> ## Documentation Index > Fetch the complete documentation index at: https://docs.bizzyco.ai/llms.txt > Use this file to discover all available pages before exploring further. # Automations > How AI-powered automations work in Bizzy Automations are how Bizzy turns your inbox from a task list into a system that works for you. Instead of manually processing every message, you define rules in plain English and let AI handle the routine work. ## What Are Automations? An **automation** is a workflow that runs when something happens in Bizzy. Every automation has three parts: 1. **Trigger** — the event that starts the automation (e.g., a new email arrives) 2. **Instructions** — what you want the AI to do, written in natural language 3. **Permissions** — what actions the automation is allowed to take What makes Bizzy automations different from traditional "if this, then that" rules is the **AI layer**. Instead of rigid conditions and fixed responses, you describe what you want in plain English and the AI interprets each situation individually. ## Automations vs. Agents Bizzy has two kinds of AI worker, and they're easy to confuse. **Automations** are event-triggered: a message arrives, a contact is created, a schedule fires, and the automation runs without anyone being there. **Agents** are conversational: you (or another workflow) chat with them, and they call tools to get things done across multiple turns. A useful intuition: an automation is a junior employee waiting for the inbox to ping. An agent is a colleague you can ask things of. Automations *can* invoke an agent as one of their steps when the work needs more nuance than rules can capture — for example, "when a new email arrives, ask the support agent to draft a reply." If you'd ever want to hit "send" yourself before the action happens, reach for an [agent](/get-started/concepts/agents). If the action should *always* happen the moment the trigger fires, reach for an automation. ## How Automations Work When an event occurs — say a new email arrives — Bizzy checks if any active automations match that trigger. If so, the AI reads the incoming content, evaluates your instructions, and takes action. Here's the flow: 1. **Event occurs** — A new email, contact creation, or other trigger event happens 2. **Trigger matches** — Bizzy identifies every automation listening for this event; all matching enabled automations run independently 3. **AI processes** — The AI reads the content and interprets your instructions 4. **Action taken** — The AI performs the appropriate action (reply, create contact, etc.) 5. **Result logged** — The execution is recorded so you can review what happened The AI doesn't follow a decision tree — it understands context. An instruction like "if this looks like a support request, send an acknowledgment" works because the AI can evaluate whether an email is a support request, even if the sender doesn't use the word "support." ## The Power of Natural Language Instructions Traditional automation tools require you to define exact conditions: "if subject contains 'help' OR subject contains 'support' OR subject contains 'issue'..." This is brittle and misses edge cases. Bizzy automations use AI instructions instead: ``` When a new email arrives: 1. If it looks like a support request, send a friendly acknowledgment 2. If it's a sales inquiry, tag the contact as "Lead" 3. If it's a newsletter or automated notification, do nothing ``` The AI handles the ambiguity. It understands that "My login isn't working" is a support request even though it doesn't contain the word "support." ## What Can Automations Do? Automations can take a range of actions depending on their permissions: * **Send replies** — Compose and send email responses * **Create and update contacts** — Add new contacts or update existing ones * **Manage messages** — Archive, label, or categorize incoming messages * **Send notifications** — Alert team members about important messages * **Create tasks** — Generate follow-up tasks from incoming requests ## Automation Lifecycle Automations have a lifecycle with different states: | Status | Meaning | | ------------- | ------------------------------------------------- | | **Active** | Running and processing events as they occur | | **Paused** | Temporarily disabled — can be resumed at any time | | **Completed** | Finished after reaching an execution limit | | **Expired** | Past its configured expiration date | You can pause an automation at any time without losing its configuration. This is useful when you want to temporarily stop processing while you refine your instructions. ## Automations and Permissions Automations run with the permissions that were set when they were created. They cannot perform actions beyond their granted permissions. This means: * An automation with read-only message access can analyze emails but not reply * An automation without contact permissions cannot create or modify contacts * Permissions are checked on every execution, not just at creation time This design ensures automations stay within the boundaries you define, even as the AI interprets instructions flexibly. ## Related Topics Step-by-step guide to building an automation Build a working AI-powered support workflow from scratch The conversational counterpart — when to reach for one instead