Glossary
Every platform uses slightly different words for the same underlying ideas. This glossary maps common APA terms to the equivalent name on each of the three platforms covered here.
The Terminology Rosetta Stone is a searchable version of the table below - useful when you're mid-project and just need to look one term up.
| General term | Automation Anywhere | UiPath | Copilot Studio |
|---|---|---|---|
| The AI agent itself | AI Agent | Agent | Agent (Copilot) |
| Reasoning/planning engine | Process Reasoning Engine (PRE) | Agent reasoning loop (LLM-driven) | Generative orchestration |
| Reusable capability the agent can perform | Skill | Tool | Action |
| Grounding the agent in your business data | Knowledge/data connections | Context Grounding index | Knowledge source |
| Human-in-the-loop step | Human-in-the-loop | Escalation (Action Center) | Human handoff / escalation |
| Rules that constrain agent behavior | AI Guardrails | Guardrails / policies | Instructions + moderation settings |
| Where agents are built | AI Agent Studio | Agent Builder (Studio Web) | Copilot Studio (web app) |
| Where agents run/orchestrate at scale | Automation Co-Pilot / Control Room | Maestro | Copilot Studio orchestrator / Agent 365 |
| A pre-scripted, rule-based bot (the "old world") | RPA Bot | Robot / Process (Studio) | Power Automate flow |
Core APA vocabulary
Agent — software that pursues a goal by reasoning about the next best action, rather than following a fixed script.
Tool / Skill / Action — a discrete capability an agent can invoke: calling an API, querying a database, running an existing bot, or asking a question.
Context / Grounding / Knowledge — the business data an agent is allowed to read in order to make informed decisions (documents, databases, knowledge bases).
Orchestration — coordinating multiple agents, tools, and workflows so they work together toward a larger business process.
Escalation / Human-in-the-loop — a defined point where the agent pauses and asks a person to approve, correct, or decide, instead of acting autonomously.
Guardrails — policies that constrain what an agent is allowed to do, to keep it safe, compliant, and predictable.