AI Agents
Autonomous AI workers that execute tasks, manage goals, and run workflows.
AI Agents are autonomous workers that can be assigned to your tasks, goals, and workflows. Each agent runs in its own conversation thread and uses tools to complete work on your behalf.
How Agents Work
When you assign an agent to a task, it:
- Analyzes the task context, description, and any related data
- Plans a multi-step approach
- Proposes actions for your approval (for sensitive operations)
- Executes approved actions using Dial8's built-in tools
- Reports results when complete
Agent Lifecycle
| Status | Description |
|---|---|
| Pending | Created but not yet started |
| Analyzing | Examining the task context |
| Working | Actively executing |
| Awaiting Approval | Proposed an action that needs your review |
| Paused | User-paused |
| Completed | Task finished |
| Error | Something went wrong |
What Agents Can Work On
- Actions — Execute specific tasks (draft emails, research topics, etc.)
- Goals — Make progress toward a goal
- Milestones — Work through milestone requirements
- Workflows — Run automated workflow steps
- Notes — Process and enhance notes
Human-in-the-Loop Approval
Agents don't act blindly. For important operations, they propose actions and wait for your approval:
- Drafts — Content drafts for your review before sending
- Plans — Multi-step execution plans before starting work
- Tool calls — Specific actions like sending emails or modifying data
- Questions — Clarifications when they need more information
Sensitive Actions
These actions always require explicit approval:
- Sending emails
- Deleting data
- Making external API calls
- Creating calendar invites
- Modifying contacts
Agent Workspace
The Agent panel in the dashboard shows:
- Left side — List of active agents with status indicators
- Right side — Selected agent's conversation thread
- Header — Pending proposals count, active agents, and controls
Auto-Spawn
Enable "Auto-spawn agents for new actions" to automatically create an agent whenever you add a new task. The agent will begin analyzing and working on the task immediately.
Agents share memory with the voice chat system, so context from voice conversations carries over to agent execution and vice versa.