Listen to this article:
For a long time, we witnessed the era where AI was seen as a "roommate" — a tool that helps us write an email or summarize a document. Today, however, we are entering a phase where AI takes on the role of an "operator." SAS has introduced a major expansion of its SAS Viya platform that integrates so-called agentic AI and governed assistants. This move is aimed at organizations that no longer just want to "chat" with ChatGPT but need AI that will actually work with their data, follow security protocols, and comply with regulations like the EU AI Act.
From Chatbots to Agents: What's Changing in SAS Viya?
The main pillar of the new announcement is the transition from passive models to active agents. The difference is fundamental. While a typical chatbot waits for your query, agentic AI has a defined goal and can plan the steps to achieve it on its own.
Within the SAS Viya platform, we can now distinguish two key types of tools:
- SAS Viya Copilot: A conversational assistant integrated directly into analytical workflows. It helps data scientists generate code, create visualizations, and propose models. It is a "copilot" that helps you fly the plane.
- Agentic AI Framework: Infrastructure for building autonomous agents. These agents can perform specific tasks within the company, such as monitoring risky transactions or optimizing supply chains, without requiring constant human intervention at every single step.
For businesses, this means that AI ceases to be an isolated experiment and becomes part of real operations. According to a study by Futurum Group, organizations using SAS Viya can accelerate their AI workflows up to 4.6 times compared to competing solutions. This is not just about speed, but about the ability to turn data into decisions faster.
Governance: The Key to Survival in the EU's Regulated Environment
For Czech and European companies, the biggest barrier to AI implementation is not its performance, but security and regulation. General models such as GPT-4 from OpenAI or Claude from Anthropic are extremely intelligent, but their use in banking or healthcare runs into the "black box" problem — we don't know exactly how the model arrived at the result, and data may leak into public training sets.
SAS addresses this problem through the concept of governed AI. The new Model Context Protocol (MCP) Server allows external AI agents to access SAS analytical tools, but only within strictly defined rules. This means the agent only has access to the data the system permits, and every decision it makes is auditable. In the context of the strict EU AI Act, this is a capability that common consumer AI tools lack.
Synthetic Data and Privacy Protection
Another significant element is the SAS Data Maker tool. At a time when personal data protection (GDPR) is a priority in the EU, SAS offers the generation of synthetic data. This data looks and behaves like real data but is not linked to any specific individual. This allows companies to train their models and test systems without the risk of violating privacy, which is ideal for the Czech banking sector or pharmaceutical companies.
Comparison: SAS Viya vs. General LLM Models
To understand the value of this solution, we must compare it with the tools we know from everyday use:
| Feature | General LLM (ChatGPT/Claude) | SAS Viya (Agentic AI) |
|---|---|---|
| Primary purpose | Text generation, creativity | Data analysis, decision-making, automation |
| Governance | Minimal (hallucination risk) | High (auditability, control) |
| Data context | General/Internet | Domain-specific (corporate data) |
| Availability in Czechia | Yes (web/app) | Yes (for enterprise clients) |
Pricing and Availability
Unlike ChatGPT, which offers an affordable subscription (approx. $20/month), SAS Viya is an enterprise solution. This means there is no universal "individual price." SAS offers various licensing models based on need, including SAS Viya Essentials, a managed cloud service designed for small and medium-sized businesses. For an exact price quote, companies must contact a SAS sales representative directly.
For the Czech market, it is important to emphasize that although the platform interface is primarily in English (which is standard for global analytical systems), the tool is fully compatible with European data standards and legislation, making it a safe choice for local corporations.
Do I need to be a data scientist to use SAS Viya Copilot?
Not necessarily. Copilot is designed to help even users with lower technical skill levels through a conversational interface. For example, it can help create a dashboard or explain a complex graph using natural language.
How does agentic AI differ from a regular automation script?
A script always does the same thing according to fixed instructions. Agentic AI can evaluate the situation within its goal; if it encounters an obstacle, it tries a different approach and can react to changes in data that occur during the process.
Is it possible to connect SAS Viya with my own models (e.g., Llama 3)?
Yes, thanks to the introduction of the Model Context Protocol (MCP) Server, it is possible to connect SAS analytical capabilities with external AI models, enabling the creation of hybrid systems.
