Unlocking the Future of Automation with Agentic AI

In the rapidly evolving landscape of intelligent automation, the concept of agentic AI is transforming how businesses approach Robotic Process Automation (RPA). This article explores the developments in UiPath’s platform, focusing on their agentic AI capabilities, Agent Builder and Maestro.

We delve into how AI agents are changing traditional automation by blending humanlike reasoning with robotic precision. It’s not just about following the market trend – cases where traditional RPA falls short started to pile up some time already. Now it’s time to finally tackle them.

Or at least give it a try.

Do you prefer watching to reading? Check out our YouTube channel and watch the webinar on which the following article is based:

The Journey from traditional RPA to Agentic Automation

The automation world has long been dominated by RPA tools that execute predefined tasks with binary precision. However, as organizations face increasingly complex workflows involving unstructured data, regulatory nuances, and human interactions, the need for more flexible, intelligent systems has become apparent. Enter agentic AI: a philosophy where AI agents think, robots execute, and humans lead.

Everyone seems to be developing agents now. This paradigm shift is vital for starting new projects and addressing cases that seemed impossible in the past, but it’s equally important to combine the RPA and AI ingredients well. Our early access and hands-on experimentation with the platform revealed a thoughtfully designed system that integrates AI agents with traditional RPA and human workflows. It’s still undergoing some changes, but it already opens up new possibilities for our client.

To appreciate what we can use today, it’s essential to look back at how the platform has matured over the past seven years. Originally, UiPath offered three core free products:

• Studio: For building robotic automations.
• Orchestrator: For managing and orchestrating robots.
• Robots: For executing automation tasks.

Since then, the platform has expanded through in-house R&D and strategic acquisitions, adding capabilities that address the complexities of modern enterprise environments:

• Computer Vision: Enabling automation in restricted environments by understanding screen content visually.
• API Automation: Simplifying integration for developers through integration services.
• Web Interface Building: Allowing the creation of modern interfaces that interact with backend automations or refresh legacy systems without altering source applications.
• Orchestration Enhancements: Robots can now escalate to humans when uncertain, such as when machine learning models yield ambiguous predictions.
• Governance and Security: Incorporating robust governance features, especially for generative AI (GenAI) components.
• Cloud Deployment: Providing flexible deployment options to meet enterprise needs.
• Discovery and Analytics: Offering tools to analyze organizational processes, communications, and generate performance KPIs.
• Testing Capabilities: Ensuring RPA automations and the applications they interact with maintain reliability through automated testing.
• Intelligent Document Processing (IDP): Expanding from structured forms to unstructured documents like policies and contracts.
• Understanding Communications: reading, analyzing and automating any text-based communication channels, like emails or tickets.

And finally, the current advancements in technology led UiPath to develop GenAI building blocks, granting robots a “brain” to interact with various AI models securely and flexibly. This continuous innovation has culminated in the current platform’s ability to support agentic automation, where AI agents can autonomously handle complex tasks with unstructured input.

Act Two: The Era of Agentic Automation

Building on the robust foundations, UiPath is now ushering in Act Two–the era of agentic automation. This phase focuses on AI agents that think and decide, working alongside traditional robots and humans. The key components of this new era include:

• Security and Governance: Maintaining strong controls to ensure safe AI deployment via Trust Layer
• Integration and Context Counting: Leveraging multiple data sources and APIs to give agents comprehensive situational awareness.
• Low-Code Agent Builder: Enabling the creation of AI agents with customizable prompts, contexts, and tools.
• Agentic Testing: Facilitating the creation of test sets and validation cases to measure and improve an agent’s quality numerically.
• Process Orchestration: Using BPMN-modeled business processes to coordinate activities among agents, robots, and humans.

This approach allows organizations to orchestrate complex workflows where AI agents handle context and decision-making, robots execute rule-based tasks, and humans oversee and intervene when necessary.

The Symbiosis of Robots and AI Agents

Drawing a parallel to the human brain, UiPath positions robots and agents as complementary forces:

• Robots: Analogous to the left brain, they are structured, logical, and deterministic. Robots excel at repetitive, rule-based tasks requiring high reliability and efficiency, such as transaction processing in ERP systems.
• AI Agents: Comparable to the right brain, they are creative, adaptive, and goal-oriented. Agents thrive in ambiguous, unstructured tasks requiring reasoning, such as interpreting documents, answering questions, and managing communications.

