> ## Documentation Index
> Fetch the complete documentation index at: https://help.meet-ariel.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Agents

> What Ariel agents are and how they work

<Badge color="blue">Agent plan</Badge>

Ariel agents are specialized, ever‑learning helpers that handle recurring tasks for you — from scheduling and follow‑ups to research and prep. You set them up once; they observe your patterns and take over with approvals and guardrails you control.

## Why agents

* Reduce busywork while keeping your voice and preferences intact
* Act across channels (phone, messages, email, calendar) in one coordinated flow
* Get smarter with every task, adapting to context, contacts, and timing
* Perfect for recurring use cases (e.g., scheduling, follow‑ups, triage)

<Info>
  Agents feel less like automation and more like a thoughtful teammate — anticipating needs,
  handling details, and returning clear outcomes.
</Info>

## What agents can do

<CardGroup cols="3">
  <Card title="Scheduling" icon="calendar">
    Negotiate times, manage invites, handle reschedules, and resolve conflicts across calendars.
  </Card>

  <Card title="Follow‑ups" icon="inbox">
    Track open threads and nudge at the right moment — so nothing goes cold.
  </Card>

  <Card title="Research & Prep" icon="magnifying-glass">
    Pull context, summarize threads, and prepare briefs before meetings.
  </Card>
</CardGroup>

<CardGroup cols="3">
  <Card title="Email Triage" icon="envelope">
    Label, prioritize, and draft replies in your tone for one‑tap approval.
  </Card>

  <Card title="Contacts" icon="address-book">
    Recognize VIPs, relationships, and history to route and personalize actions.
  </Card>

  <Card title="Calls & Messages" icon="phone">
    Answer, qualify, and capture next steps with clear, human responses.
  </Card>
</CardGroup>

## How agents work

<Steps>
  <Step title="Define the job">
    Choose an agent type (e.g., Scheduling, Follow‑ups) and set preferences like tone, working hours, and approval rules.
  </Step>

  <Step title="Observe your patterns">
    Agents learn from your past actions, contacts, and calendar behavior to mirror your style and priorities.
  </Step>

  <Step title="Learn from previous runs">
    After each run, agents incorporate feedback from your approvals, edits, and outcomes to refine prompts, rules, and timing — steadily improving results over time.
  </Step>

  <Step title="Act with approvals">
    Start with safe‑draft workflows: agents propose actions and you approve with one tap from anywhere.
  </Step>

  <Step title="Autonomy with guardrails">
    As trust grows, allow fully autonomous actions for specific scenarios while keeping clear audit trails.
  </Step>
</Steps>

<Info>
  To get great outcomes, give agents detailed instructions. Specify the goal, constraints,
  preferences (tone, contacts, timing), and examples of good responses. Clear guidance helps
  agents learn faster from previous runs and perform with confidence.
</Info>

## Examples

* A Scheduling agent proposes three time windows, sends holds, and confirms the meeting — respecting time zones and buffers.
* A Follow‑up agent tracks proposals and pings after 3 days with a friendly nudge in your tone.
* An Email Triage agent labels VIP threads as Urgent, summarizes long chains, and drafts replies for approval.