# AI agent (agentic AI)

> Software that uses an AI model and tools to pursue a goal across several steps, with defined limits on what it may do.

## What it means

Plain-English definition: software that uses a language model to pursue a goal over several steps, choosing actions and tools, as against a chat that answers once; ‘agentic AI vs generative AI’ in one clear paragraph (searchers ask exactly this). Where autonomy should stop and a person approves (link human-in-the-loop). Workplace examples: keeping a running summary of meeting notes; researching and filling a table.

## A practical example

A desk agent reads a project table, drafts a weekly summary and proposes status changes for a person to approve.

## Put it into practice

- Name the concrete situation before applying the label. AI agent (agentic AI) is useful only when it changes a choice, boundary or next action.
- Make the state visible: write the owner, time, source or stopping condition down instead of relying on memory.
- Review what happened. Keep the part that reduced ambiguity and change the part that added ceremony without helping the work.

## How it relates to Plexii

the Desk agent widget — a role, an instruction, widgets wired in as inputs, run by hand, when inputs change or on a timer, results and proposed changes arriving as cards (/help/widget-desk-agent); the Automations panel shows what runs on its own on a desk (/help/canvas-automations).