However, AI agents’ flexibility comes with inherent unpredictability. To address this, UiPath employs a concept of “controlled agency” – a set of tactics and features designed to enhance agent trustworthiness and reliability.

Ensuring Trustworthy AI Agents

The concept of controlled agency involves iterative development and testing of agents through various mechanisms:

• Prompt Experimentation: Developers can interactively test and refine the natural language prompts that define an agent’s behavior.
• Dynamic Evaluations: Collections of test cases generate a quality score (0-100) that guides ongoing improvements.
• Context Grounding: Utilizing retrieval-augmented generation, agents access large amounts of company-specific data in a controlled manner to inform decisions.

These elements combine to build AI agents capable of accessing knowledge and understanding context – consolidating information, summarizing data, and making decisions – while maintaining transparency and control.

Building Blocks of an AI Agent

Each AI agent comprises several key components:

1. Natural Language Prompt: The agent’s “job description,” detailing the goal and approach.
2. Context: Company-specific knowledge such as databases, documents, and policies that inform the agent’s understanding.
3. Tools: Integration service activities, other automations, or even other agents that the AI agent can invoke to execute tasks.
4. Escalations: Mechanisms for the agent to request human intervention when encountering uncertain or abnormal situations.
5. Memory: After human feedback, the agent stores lessons learned to avoid repeating mistakes.

One of the most important aspects of agentic AI is maintaining human oversight. Agents are designed to escalate uncertain cases to humans, ensuring accuracy and control. This is especially crucial when dealing with generative AI, which can sometimes produce unpredictable outputs.

This modular design allows AI agents to interact with complex enterprise systems, collaborate with robots, and seek human guidance when necessary, making them flexible components of automation.

Managing Complex End-to-End Processes with Maestro

In real-world enterprises, business processes rarely reside within a single system. They often span multiple applications – Salesforce, SAP, Excel – and involve many human actors. To manage this complexity, UiPath introduces Maestro, a process-oriented orchestration tool that offers both a wide and detailed view of automation workflows.

Maestro enables organizations to:

• Model processes visually using BPMN 2.0 standard notation or import existing process maps.
• Implement processes by linking BPMN blocks to UiPath activities, including agentic tasks, traditional RPA, human-in-the-loop actions, and third-party agents.
• Operate and monitor process executions with detailed visibility into individual instances.
• Employ retry mechanisms and fault tolerance to preserve progress in long-running processes.
• Analyze performance using heat maps and process mining to identify bottlenecks and optimize workflows.

Maestro acts as a centralized canvas where agents, robots, and humans collaborate.

Benefits of Maestro for Different Personas

• Automation and AI Leaders: Securely integrate AI agents from various silos, orchestrate complex workflows, and monitor performance.
• Process Owners: Manage the entire process lifecycle–from modeling and implementation to monitoring and continuous improvement–using a single tool.
• IT Leaders: Reduce total cost of ownership by consolidating workflow management and governance within one platform, supporting business rules and guardrails.

Practical Demo: Automating Recruitment with AI Agents and Robots

To illustrate the power of agentic AI in action, consider a recruitment process automated using UiPath’s Agent Builder and Maestro. In this scenario, candidates apply for an RPA developer position by sending emails with resumes attached. The process involves several manual tasks for HR recruiters:

• Reviewing resume completeness.
• Answering common candidate questions based on internal policies.
• Aggregating candidate applications.

The automated workflow addresses these challenges by combining AI agents, robotic automations, and human-in-the-loop validation:

1. Email Reception: The system detects a new application email.
2. Initial Screening Agent: Checks for suspicious content in the email.
3. Resume Check Agent: Analyzes the attached CV for missing information, such as language proficiency.
4. Answer Generator Agent: Responds to candidate questions using company policy documents as context.
5. Response Assembly Agent: Combines resume feedback and answers into a draft reply.
6. Human Validation: The recruiter reviews and approves the response before sending.
7. Data Storage: Candidate information is stored in a database, and incomplete applications are flagged.
8. Follow-up Agents: Generate interview questions and evaluate candidate fit based on profile data.
9. Final Data Handling: Pure RPA flows save candidate data to SharePoint for HR review.

