The world of enterprise analytics has just undergone a fundamental shift. While in past years we learned how to "chat" with data using large language models (LLM), now an era is coming when AI ceases to be merely a passive advisor and becomes an active participant in processes. SAS announced the expansion of its flagship platform SAS Viya with capabilities of agentic AI and governed AI assistants.
From text generation to autonomous action: What is agentic AI?
To understand this shift, it is important to distinguish between common generative AI, known from ChatGPT or Claude, and so-called agentic AI. While a common model generates a report or summary based on a query, agentic AI has the ability to plan, use tools, and execute steps to achieve a set goal.
Within the SAS Viya platform, this means that the assistant will not only answer the question "What were our profits last quarter?" but can be instructed: "Analyze profit trends, identify three main causes of decline in the Central Europe region, and prepare a cost reduction proposal to email to managers." In this case, the agent must search databases, perform statistical analysis, interpret results, and subsequently use other software for communication.
Key pillars of the new SAS Viya expansion
The new features rely on three main areas that define modern enterprise AI:
- Agentic capabilities: The ability of AI to interact with existing systems and carry out complex workflows.
- Governed AI: Implementation of strict rules ensuring that AI does not hallucinate, does not misuse sensitive data, and its decision-making is transparent.
- LLM integration: SAS Viya allows companies to use various models (e.g., from OpenAI, Google, or Anthropic) in a secure, isolated environment.
Governance: Safety fuse for European companies
For Czech and European companies, the biggest barrier to AI adoption is not a lack of technologies, but risk. Concerns about corporate data leaks, opaque decision-making processes, and legislative non-compliance are legitimate. This is where governance comes into play.
In the context of strict European regulation, especially EU AI Act, the ability to control how an AI model arrives at a result is absolutely critical. SAS Viya offers a framework where every agent action is recorded, auditable, and subject to company-defined rules. This means that if an AI assistant in a bank proposes a credit limit change, the decision-making process must be fully transparent and in compliance with the bank's internal regulations and European laws.
This approach positions SAS not just as a "layer over data", but as a safe haven for enterprise data. Unlike public ChatGPT interfaces, where data may be used to train future models, SAS Viya ensures that your data remains yours and the model uses it only within your controlled instance.
Comparison: SAS Viya vs. general LLM (GPT, Gemini, Claude)
It is important to realize that SAS Viya is not a direct competitor for individuals using ChatGPT. It is an entirely different category of tools. Below is a comparison for the enterprise segment:
| Feature | General LLM (e.g., ChatGPT/Claude) | SAS Viya (Agentic AI) |
|---|---|---|
| Main purpose | Creativity, text, general queries | Data analysis, process automation, decision-making |
| Governance | Minimal (depends on company settings) | High (Auditability, compliance with EU AI Act) |
| Data handling | Need to upload files (leakage risk) | Direct integration with enterprise databases |
| Autonomy | Reactive (waits for prompt) | Proactive (agentic workflow) |
Practical impact: What does it mean for the Czech market?
For Czech industry, banking, and public administration, this has several direct impacts:
- Supply chain optimization: Czech manufacturing companies can use agentic AI to monitor inventory and automatically negotiate with suppliers when a shortage is detected.
- Banking sector: Automation of compliance and fraud detection, where agents not only identify an anomaly but prepare a complete report for the regulator.
- Availability and price: SAS Viya is an enterprise solution. There is no "free tier" for ordinary users. The price is set individually based on need (subscription model) and is intended for medium and large corporations. In the Czech environment, these tools are implemented through specialized partners and require an investment in data infrastructure.
While tools like ChatGPT are great for everyday individual productivity, SAS Viya positions them as the operational brain of a modern organization, operating within the strict rules required by the European market.
Can agentic AI in SAS Viya make a mistake and what happens to the data?
Thanks to the governance layer, every agent action is monitored. If the agent evaluates that an action requires a higher degree of certainty than is set, it stops and asks a human for approval (Human-in-the-loop). Data remains in an isolated environment and is not shared with public models.
Is SAS Viya available in Czech?
The platform itself and its analytical functions are primarily in English, which is standard for global enterprise software. However, thanks to LLM integration, users can communicate with assistants in Czech if they use models with Czech language support (such as GPT-4 or Claude), which are connected through SAS Viya.
What is the difference between generative AI and agentic AI in practice?
Generative AI will write you an email that you are running low on stock. Agentic AI finds out that you are running low on stock, finds the cheapest supplier, compares prices, and prepares an order for your approval.