Agents & Automations
Let the system act when something happens — from fixed steps to an AI agent that reasons over your records.
- GuideAgents and automations from first principlesAn automation watches for a trigger — a record changing, a scheduled time, an incoming message — and runs steps in response: a saved "when this, do that" the system carries out for you. When a response needs judgment rather than a fixed rule, a step hands off to an AI agent that reads the record and decides what to change — which is where a fixed workflow becomes agentic.
- ArticleCreate and activate an automationBuild an automation by choosing its trigger, adding the steps it should run, then testing it before it goes live. A new automation starts inactive on purpose — validate the trigger, do a dry run that changes nothing, and only then activate it.
- ArticleAutomation triggersTriggers come in four families: something happens to a record, a person or system starts it, time arrives, or a message comes in. The setting people miss most is which origins a record trigger listens to — by default it watches changes coming from your website, so edits made by hand in the CRM will not fire it until you say they should.
- ArticleWhat an agent is (and how it differs from an automation)An agent is AI that acts on your records: it reads them, reasons, and changes them. An automation is the deterministic engine underneath, a trigger followed by steps that run the same way every time. The two meet when one of those steps hands control to an AI model, and that is where an automation becomes agentic.
- ArticleAdd an AI step to an automationTo add an AI step, connect an AI provider, add the step to an automation, choose the provider and model, write the prompt with values from the entry, and choose what the step produces: text, an email, or values for fields. Later steps use that output, and a test run shows the prompt without calling the provider.
- ArticleSteps and actions overviewAn automation is built from steps of two kinds: actions that do work, such as sending an email, updating entries, or calling an AI model, and flow steps that shape the run, such as conditions, loops, and delays. Every automation has exactly one starting step, and the connections between steps decide the order.
- ArticleInspect an automation runTo see what an automation did, open its recent runs, pick one, and read each step's status and, for a failed step, its error. A run can look successful and still have done nothing, such as sending zero emails, so read the skip reasons too. Fix the cause, check it with a dry run of the step, then replay the run.