From Bots to Brains: Getting Started with Agentic Automation in UiPath

From Bots to Brains: Getting Started with Agentic Automation in UiPath

• 1.
  From Bots toBrains Getting Started with Agentic Automation with UiPath Speaker : Srinivas K
• 2.
  Agenda Journey Towards AgenticAI High-level view of how Agentic AI works Characteristics of Agentic AI Guidelines for Effective UiPath Agent Effective Prompt Engineering Context Grounding Use case Explanation Demo
• 3.
  WHO AM I SRINIVASK(Srini) SeniorTech Lead @ Proservartner 1. UiPath MVP (Most Valuable Professional) 2022-2025 2. HyperHack Team member and Team secure 2nd Place. 3. 𝐀𝐈 𝐚𝐭 𝐖𝐨𝐫𝐤 𝐇𝐚𝐜𝐤𝐚𝐭𝐡𝐨𝐧 Team member and Team secure Runner-up Prize.
• 4.
  From Bots toBrains Journey towards Agentic AI from rule-following bots to intelligent agents capable of independent action.
• 5.
  How Does AgenticAI work How the bot Think
• 6.
  How Does AgenticAI work How the bot Think Autonomous The most important advancement of agentic systems is that they allow for autonomy to perform tasks without constant human oversight. Proactive Agentic systems provide the flexibility of LLMs, which can generate responses or actions based on nuanced, context-dependent understanding, with the structured, deterministic and reliable features of traditional programming. Specialized Agents can specialise in specific tasks. Some agents are simple, performing a single repetitive task reliably. Others can use perception and draw on memory to solve more complex problems. Adaptable Agents can learn from their experiences, take in feedback and adjust their behaviour. With the proper guardrails, agentic systems can improve continuously. Intuitive Because LLMs power agentic systems, users can engage with them with natural language prompts. This means that entire software interfaces—think of the many tabs, dropdowns, charts, sliders, pop-ups and other UI elements involved in the SaaS platform of one’s choice—can be replaced by simple language or voice commands.
• 7.
  Guidelines for EffectiveUiPath Agent Guardrails • Built-in safety checks & control tools • Trigger actions: Log, Block, Filter • Compliance path enforcement • Clear escalation proceduressyste Effective Prompt Writing • Use clear, structured prompts to guide AI behavior • Prefer few-shot or few-chain-of-thought prompting • Write thoughtful, outcome-focused prompts • Define expected format, tone, and style Context Grounding • Provide agent-specific knowledge and parameters • Importance: Ensures accurate reasoning and reliable execution Escalations • Enable human-in-the-loop handoff for critical decisions • Review sensitive agent actions (e.g., JSON/XML operations) • Pro Tip: Validate outputs for accurate agent execution • Example: Triggered issues → realistic, explainable outcomes
• 8.
  Effective Prompt Engineering Zero-ShotPrompt • Asking the model to perform a task without providing any examples • The model relies purely on its pre-trained knowledge to understand and answer • Examples: • Summarize this document in one paragraph. • List three benefits of using RPA in financial services. Few-Shot Prompt • Providing a few examples of Input- Output pairs • The model learns the pattern or Style from the given examples • Examples: • Input: "Hello" → Output: "Hi there! How can I assist you today?“ • Input: "Write a formal thank-you note for a client meeting." Output: "Dear [Client], thank you for your valuable time today. We appreciate your insights and look forward to our collaboration." Chain-of-Thought Prompt • Encourage the model to show intermediate reasoning steps before giving the final answer • You instruct the model to think “Step by Step” • Examples: • "Provide the final decision and a short risk assessment (3 bullets). Then add a 3-step rationale summarizing how the risks and benefits were compared to arrive at the decision. Scenario: <insert scenario>.“ • Answer the question concisely. Then provide a numbered summary (3–5 short steps) showing the key reasoning used to reach that answer. Question: <insert question>."
• 9.
  CONTEXT GROUNDING Context Groundingis a component of the UiPath AI Trust Layer which allows you to bring in your data to generate more accurate, reliable GenAI predictions. • Ingestion: Convert business data into representative embeddings using UiPath- managed embedding models. • Embedding: A representation of business data that an LLM can understand and search through. • Search through LLM-ready business data to find the most relevant information. Context Grounding uses a variety of extraction, chunking, retrieval, and re-ranking techniques that are optimised based on different data formats and queries. • Ground and update prompts with the most relevant information from the semantic similarity search results, then execute a generation via an LLM hosted through the LLM Gateway of the AI Trust Layer.
• 10.
  Manual Typical AccountsPayable Process Heavily manual processes that is reliant on Humans to perform all the steps
• 11.
  Automation &AI evolutionfor Accounts Payable Automation & Intelligent Automation with UiPath(Act One)
• 12.
  Automation &AI evolutionfor Accounts Payable Agentic Automation with UiPath(Act Two)
• 13.
  UseCases – Whento Use Agent / Workflow Help Desk Automation Sales email generation Invoice Processing Employee Onboarding Agent Agent Depends, Agent / RPA Workflow RPA Workflow
• 14.
  DEMO