AA Advanced RPA key terms

A bot that runs on a user's workstation and is triggered by that user in real time, typically handling small tasks while the person waits or works alongside it.

A bot that runs on a server or virtual machine with no human present, usually started by a scheduler or queue and operating on its own through the whole task.

A design pattern where a bot pauses and routes a decision, approval, or exception to a person before continuing, combining automation speed with human judgment.

A technique that analyzes system event logs to reconstruct how a process actually runs, revealing bottlenecks, rework, and the steps worth automating.

Recording and analyzing desktop user actions to map detailed step-level work, useful for finding small high-frequency tasks suitable for attended bots.

The structured effort to document a candidate process, its steps, systems, inputs, and exceptions before deciding whether and how to automate it.

Technology that extracts structured data from scanned or unstructured documents using OCR plus machine learning, feeding clean fields into downstream automation.

Conversion of images of text, such as scans or photos, into machine-readable characters so software can read and act on the content.

A statistical component trained on example data to make predictions or classifications, such as sorting emails or judging document type, that a bot can call.

A cutoff score above which a model's output is accepted automatically and below which the item is sent to a person or exception path for review.

A field of AI concerned with software understanding and generating human language, used in automation to read messages, classify intent, and extract entities.

Models that produce new text, images, or code from prompts, increasingly used inside automation for drafting content or summarizing records, requiring output validation.

A constraint placed around an AI or automation component, such as allowed actions, output checks, or blocked topics, to keep behavior within safe and approved limits.

An attribute or path a bot uses to locate a specific user-interface element, such as a button or field, so it can click or type in the right place.

A selector built partly at runtime from variables or wildcards so it still matches elements whose identifiers change between sessions or records.

Driving an application through its user interface, simulating clicks and keystrokes, which works broadly but is more fragile than calling an API.

Connecting a bot directly to a system's programmatic interface to read or write data, generally faster and more stable than screen-level interaction.

A prebuilt or custom component that handles authentication and data exchange with a specific external system, sparing the bot from low-level integration work.

A holding structure where work items wait to be processed, letting bots pull tasks at their own pace and enabling parallel processing across many robots.

A single unit of work in a queue, carrying its data and status, which a bot processes independently so one failure need not stop the batch.

The designed logic for detecting errors during a run, deciding whether to retry, skip, or escalate, and recording the outcome for follow-up.

A failure caused by invalid data or a rule violation rather than a technical fault, which is typically routed to a person instead of being retried automatically.

Rules defining how many times and how often a failed step or transaction is attempted again before being marked as failed for good.

A reusable project template that standardizes logging, error handling, configuration, and queue processing so every bot in an estate is built the same way.

A packaged piece of automation logic, such as a login routine or data formatter, that multiple processes can call instead of rebuilding the same steps.

A published collection of reusable activities or components that projects can install and invoke, supporting consistent behavior and easier maintenance across teams.

An external file holding settings such as credentials references, URLs, and thresholds, so a bot can change environments without editing its code.

A secured store that holds passwords and keys, letting bots retrieve secrets at runtime instead of embedding them in project files or logs.

The central management platform that deploys, schedules, and monitors bots, distributes queue items, and stores logs and configuration across the estate.

The stages a bot passes through, from requirements and build to testing, deployment, production monitoring, maintenance, and eventual retirement.

A tamper-evident record of actions taken by bots and users, used to trace what happened, support investigations, and satisfy compliance requirements.

A control that prevents one person from both building and approving the same change, reducing the risk of undetected error or misuse in the bot estate.

The governed process for requesting, reviewing, testing, and approving modifications to a bot or its environment before they reach production.

A central team that sets standards, provides shared tooling and expertise, and governs automation demand so efforts stay consistent as the program grows.

The managed backlog of assessed automation opportunities, ranked by value and feasibility, that feeds delivery teams in a steady prioritized flow.

Scoring candidate processes on criteria such as volume, rule stability, error rate, and system access to choose which ones to automate first.

A business-side employee who builds small automations with governed low-code tools, supported by the CoE with standards, training, and review.

Software that mimics user actions across applications to execute repetitive rule-based work, connecting systems without changing their underlying code.