The world of automation is undergoing a fundamental shift. While the previous era was defined by robotic process automation (RPA), where software strictly followed predefined rules, we are now entering the era of agentic AI. In this phase, we are no longer just programming "if A happens, do B" — we are delegating complex tasks to agents that can independently assess a situation and choose the best path to the goal.

According to a recent announcement from UiPath, this new capability is now available within the Automation Suite, enabling deployment even in environments not connected to the public cloud. This is a critical parameter for the public sector, where data security is not a choice but a legal obligation.

What is agentic AI and how is it different from ordinary automation?

To understand the significance of this technological shift, it is necessary to distinguish between classic RPA and agentic automation. Traditional RPA functions as a very fast and precise digital employee that follows instructions. However, if the process changes (for example, the invoice format changes or an unexpected error occurs), the RPA process stops.

Agentic AI, on the other hand, uses large language models (LLMs) to gain the ability of reasoning (logical thinking). An agent can:

• Plan: Break down a complex goal into individual steps.
• Use tools: Access databases, emails, or web interfaces to obtain necessary information.
• Adapt: If it encounters an obstacle, it tries a different approach instead of simply failing.

In the context of the public sector, this means that an agent can, for example, process a grant application — not only filling out the form but also comparing attachments with legislative requirements, searching for missing information in other systems, and if necessary, proposing a solution for the official.

Security and data sovereignty: The key to adoption in government

The biggest barrier to the adoption of advanced AI by government institutions has always been the question: "Where do our data flow?" Most popular models, such as ChatGPT from OpenAI or Gemini from Google, require sending data to the provider's cloud. For public administration, which works with citizens' sensitive data, this risk is unacceptable.

The new version of UiPath Automation Suite solves this problem by enabling on-premises deployment (on own servers) or within a controlled cloud environment (e.g., AWS, Microsoft Azure, or OpenShift). Institutions can choose which LLM model they will use. They can utilize closed instances of models or even their own, locally trained models. This ensures full data sovereignty — data never leaves the organization's infrastructure.

Relevance for the Czech Republic and the EU

For the Czech public sector, banking, and healthcare, this announcement is highly relevant. With the arrival of the EU AI Act regulation, organizations in the European Union will have to prove that their artificial intelligence systems are safe, transparent, and comply with strict personal data protection standards (GDPR).

The ability to run agentic AI locally within infrastructure in the Czech Republic or within EU data centers is the only way for our authorities to effectively modernize services for citizens without endangering state security. Czech administration, for example in the digitalization of state services, can use these tools to massively reduce administrative burden without fear of data exfiltration to foreign servers.

Comparison with competition and price

UiPath is positioning itself in direct competition with giants such as Microsoft (with their Power Automate platform and Copilot) and Salesforce (with their Agentforce).

While Microsoft bets on deep integration into the Office 365 ecosystem, UiPath offers greater flexibility in which type of LLM (e.g., Llama 3, GPT-4, or custom models) you can "feed" your agents with. UiPath also profiles itself better in complex, hybrid scenarios where AI must interact with very old (legacy) systems that are not prepared for modern cloud APIs.

Pricing policy: UiPath is primarily an enterprise solution. There is no "free tier" for the full version of Automation Suite. The price is always individual and depends on the scope of deployment, number of robots, and type of licensing. For smaller companies and individuals, UiPath offers basic cloud versions, but for agentic AI capabilities in on-premises mode, it is necessary to contact the sales department for an individual price quote (usually in the range of thousands to tens of thousands of USD annually for enterprise deployment).

Summary of advantages and disadvantages

Advantages:

• Maximum security: Ability to operate entirely outside the public internet.
• Model flexibility: Choice between different LLMs according to needs.
• Efficiency: Transition from simple automation to autonomous task solving.

Disadvantages:

• High infrastructure demands: On-premises deployment requires powerful own servers and experts for management.
• Price: This is not a tool for small businesses with limited budgets.