UIAAAV1: UiPath Agentic Automation Associate ExamDemo

Input properties are values derived from the agent context and presented to the user for a particular agent run. They cannot be changed during review; Input/Output properties are used when the reviewer must modify values, and Output properties hold additional values entered by the user and returned to the agent.

A clear, descriptive lowercase snake_case name makes a tool’s purpose easy to identify and follows a common function-naming convention. pdf_summary specifically conveys that the tool produces a PDF summary; file_processor is overly broad. Official Gemini function-calling documentation permits underscores or camelCase for function names, and its examples commonly use descriptive underscore-separated names.

The feature starts a UiPath agent and waits for that agent to finish execution before the process proceeds. UiPath documents external agents as a separate action, “Start and wait for external agent.”

Zero-shot prompting relies on a model’s existing pretrained capabilities to perform a requested task without task-specific examples or prior training for that particular request.

Traditional automation is designed for predictable, rules-based processes and therefore has limited capability to automate unstructured work that depends on judgment and contextual awareness.

Agentic automation can adapt to uncertainty by making probabilistic decisions from learned patterns and real-time data. This enables behavior to change as conditions and available information change.

Publishing makes the agentic process available in UiPath Orchestrator as a process that can be deployed and run by an automation agent or robot. Debugging is a design-time activity, while sharing and scheduling are separate configuration capabilities.

A deterministic string-check evaluation compares the extracted invoice_date value directly with the expected ground-truth value, 2024-09-01, using exact equality.

UiPath external agents enable coded or low-code agents that were built on other platforms to be incorporated into UiPath, supporting end-to-end, cross-platform business-process orchestration through UiPath Maestro.

LLM-as-a-Judge can evaluate open-ended responses using contextual criteria such as answer quality, semantic correctness, and reasoning. Exact-match, JSON-similarity, and deterministic evaluators are better suited to outputs with fixed expected text, structure, or rules.

The threshold is the minimum relevance or similarity score a retrieved chunk must satisfy before it is included in the agent’s context. A higher threshold excludes more low-scoring chunks, yielding more precise retrieval results. Microsoft’s file-search API documentation defines the score threshold as a value from 0 through 1.

Agentic automation coordinates interdependent work by orchestrating AI agents, robots, and human participants within one unified automation ecosystem.

The Argument input method populates the field from an agent input argument that was defined in advance, enabling a workflow to pass a runtime value into the agent.

Few-shot classification is most effective when examples explicitly map an input to its intended class label in a consistent format. A labeled email-to-category demonstration teaches the agent the expected classification behavior.

Clipboard AI is intended for a single AI-powered paste of a table on a Windows machine, such as moving a table from a PDF into another application directly through the chat interface. It does not provide unrestricted cross-platform pasting or automatic queuing of multiple pastes.

A BPMN Service Task is identified by the cog/gear marker and represents an automated activity performed by a service or application, such as moving an email into a folder.

A UiPath process added as an agent tool needs a description in the system prompt that explains its function and the circumstances in which it should be invoked. That guidance enables appropriate tool selection by the agent.

A current-state (“as-is”) analysis establishes how work flows, who performs it, which systems or teams it depends on, and where pain points occur. That information is necessary to identify appropriate and feasible areas for automation.

Agentic orchestration promotes reliable, repeatable behavior by using defined process models, business rules, and guardrails to constrain and coordinate AI-agent actions. BPMN is a standard notation for modeling such business processes and decision flows.

Indexes provide agents with access to permissioned knowledge bases, enabling them to retrieve and reason over business-specific data as context. UiPath Agents — Context

Agent Memory key-value pairs have a default retention period of 12 months, after which they expire unless their retention is otherwise managed.

A system prompt establishes an AI agent’s assigned role, objectives, operational constraints, and behavioral rules. It provides the foundational direction that governs how the agent handles subsequent user requests.

Few-shot prompts should clearly state the task and provide structured, representative input-output examples. The instruction establishes the intended behavior, and the examples demonstrate how customer queries should be mapped to answers.

A robust UiPath agent prompt defines the agent’s role and constraints, provides an ordered workflow with conditions for tool use and human escalation, specifies output formatting, and references runtime input arguments using the exact {{argumentName}} syntax. The complete invoice-only guardrail, clear approval/escalation threshold, required tool action, constrained tagged output, and {{Invoice_ID}} reference satisfy those requirements together.