Help Center
DocsAI Agents Overview
What an AI agent is, how agents fit alongside the assistant, automation rules, and support autonomy, and who on your team can use them.
Laureo has four distinct AI surfaces. They look similar but do different jobs.
It helps to know which surface you are using before you reach for it. Each one has its own home in the app and its own way of working.
- The assistant (Laureo chat). You ask, it answers and can perform reviewed work as your signed-in user. Accepted work is durable, so it can continue while you are away and pause only for input, approval, Stop, completion, or an unrecoverable failure. Read AI Assistant Overview.
- Agents. Workers you brief once that then run on their own. They start from a schedule or from something happening, and they use the same reviewed capabilities the assistant does, but only the ones granted to them, only on the records in their scope, and only as far as the level you set lets them go. This is the product these articles cover.
- Automation rules. Fixed, deterministic flows. When this happens, do exactly these steps, the same way every time. No reasoning, instant, and free. Read Automations Overview.
- Support autonomy. The AI support copilot that can draft, approve, and (when you allow it) send replies to customer tickets on its own. Read AI Autonomy Controls.
An agent is given a plain-English brief, an explicit set of reviewed actions, the records it may work on, a level saying how much it may do without asking, and a trigger the server owns. When that trigger fires, the agent can make several bounded hops while keeping its lease and its safety state.
- Read. It looks at the records in its scope using the reads it was granted.
- Decide. It works out what matters and records a safe decision summary, never private reasoning.
- Propose. It writes up each action it wants to take.
- Apply the level. What runs without asking is decided by the level you set and by the decision on that individual action. Anything that leaves the product or cannot be undone waits for a person, at every level.
Always asks, whatever you choose: deleting records · sending anything to a customer · taking or refunding a payment · anything that cannot be undone.
Looking things up never waits, at any level, and a draft is only ever saved as a draft for you to send. For an action that changes something, Ask before acting asks first; Act automatically makes any change that is not on the always-asks list on its own. The always-asks list also holds any change Laureo rates as high or critical risk. A decision you set on one action comes before the level: Allow runs it without asking even on the first level, and Never stops it at every level. No score or threshold decides a change.
One home, opening on the work: what needs you first, then your agents, then what they ran.
The Agents home in the sidebar is one page with three sections under a single header, and it opens on the Inbox: everything your agents have proposed and are waiting on you for. Agents is your roster with the month’s outcomes above it, and Activity is the feed of recent runs. See Runs and Approvals.
Two sentences sit directly under the section control, on every section, because they describe the state of the whole thing rather than of one list. One is your AI credits, Used about 8,000 of 10,000 AI credits this month. The count appears only for people who may see the workspace's AI credits (the billing.view permission, or everyone once member usage visibility is on). The other appears only when work is not being handed out: Agent runs are paused right now. Nothing is lost; they will resume automatically.
You build agents from the same header. New agent opens the agent’s Build page: you write what the agent should do, Set up from instructions fills in when it runs and what it may do, you check what it assumed, try it on a real record and publish. Changes save as you go. Publish to make them live. See Building Your Own Agent.
The assistant and agents use the same versioned capability schemas, executor contracts, leases, events, approvals, effect receipts, idempotency, usage accounting, cancellation, and retry rules. This keeps the meaning of an action consistent across both surfaces.
Authority remains different. The assistant acts as the current user. An agent may use only the actions an administrator explicitly granted it after review, on the records inside its scope, and within the limits the workspace applies to every run. A trigger payload can provide context but cannot become authority, change the agent's scope, or turn a test run into a live one.
Two things gate agents: your plan and your role.
| Field | Description |
|---|---|
| Preset agents | The ready-made agents Laureo installs on the Agents page run on every paid plan. Ticket Triage, which sorts new support tickets inside Service, comes with the Ultra plan. |
| Building your own | Building your own agent is available on the Business and Ultra plans. On lower paid plans you can still run the preset agents and build deterministic automation rules. |
| Support autonomy | The AI autonomy that drafts, resolves, and answers tickets for your own support team is an Ultra-plan feature, configured separately in the Service hub. |
Within your plan, what each teammate can do is set by their role. Every role can see the Agents home and the roster, so anyone on your team can read what the agents are doing. Creating and editing agents is held to admins and the marketing role. Approving what an agent proposes is held to admins. Running an agent now or stopping a run is open to admins, the marketing role, and sales managers for their own team. The full table is in Runs and Approvals.