how SAS Copilot automates model governance—analyzing documentation, completing policy assessments, and preserving full auditability with human oversight.   Using AI to Enhance SAS Programming for AML and Data Analysis in Banking Presented by BMO Financial Group, this session explores how GitHub Copilot can modernize and optimize SAS programming workflows for AML and banking analytics using AI‑assisted code generation. Attendees will see practical applications including PROC SQL optimization, documentation automation, prompt engineering, style guide development, macro generation, and SAS‑to‑open‑source code conversion demonstrated using a real AML transaction monitoring dataset.   If this got you excited Save the date for SAS Innovate 2027 in Las Vegas and be part of it from the start.  ... View more

Managing digital marketing campaigns across multiple markets has taught me one consistent lesson: multilingual execution rarely fails because of ambition but because of unmanaged complexity. Teams expand into new regions, add languages incrementally, and suddenly find themselves maintaining dozens of parallel campaign variants with little shared structure.    The risk isn’t just operational overhead. Research consistently shows that customers are more likely to engage and convert when content is delivered in their native language. When localization breaks down, performance erosion follows, often quietly at first, then at scale.      Where multilingual complexity really comes from  Multilingual campaigns become difficult to scale when they are handled as a series of translation tasks rather than as a structured operating model. In practice, that usually shows up through a familiar set of anti-patterns, each of which creates both operational friction and commercial downside.    Asset duplication without a clear source of truth increases maintenance effort and makes errors harder to catch, often leading to inconsistent customer experiences across markets.  Weak separation between global standards and local adaptation creates tension between brand consistency and market relevance, reducing engagement when campaigns feel either too generic or too rigid.  Language logic embedded inconsistently across segments and journeys adds technical debt, slows execution, and can push customers into the wrong language experience, affecting conversion and average order value.  Limited cross-market visibility makes it harder to compare performance across localized versions, weakening insight generation and reducing the ability to optimize journeys over time.  Manual localization workflows delay activation and limit experimentation, making it harder for teams to respond quickly and improve performance at scale.    What I see working in scalable multilingual campaigns    Treat assets as systems, not one‑offs  To avoid asset duplication and the confusion that comes from having no clear source of truth, teams that scale effectively design assets with reuse in mind. Rather than duplicating entire emails or messages per language, they standardize structure and vary only what genuinely needs localization.  A common pattern is to maintain a shared master structure, layout, modules, and brand elements, while isolating language‑specific content. This reduces confusion, accelerates updates, and makes governance far easier as campaign volume grows.    Separate brand consistency from local expression  To prevent the common failure mode where global consistency and local relevance pull in opposite directions, high-performing teams separate brand standards from local expression. Rather than treating the two as a trade-off, they design for both from the outset.  Brand-critical elements, visual identity, tone guardrails, and structural components are standardized. Local teams are then given controlled flexibility within that framework to adapt messaging, references, and emphasis based on market context.  Templates and synchronized components help enforce consistency without constraining localization. The key is clarity: teams need to know which elements are fixed and which are intentionally flexible.    How principles 1 and 2 work together in practice    This shared-template approach helps solve two common problems at once: it reduces asset duplication by using a shared template, while also separating global brand consistency from local language variation.  When creating a campaign in SAS Customer Intelligence 360, such as an email promoting a new hotel, you can then add language-specific modules that are shown only when the relevant display condition is met.    One module contains the English version of the promotional content.  A second module contains the Spanish version.  Both modules sit within the same email, but only the version that matches the customer’s language is displayed.            Make language a first‑class data attribute  To avoid embedding language logic inconsistently across segments and journeys, multilingual execution becomes much simpler when language preference is treated as a core customer attribute rather than an afterthought.  Whether derived from declared preferences or behavioral signals (such as browser language picked up by a CI 360 event), this data enables segmentation logic that remains clean and scalable. When language is embedded upstream, at the audience or segment level, campaign logic downstream becomes far easier to manage and reason about.    Design journeys for visibility, not just delivery  To solve the visibility problem that emerges as campaigns spread across languages and markets, journeys need to be designed for monitoring and comparison, not just delivery. Otherwise, teams quickly lose sight of where performance differs and why.  Journey‑based orchestration helps address this by keeping language variations within a shared structural view. The objective isn’t to force all markets into identical execution, but to maintain a coherent model that allows performance to be monitored and optimized holistically.     In the example below, I use the CI 360 agent to generate an onboarding journey with three language-specific paths: French, English, and Dutch.  The overall onboarding design remains consistent across all three paths. Each one follows the same:    end goal  journey flow  communication cadence over the 90-day onboarding period  channel mix  What changes is the localized execution. By separating each language version into its own path, I can maintain a clear view of the overall journey while also monitoring how each localized path performs once the journey is live.            Use AI to accelerate localization—responsibly  To reduce the delays and bottlenecks created by manual localization workflows, generative AI has materially changed the economics of multilingual content creation. What once took weeks can now be produced in minutes, enabling faster iteration and broader experimentation.  That said, the teams seeing the most value treat AI as an accelerator, not an autopilot. Human validation remains essential, particularly for tone, cultural nuance, and regulatory sensitivity. Used this way, AI becomes a force multiplier rather than a risk.  Once multiple localized variants exist, structured experimentation becomes critical. A/B testing within journeys allows real customer behavior—not internal opinion—to determine which messages resonate most strongly in each market.    In the screenshot below, I use AI in CI 360 to generate five French versions of an original English push notification. This makes it much faster to brainstorm and develop suitable localized variations.            From these AI-generated options, I can select the three strongest French variants and test them within the French path of the onboarding journey using an A/B optimization node. The goal is not to compare languages against one another, but to identify which version performs best for that specific language audience based on real customer behavior during execution. The same optimization approach can then be applied independently within the other language paths.              Where to start this week    For teams early in their multilingual maturity, the goal is not perfection but leverage. A few high-impact starting points:    Standardize asset structure before adding new languages  Ensure language preference is reliably captured and accessible  Review where duplication is occurring and why  Identify one journey where multilingual visibility really matters and redesign it deliberately    Small structural improvements early on prevent exponential complexity later.    A closing perspective  Multilingual execution is no longer a niche capability reserved for global enterprises. As digital reach expands, it’s becoming a baseline expectation. The teams that scale successfully are the ones designing systems that make localization repeatable, visible, and sustainable.  Complexity will always increase as you grow, the difference is whether it’s intentional or accidental.    If this resonates with you, I’d love to hear your perspective in the comments. And if you’ve seen other approaches work well, please share them too.    ... View more

