SAS Embraces Third-Party AI: Viya Integrates External LLMs, Unified Governance on the Way

In 2026, AI in business ceased to be a singular pilot project. In more mature organizations, proprietary models, machine learning services from hyperscalers, agents built on commercial LLMs, and internally developed prototypes coexist. The challenge is no longer just to launch AI initiatives but to keep them cohesive, map them, monitor them, and ensure they operate within both internal rules and those of regulators. The phenomenon of shadow AI—the adoption of tools outside the IT perimeter—has become a tangible concern for CIOs, and according to Gartner's forecasts, by 2030, over 40% of companies will experience security or compliance incidents related to the unauthorized use of these technologies.

In this scenario, SAS, at Innovate 2026, not only celebrated its fifty years of activity but also presented two innovations that significantly change how the company addresses the enterprise AI market. The first is the SAS Viya Model Context Protocol Server, and the second is SAS AI Navigator. Edge9 is present in Dallas to cover the event.

Viya's Opening to Third-Party Agents

The SAS Viya Model Context Protocol Server adopts the open MCP standard, already prevalent in the LLM ecosystem, to expose Viya's analytical functions, models, and decision automation logics as callable tools by external AI agents. A company that has chosen Claude as its conversational engine, for example, can make its agent directly access Viya's pipelines, risk models, and scoring rules developed in SAS, without duplicating code and while remaining within internal governance boundaries.

This move aligns with the viewpoint of Jared Peterson, Senior Vice President Global Engineering at SAS, who states that "the role of human expertise, when deploying agentic AI, does not diminish with automation; rather, it amplifies." Practically translated, the company accepts that the conversational agent does not necessarily have to be a SAS product, as long as the resulting decisions are made through models, data, and rules under governance. This package includes two additional components that align with this direction.

SAS Agentic AI Accelerator is a no-code and low-code framework for teams looking to design and deploy their own agents within Viya. SAS Retrieval Agent Manager is a no-code RAG solution, currently distributed as a standalone product and on the roadmap to integrate natively into the platform.

AI Navigator: Cross-Cloud Governance

The second announcement is SAS AI Navigator, a SaaS service expected in the third quarter of 2026 on the Microsoft Azure Marketplace. It is designed for AI, data, compliance, and risk managers and allows for a unique inventory of all AI use cases present in a company, the underlying models, and the rules that apply to them. What distinguishes it from competing offerings is that the governance also applies to models and agents not developed by SAS. A corporate chatbot based on Claude or Microsoft Copilot, for example, falls under the same policy perimeter that applies to Viya's internal models.

Reggie Townsend, Vice President SAS AI Ethics, Governance, and Social Impact, describes the product as a growth tool rather than a control measure. "AI governance is often seen as a compliance measure; instead, it is a growth engine. It allows pushing the limits of technology within a structured, transparent, and secure perimeter." A second point he makes addresses the operational dimension: "The greatest risk of an AI governance program is not regulation, but adopting such a complex tool that no one ends up using it." The implicit message is that SAS wants to present AI Navigator not as a compliance burden but as a usable starting point for CIOs who want to see, in one place, where their AI stands.

Together, these two announcements signify a shift in direction for SAS. The historical closure of the platform gives way to a declared openness towards commercial LLMs, as a company's analytical stack in 2026 is inherently heterogeneous. Concurrently, SAS aims to occupy a level that hyperscalers and application vendors are only partially covering—namely, the horizontal control of the models and agents already in use. External confirmation comes from Chartis Research, which at the beginning of 2026 placed SAS among the category leaders in its RiskTech Quadrant for AI governance, valuing the Viya platform for managing model risk, detecting biases, protecting privacy, and providing end-to-end monitoring of flows.