A foundation for AI

SAS has taken a measured approach to AI since OpenAI's November 2022 launch of ChatGPT marked significant improvement in generative AI (GenAI) and sparked rising interest in AI development.

While some vendors were quick to integrate with ChatGPT and others including Domo and Qlik quickly put together GenAI development environment, SAS was careful not to provide similar tools until it addressed the accuracy and security problems that plagued early GenAI development. As agentic AI has become the vanguard over the past two years, SAS has similarly been more patient than some of its peers to provide capabilities that help customers build and manage agents.

Now, however, as the experimental phase of agentic AI development is closing and a production phase is beginning, SAS is launching tools specifically aimed at helping users build governed, trustworthy AI tools that perform as intended, according to Jared Peterson, the vendor's senior vice president of global engineering.

"We're designing for a moment where AI has to actually work in production, where the stakes are real and the answers have to be defensible," he said. "Every announcement … is essentially SAS saying, 'Here's something you can stand behind' -- governed data, traceable decisions, agentic capability with trusted execution underneath."

Meanwhile, customer feedback provided the impetus for adding the new capabilities, Peterson continued.

"I get to meet with customers regularly, and a lot of what we're announcing starts in those conversations," he said. "We try to distill themes across them and then pressure-test them against what we're seeing in market dynamics, research and academia."

David Menninger, an analyst at ISG Software Research, noted that SAS's comparatively measured approach to GenAI and agents is somewhat ironic given that its commitment to AI long predates ChatGPT. However, with longtime customers expecting production-ready capabilities, SAS's caution is appropriate.

"SAS typically moves slowly into new technology areas, and agentic AI has been no exception," Menninger said. "This approach has enabled SAS to maintain a multi-billion-dollar business supporting the data and analytic needs of some of the world's largest enterprises."

Additionally, what is significant about the new capabilities beyond their mere addition to SAS's array of offerings is that they're accessible through Viya, he continued.

"Data and analytics vendors are all converging on ways to make data management and analysis easier via the use of AI," he said. "What's important here for the SAS customers is that these are now being delivered on the SAS Viya platform."

The new tools

Specific new SAS capabilities -- among others -- include the following:

• The Agentic AI Accelerator development framework, featuring no-code and low-code options in addition to code-first capabilities for experienced developers.
• An MCP server to standardize connecting SAS models and other capabilities with external agents.
• Viya Copilot, a conversational assistant grounded by an organization's security and governance that helps users throughout the analytics lifecycle including data discovery, writing code, building data models, developing dashboards and analyzing data.
• Industry-specific copilots generally available for asset and liability management and health clinical data discovery.
• An industry-specific copilot in preview for supply chain management.
• SpeedyStore to deliver AI and analytics to data rather extract data out of its storage environment and move it into SAS.
• AI Navigator, a software-as-a-service feature expected to be GA during the third quarter that will enable enterprises to better inventory and govern their use of AI, including models and agents and the policies applied to them.

While not groundbreaking, SAS's industry-specific capabilities somewhat distinguish its AI and AI development capabilities, according to Menninger.

"Enterprises are focused on meeting specific business needs, not building better data engineering pipelines," he said. "These domain-specific tools help enterprises get one step closer to real business issues."

Farmer called SAS's new capabilities incremental improvements, though he noted that the MCP server, while catching up to what competing vendors already provide, is "genuinely important." Meanwhile, like Menninger, he noted the value of SAS's industry-specific agentic AI capabilities while also highlighting SpeedyStore.

"The industry agents are interesting, especially because SAS has such a strong base of users who have built their business around SAS's vertical applications … but my personal favorite would be SpeedyStore," Farmer said. "It is catching up again … but what makes it significant is that it closes a long-standing gap where customers had to choose between SAS Analytics and a modern lakehouse foundation."

Future plans

As SAS plots its next wave of new products, adding more AI agents and copilots is a prominent initiative, according to Peterson. In addition, making it easier for users to migrate workloads to SAS and investing in improving Viya's architecture are part of the vendor's roadmap.

"And underneath all of it [is] governance that keeps pace with adoption," Peterson said. "As AI scales inside organizations, governance has to become operational. That's not separate from the innovation agenda. It's what makes it durable."

Menninger, meanwhile, noted that SAS has been slowly developing an ecosystem beyond its platform by supporting cloud deployments, open source platforms and open data formats. With agentic AI demanding greater interoperability between systems, making SAS even more open would be wise, he advised.

"Continuing to build out and expand the ecosystem of partners, tools and technologies it supports will help SAS customers deploy its products in more scenarios across their enterprises," Menninger said.

Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than three decades of experience. He covers analytics and data management.