If you read my earlier post on using the SMOTE method in SAS Data Maker, you may recall that I started out with an unbalanced 40,000-observation dataset, and discovered that it generated synthetic data that looked exactly as unbalanced as the data I put in. The fix turned out to be obvious in hindsight: feed SMOTE only the minority class. The lesson was as much about understanding what a method does as it was about using the software correctly.   That experience got me thinking about the philosophy behind different synthetic data generation methods, and I fell into a rabbit-hole of learning more. So come into the rabbit-hole with me! SMOTE and Bayesian networks approaches are available in SAS Data Maker, SAS's no-code/low-code platform for generating privacy-preserving synthetic tabular data. But they are built on fundamentally different premises about what it means to "understand" a dataset well enough to generate new data from it. This post compares those two families of methods, examines their respective strengths and challenges, and considers when you might choose one over the other. It also lays the groundwork for a follow-up post that will go deeper into the two marginal model-based methods available in SAS Data Maker specifically: MST and PrivBayes. ... View more

(2× reproduced)   Offer: SAS Viya (Pay-As-You-Go), plan sas-viya-on-azure, publisher sas-institute-560503, version 26.05.01 Managed app: datafold-sas-viya — subscription 5f7a1c8d-22de-4925-b5ac-491634cad550, RG sas-viya-rg, region Central US Managed resource group: mrg-sas-viya-on-azure-20260612110733 Failing resource: deploymentScripts/datafold-viya-ds-viya-deploy → DeploymentScriptExceededMaxAllowedTime (PT2H). Reproduced on two separate attempts (~09:21–11:21 UTC, 2026-06-12). What we've verified on our side: - All infrastructure provisions successfully (VNet, NSG, AKS, NFS VM, jump VM). - AKS node pools never scale past 1 node each; the NFS share (/viya-share) stays empty for the full 2 hours → the install hangs before deploying SAS workloads. - The deploy ACI container (uxtgshwbglkzo...azscripts) is still Running at 2h+ — stalled, not crashed. - Azure compute quota is confirmed sufficient (Standard DSv4 = 64, EDSv4 = 64 in Central US; never approached). - We cannot access the cluster or the clusterSetup.sh logs due to the managed-app deny assignment.   Request: Please inspect the clusterSetup.sh / orchestration logs and in-cluster pod state for this managed RG to identify what the deployment is waiting on (suspect image-pull/order entitlement or a bootstrap pod not reaching Ready) and tell us what we can do to fix it, so we can redeploy.    ... View more