SAS update targets responsible agentic AI development
The longtime independent analytics vendor's latest Viya platform update includes Intelligent Decisioning, a set of governance-infused capabilities for developing agents.
SAS on Wednesday launched Intelligent Decisioning, a new set of capabilities within its Viya analytics platform that enables customers to responsibly develop, deploy and scale agentic AI applications.
The Viya update -- unveiled during SAS Innovate, the vendor's user conference in Orlando, Fla. -- also includes Viya Copilot, a synthetic data generator called Data Maker in preview that follows its November 2024 acquisition of Hazy, and new features in a cloud-based coding environment titled Viya Workbench. In addition, SAS introduced new AI models for different industries such as healthcare and manufacturing, and new digital twin capabilities.
Unlike generative AI assistants that enable users to ask questions and receive responses using natural language, GenAI-powered agents don't need to be prompted and can act autonomously.
For example, agents can be trained to take on processes such as documentation to relieve humans of time-consuming work to make them more efficient. In addition, they can be trained to search and monitor an enterprise's data and alert users to new insights that make them better informed.
SAS Intelligent Decisioning provides the tools to develop and deploy agentic AI applications. In addition, it comes with governance capabilities that embed auditability, bias detection and compliance into all agents.
SAS is not the first analytics vendor to offer agentic AI development capabilities. Domo, for example, has launched such capabilities.
SAS Intelligent Decisioning is an important move because it packages SAS' traditional strengths in advanced analytics, governance and compliance, and auditability -- especially important for industries focused on responsible AI -- into their new agentic AI framework.
Michael NiAnalyst, Constellation Research
However, SAS's measured approach that emphasizes governance and responsible development is important, according to Michael Ni, an analyst at Constellation Research. It not only adds what now represents advanced analytics capabilities but also does so in a way that emphasizes business outcomes rather than shiny new tools.
"SAS Intelligent Decisioning is an important move because it packages SAS' traditional strengths in advanced analytics, governance and compliance, and auditability -- especially important for industries focused on responsible AI -- into their new agentic AI framework," Ni said. "It's SAS' way of saying, 'We're not trying to be the flashiest; we're the grown-up in the room for AI decisioning.'"
Based in Cary, N.C., SAS is a longtime analytics vendor that, like many of its competitors, has expanded into AI development since OpenAI's November 2022 launch of ChatGPT represented a significant advancement in GenAI technology.
New capabilities
Agents began to emerge as the next evolution of enterprise AI during the second half of 2024. But just as it took considerable time for enterprises to begin developing GenAI assistants and for vendors to provide customers with GenAI tools that speed and simplify developing natural language processing applications, it's taken time for agentic AI to move from idea to reality.
In recent months, data management and analytics vendors such as Databricks and Snowflake have introduced agentic AI development capabilities while others, including tech giant Google Cloud and data observability specialist Monte Carlo, have launched agents of their own that make their platforms easier and more efficient for users.
Now, SAS is putting out its own agentic AI development suite. And just as it took a measured approach to GenAI development during the initial stage of excitement about GenAI following the launch of ChatGPT, waiting 10 months to unveil GenAI capabilities as competitors raced to reveal their GenAI plans, SAS is taking a careful approach with agentic AI.
"In a world of expanding analytical and generative AI, SAS is carving out a space on the responsible edge of AI where trust, domain knowledge and decision quality matter more than trend-chasing," Ni said.
SAS Intelligent Decisioning is built on three pillars, according to the vendor. They include the following:
Decisioning that combines deterministic analytics -- an approach to BI based on fixed rules and known relationships between data points -- and the reasoning capabilities of large language models.
A balance between AI autonomy and human involvement determined by each individual organization.
Built-in governance that ensures agents adhere to ethical standards and comply with both organizational values and regulatory statutes.
Given that Intelligent Decisioning provides a framework for agentic AI development, it's a significant addition for SAS users, according to Mike Leone, an analyst at Enterprise Strategy Group, now part of Omdia.
"Embedding governance controls is particularly important because enterprises need the most guidance with responsible AI implementation," he said. In this case, SAS is providing customers with a valuable balance of AI agent autonomy with human oversight.
Collaboration with customers provided the impetus for developing Intelligent Decisioning, according to Jared Peterson, senior vice president of platform engineering at SAS.
"Its existence is rooted in a deep collaboration with our customers and an understanding of the things they struggle with while trying to get real business value from applying AI and analytics in the enterprise," he said.
Beyond Intelligent Decisioning, Viya Copilot and Data Maker highlight SAS's Viya update, according to Leone.
Viya Copilot -- currently available in invitation-only preview -- is a GenAI-powered conversational interface embedded into the platform that provides developers, data scientists and analysts with an assistant. Initial applications for Viya Copilot, which is built on Microsoft Azure AI Services, include code assistance and model development.
Data Maker, expected to be generally available by the end of September, is a synthetic data generator that helps enterprises address data privacy and scarcity challenges as they develop analytics and AI applications.
"Data Maker stands out because it helps organizations tackle common roadblocks of data privacy and limited data availability," he said. "Viya Copilot brings significant value to customers by embedding an assistant directly in the platforms. For SAS customers, there is immense value in the potential to democratize access to advanced analytics capabilities within the Viya platform."
Ni likewise cited Viya Copilot and Data Maker as significant additions.
"Viya Copilot and Data Maker deliver the kind of AI features SAS's core customers need," Ni said. "Copilot accelerates analytics development, while Data Maker tackles two of the biggest blockers to AI adoption -- data privacy and data scarcity. Together, they reinforce SAS' differentiation in tailored solutions that are governed and trusted."
As with Intelligent Decisioning, customer feedback played a role in the development of the other new Viya features, according to Peterson. Each is geared toward different personas within organizations. For example, Viya Workbench aims to meet the needs of developers and data scientists, while Viya Copilot is targeted at business analysts.
Looking ahead
As SAS plots its product development roadmap, making Viya Copilot pervasive throughout the entire SAS platform and providing more tools to assist customers as they develop AI agents are priorities, according to Peterson.
"We're giving them the ability to deploy and scale AI agents with balanced human and AI autonomy," he said.
In addition, SAS is exploring quantum AI and how to guide customers as they apply quantum technology to their operations.
Ni, meanwhile, suggested that in addition to continuing to invest in agentic AI development, SAS could do more to attract new users, such as embracing open source technology and simplifying the use of its platform.
"What continues to be true … is for SAS to continue agentic innovation, enhance open source appeal to streamline integration with the open source ecosystems to attract developers who are wary of proprietary systems, and expand low/no code tools and training to ease the transition for legacy users and new adopters," he said.
Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than 25 years of experience. He covers analytics and data management.
