Agentic Automation Journey Series Day 2 – Prompt Engineering for UiPath Agents
- 1. Agentic Automation Journey Series Day 2 – Prompt Engineering for UiPath Agents Lakshman Ganta RPA Tech Lead, 4 x UiPath MVP
- 2. Lakshman Ganta RPA Tech Lead Broadridge Financial Solutions
- 3. 1 2 3 4 Agenda The Power of prompt engineering Best practices for creating effective prompts Agentic AI Prompt Engineering Agent prompt optimization 4 5 Quick Demo
- 4. What is LLM? An LLM (Large Language Model) is a super-smart computer brain trained on huge amounts of text—like books, websites, code, emails, etc.—so it can understand and generate human-like language. Famous LLMs: ChatGPT (OpenAI) Gemini (Google) Claude (Anthropic) LLaMA (Meta) UiPath AI Agents (powered behind the scenes by LLMs)
- 5. What is Prompt Engineering? Prompt Engineering is the art and science of talking to AI in the right way to get the best results. It’s about crafting clear, smart, and purposeful instructions (called prompts) so that AI agents like ChatGPT or UiPath’s AI-powered Agents know exactly what to do, how to do it, and in what format. We typically craft prompts for: Content generation Building chatbots Automating tasks Developing intelligent AI agents How you ask determines what you get.
- 10. Prompting best practices •1. Be clear and specific •2. Provide context •3. Encourage step-by-step thinking •4. Use examples •5. Avoiding bias •6. Iterate and refine
- 11. Prompt Enginering Focusing on Agentic AI
- 12. Traditional vs Agentic AI Prompt Engineering Aspect Traditional Prompt Engineering Agentic AI Prompt Engineering Definition Designing prompts to guide a static LLM response. Designing prompts and tools for AI agents that act, reason, and iterate. Focus One-shot or few-shot outputs. Multi-step task completion and autonomous decision-making. Execution Flow Linear and manual. Dynamic, with iterative reasoning and tool usage. User Role User controls the logic and flow. AI agents take initiative within a defined framework. Prompt Type Static text prompts. Dynamic system + tool + user + memory prompts. Context Handling Limited context window; usually one interaction. Maintains memory across steps, leveraging state and feedback. Tools Integration Rare or manual. Integrated APIs, databases, web browsing, code execution, etc. Typical Use Cases Text summarization, Q&A, classification. Workflow automation, research agents, coding assistants, data extraction. Prompt Optimization Manual testing and refining of individual prompts. Systematic testing across agent behaviors, tools, and task branches. Error Handling Errors require re-prompting or human correction. Agents can detect, reason about, and recover from errors. Examples “Summarize this article.” “Research the latest trends, extract insights, and generate a report.”
- 13. Elements of an AI agent prompt A high-performing AI agent relies on a prompt that clearly outlines its objectives, structures user inputs and context effectively, and provides guidance on when to invoke tools, retrieve enterprise data, or escalate to a human when necessary. Key components of a well-structured AI agent prompt: Clear role definition Explicit task breakdown Methodology of reasoning Output formatting requirement Contextual background information Tool usage instructions Fallback and escalation protocols
- 15. Prompts & Demo
- 16. Write effective prompts Example 1: Bad prompt: Write a follow-up email to the client I met this morning. Good prompt: Write a follow-up email to the client I met with this morning, thanking them for their time and summarizing the three key points we discussed: project timeline, budget approval, and next steps. Keep the tone professional and appreciative, and include a sentence inviting them to reach out with any further questions. Example 2: Bad prompt: Summarize this weekly project update. Good prompt: Summarize this weekly project update in 3–5 bullet points for the executive team. Focus on progress, blockers, and next milestones. Keep the language business-focused and avoid technical jargon.
- 17. Agent prompt optimization Prompt engineering is a dynamic, iterative craft requiring constant experimentation and fine-tuning. A major challenge is the lack of robust feedback loops, making optimization difficult. UiPath solves this with a powerful prompt optimization solution, enabling smart testing, adaptive refinement, and the creation of reliable, efficient, and high- performing prompts for AI agents.
- 18. Agent prompt optimization Prompt engineering is a dynamic, iterative craft requiring constant experimentation and fine-tuning. A major challenge is the lack of robust feedback loops, making optimization difficult. UiPath solves this with a powerful prompt optimization solution, enabling smart testing, adaptive refinement, and the creation of reliable, efficient, and high- performing prompts for AI agents.
- 19. Vibrant ecosystem of more than 3 million professionals and citizen developers learning, getting support, and succeeding together in their automation careers. Academy • Get crowdsourced support and share product feedback on UiPath Forum • Check the product documentation Forum Community Events Join the UiPath Community • Connect with like-minded people and share best practices with the UiPath Community • Solve challenges in engaging hackathon competitions • Join meetups and conferences • Learn the skills of the future on UiPath Academy or through our Academic Alliance • Earn globally recognized credentials with UiPath Certifications
- 20. Questions? Ask your Questions in chat or speak up!