This demo highlights the capability of agentic AI to handle unstructured data (emails and resumes), interpret policies, and interact with humans – a significant leap beyond traditional rule-based robots.

In practical terms, this means recruiters no longer need to go through hundreds of emails manually, hunting for missing data or drafting repetitive responses. Agents can pre-screen applications, request any missing information, and even answer standard candidate queries – freeing up valuable time and reducing the risk of overlooking a qualified applicant due to the huge amount of applications. Automating these early, repetitive tasks ensures consistency and responsiveness across all applications, regardless of scale.

Importantly, the human element remains central. While AI agents streamline the initial touchpoints and prepare structured feedback, it’s still the recruiter who validates communications and makes hiring decisions. This hybrid model strikes the right balance: boosting efficiency without surrendering critical judgment to algorithms. In high-volume recruitment scenarios, it’s not just a matter of saving time – it’s about enabling HR teams to focus on what matters most: selecting the right people.

Frequently Asked Questions (FAQ)

What did the webinar participants ask?

How well do AI agents handle complex documents like resumes with tables?

AI agents generally handle complex documents very well, including resumes containing tables and various formats. However, some edge cases–such as ratings represented by stars–may pose challenges. Continuous testing and refinement help improve accuracy in these scenarios.

Can the human-in-the-loop interface show original documents or links during validation?

Yes, UiPath applications and Action Center offer extensive customization to include links to original documents or attachments, allowing human reviewers to access all necessary information during validation.

Are data sent to external AI providers like OpenAI when using UiPath agents?

No, the current models used by UiPath agents are hosted in UiPath’s private cloud infrastructure, ensuring data privacy and security. Additionally, a “bring your own model” option is on the roadmap for organizations with strict confidentiality requirements.

How much does it cost to build and run an AI agent?

Pricing is consumption-based and depends on the complexity and usage of the agent. More complex agents with higher call volumes incur higher costs. For a precise estimate, it’s recommended to consult with UiPath or Office Samurai experts.

How long does it take to build a reliable AI agent?

Building a simple agent can take 1-2 hours, but thorough testing to ensure reliability may require more time. The best practice is to keep agents focused on specific tasks rather than attempting to automate an entire process within a single agent.

When should I use UiPath Maestro versus other AI orchestration tools like N10?

Maestro is ideal for enterprise-scale automation requiring robust security, reliability, and scalability. In contrast, tools like N10 may be more suitable for niche or custom solutions developed from scratch. Maestro integrates deeply with UiPath’s ecosystem, offering a comprehensive workflow management engine.

What does the “temperature” setting mean in agent testing?

Temperature controls the creativity or randomness of the AI model’s responses. A value of 0 means deterministic and precise outputs, while 1 allows more creative freedom. For most business automations, a low temperature is preferred to maintain consistency.

How does Maestro relate to UiPath Orchestrator?

Maestro serves as a process modeling and orchestration interface that integrates with UiPath Orchestrator. After designing processes in Maestro, you can deploy and trigger them through Orchestrator, benefiting from its scheduling, monitoring, and governance capabilities.

Embrace the Agentic AI Revolution

The fusion of AI agents with Robotic Process Automation initiated a new era of business automation – one where computers don’t just follow instructions blindly but can “think”, adapt, and collaborate with humans. UiPath’s Agent Builder and Maestro provide a powerful, integrated platform to utilize this potential, enabling enterprises to automate complex, unstructured workflows at scale securely and reliably.

If you’re facing challenging processes that involve communication, document interpretation, or require flexible decision-making, it’s time to consider agentic AI solutions. Whether you’re an automation leader, process owner, or IT professional, these tools allow you to create intelligent assistants that work alongside your teams, boosting productivity and innovation.

Ready to start your agentic AI journey? Reach out to experts who can guide you through building and deploying AI agents matching your business needs.

The future of automation is taking shape – now’s the perfect time to join the journey.