SAS Enhances Agentic AI Platform with Improved Governance and Industry-Specific Models

SAS extends business rules governance experience to agentic AI. Here's why

Silicon Valley start-ups might be excited about the prospects of letting AI agents run around their infrastructure. However, traditional enterprise buyers will need to see a few more guardrails in place. With that in mind, SAS is extending its Intelligent Decisioning tooling to support trustworthy and agentic AI. The company was an early leader in decision management governance, with Business Rules Manager in 2013. Its more flexible agentic variant helps design, deploy, and scale AI agents with human stewardship, embedded governance, and explainability of decisions. This takes advantage of the work on the Viya AI platform, which has a rich ecosystem of supporting tools for data ingestion, performance tracking, governance, and security. On the governance side, this includes embedding auditability, bias detection, compliance across various processes, and now agents. Marinela Profi, Global AI Market Strategy Lead, SAS, explains: The evolution of SAS Intelligent Decisioning positions the product as the ideal technology for building, orchestrating, governing, and validating agentic workflows, particularly where both deterministic and generative models are layered with decisioning constructs, established business rules, and data governance. The broader context is to help enterprises prepackaged agents that complement agentic orchestration frameworks like Agentspace and AutoGen by delivering domain-specific agents built on validated, production-ready AI models. Profi says: Unlike approaches that begin with general-purpose LLMs (Large Language Models), we see agents as a natural extension of our models' initiative, making them easier to govern, deploy, and integrate. The new platform will help democratize agent development with low-code development tools accessible to business users and programming-savvy developers. Business users will be able to define business rules where human review is needed using the graphical interface. Profi explains: The appropriate balance of human and AI collaboration is dictated by established business policies, protocols, and processes. All business rules can be made available in a governed repository, for use across different scenarios and use cases, and updated as the business evolves, with versioning and full auditability. SAS is also working to develop a training program to help employees at all levels grasp the risks and best practices for using agentic and autonomous decisioning tools. This includes workshops, hackathons, certifications, and media that explain complex agentic governance in practical terms. Profi says: We are also collaborating with industry partners and academic institutions to develop broader community initiatives, ensuring that both current and future professionals are equipped to handle the evolving landscape of autonomous AI with a strong foundation in ethical decision-making and risk management. The new tooling is being designed to help with practical aspects around making it easier to configure, monitor, and adjust levels of human oversight. Decisions are logged, auditable, and available for tracking. Business users can investigate the impact of each variable and asset version used across the decisioning environment. They can also perform outcome analysis in natural language to highlight the effects depending on scenario risk, context, and business goals. Agents can also be tested across various scenarios to validate performance and track data paths through decision gates. Profi hopes that this attention to detail will spark discussion across the industry: Our focus on highlighting the complexities and risks of autonomous agents, coupled with integrated governance and decision management, is designed to shift the conversation toward responsible AI deployment. By emphasizing the importance of transparency, accountability and ethical oversight, we aim to foster a broader industry-wide recognition that AI autonomy must be carefully balanced with human oversight. This will encourage other vendors and enterprises to adopt more rigorous governance frameworks, ensuring that AI-driven decisions are not only efficient but also trustworthy and aligned with organizational values. The current crop of generative AI models is trained on static data sets and then sporadically retrained and updated. These models are helpful in generating responses based on patterns in historical data, but cannot interact with and adapt to dynamic, real-time environments. More autonomous AI systems and processes will require adopting embodied AI approaches using techniques like reinforcement learning and active inference to improve over time, making them more agile and responsive, and adapting based on feedback loops. Profi explains: Governance and risk management considerations must evolve significantly with this shift. In a static generative AI model, governance focuses on ensuring that the model is trained on diverse, unbiased data and is regularly updated to remain relevant. However, for embodied AI, the challenge is more complex. These systems are learning and adapting continuously, which introduces dynamic risks, such as the potential for the model to adapt in ways that are unintended or introduce new biases as it interacts with evolving data. At SAS, we address this by emphasizing hybrid governance models that balance AI autonomy with human oversight. For example, we integrate strong decision traceability and auditability mechanisms into the AI systems, enabling organizations to track not only the actions taken by the AI but also the rationale behind those decisions in real-time. We also focus on embedding ethical guidelines, risk mitigation protocols, and bias detection mechanisms into the very fabric of the system, ensuring that as the AI adapts and learns, it does so in a way that is transparent, explainable, and aligned with organizational values. Five years ago, enterprise automation looked like tooling for API management, low-code development, process mining, and business rules management. Vendors across these disparate aspects are extending their tools to support the emerging agentic AI imperative. SAS's current approach is interesting since it builds on its extensive experience in helping enterprises address compliance and risk management concerns. This experience could also play an important role in guiding cross-industry efforts to shape agentic AI governance and trust as well.

SAS calls agentic AI its 'bread and butter' - SiliconANGLE

SAS builds on agentic offerings with industry-specific models, better governance The new star on the artificial intelligence scene is agentic AI, a shift SAS Institute Inc. saw coming. The company calls the "decisioning space," or software based on autonomous AI agents, its "bread and butter." This week, SAS announced it was building on its Viya cloud-based and AI analytics platform with updates focused on improving AI governance and giving customers more control. "When we look at the macro audience of who we're targeting here, it's really the builders and buyers of AI ultimately," said Alice McClure (pictured), senior director of product marketing at SAS. "The teams that [builders] work on need to be personally productive. The buyers need to be able to address very complicated business cases ... and so it's those two audiences that ultimately our agentic AI strategy is really all about and making sure that we're serving them with the right tools and the right solutions." McClure spoke with theCUBE's Rebecca Knight and Scott Hebner at SAS Innovate, during an exclusive broadcast on theCUBE, SiliconANGLE Media's livestreaming studio. They discussed SAS' agentic AI platform and industry-specific agents. (* Disclosure below.) Many users are concerned about the autonomy and accuracy of agentic AI, according to McClure. SAS has attempted to alleviate those concerns with better governance and pre-packaged agents that can perform industry-specific tasks. "The really core agent space around decisioning is the building, the deployment and the governance of agents," she said. "You're able to bring in business rules, you're able to bring in the workflow, the path to decisioning ... and be able to track the path of that decision and govern it along the way and have lineage all the way back again to where you started with the data." SAS also reinforces a "human-in-the-loop" strategy -- one that doesn't just insert oversight, but defines it. The critical distinction is understanding precisely what role the human plays in the decision chain: what requires human scrutiny, what can be safely delegated to agents and where accountability resides across that continuum. This clarity is essential to building trust in automated outcomes, according to McClure. "I'm excited for the fear to get out of the system a bit more," she said. "I'm excited to see customers winning with agent deployments. I'm excited for us to, for the market in general, to have a level of confidence and feel empowered around these kinds of decisions." Here's the complete video interview, part of SiliconANGLE's and theCUBE's coverage of SAS Innovate: