Help Center
DocsAI Agent Steps
Hand a step to an AI agent that decides what to do inside the limits you set: the actions it may take, how much it may do on its own, and your plan's AI allowance.
Most steps do exactly one fixed thing. An AI Agent step is different: it hands the record to an agent that looks at the context and decides which of its allowed actions to take. You do not script the decision; you set the boundaries, and the agent works inside them.
Use an agent step when the right move depends on judgement a fixed rule cannot capture, like drafting a reply that fits the thread or deciding which of several follow-ups suits a given deal.
| Field | Description |
|---|---|
| Name* | A label for the agent step so its purpose is clear in the flow and the run history. |
| Allowed actions | The exact set of CRM actions this agent may take, split into Find data and Take action. The agent can never act outside this list. |
| How much it may do on its own | One of three levels, deciding how much runs without asking. See the levels below. |
| Per-action permission | On each action that changes something: Allow, Ask first or Never. Reads are never permissioned. |
| AI allowance | Automatic. Agent usage counts against your plan’s AI allowance, and an agent that reaches its share pauses and resumes on its own. Nothing to configure. |
| Field | Description | Default |
|---|---|---|
| Ask before acting | Looking things up and drafting. Nothing is saved, sent or changed. Your first weeks with an agent, and anything a customer will see. | Asks first |
| Act on routine work | Everyday updates you can undo, like logging a next step or tidying a record. Work you would approve every time anyway. | Routine work |
| Act automatically | Everything you have allowed. You read what it did in the run history. Summaries and internal briefs you read after the fact. | Automatic |
Always asks, whatever you choose: deleting records · sending anything to a customer · taking or refunding a payment · anything that cannot be undone.
That line holds at every level and is not a setting. Each action that changes something also carries its own permission, and when the two disagree the order is fixed: Never wins, then the always-asks floor, then Ask first, then Allow, and only then the level. There is no score and no threshold anywhere in that decision.
A preset agent that drafts a reply to the mail worth answering, so you only edit and send.
The Inbox Assistant is the preset agent that watches your inbox. When mail arrives, it decides which messages deserve a reply and prepares one draft for each, grounded in the message and in the linked record where there is one. When you open the email the draft is waiting in the reply box: edit anything you like, then send it when you are happy. Newsletters, no-reply senders and automated mail are filtered out before a run even starts.
Its Which emails this agent reads setting decides which mail it reads: All emails, Only people and companies in your CRM, Only cold outreach replies, or Choose to name the mailboxes yourself.
You manage it from the Agents home like any other preset agent: pause or resume it, and open its run history to see every draft it produced and what happened to it. Drafts are only visible to the mailbox owner.
Because an agent step uses your AI allowance, it is often worth placing a Filter or If / Else before one so the agent only runs on records that matter. The agent step itself can also carry a condition the record must meet first, which keeps the allowance pointed at the work you care about.