The home for rules: deterministic automations that run the same fixed steps every time a trigger fires.
The place to create, run, and watch your rules.
An automation is a saved rule that starts when a trigger fires and then runs a fixed flow of steps against a record. The hub at Automations in the sidebar lists every rule you have, with its status, last run, runs in the last 7 days, and its goal all in one row.
AI agents are a separate feature with their own home. The Agents page in the sidebar is where you create and supervise agents; the Automations hub is where you build rules. The Automations hub carries a slim link to the Agents home in case you came to the wrong one.
Rules live here; AI agents have their own home. Here is how the two differ.
| Rule | AI Agent | |
|---|---|---|
| How it decides | Deterministic: same input, same result | Reasons within the approvals and limits you set |
| Cost | Free, no model spend | Metered against your plan’s AI allowance |
| Speed | Runs the moment the trigger fires | Runs a model, then acts |
| Oversight | Every step is auditable | Approval-gated by default |
| Best for | Fixed if-this-then-that work | Judgement calls across several signals |
A Rule runs exactly the steps you laid out, the same way every time. Use it when the right action is obvious: when a deal is won, create a follow-up task; when a form is submitted, assign the lead round-robin. Everything on this hub, and the rest of this section, is about rules.
An AI Agent is a worker you brief in plain English that reads your CRM, reasons, and proposes actions for your approval. Use one when the call depends on context a fixed rule cannot capture, like drafting a tailored reply or triaging which deals look at risk. Agents are created and supervised from the Agents home, not here.
A rule stays deterministic end to end: every step in its flow is a fixed action, never a judgement call. When a step of the work needs judgement, create an agent from the Agents home instead and keep the rule for the fixed part. How agents behave when they run, including approvals and limits, cost, and confidence, is covered in AI Agent Steps below.
Every automation moves through the same statuses.