# Set your AI agent's rules, goals and handover

> Use AI agent › Agent and rules to decide what each conversation should achieve, the rules the agent cannot break, and when it hands the customer to a person.

Source: https://www.linkubit.com/docs/agent-rules-and-goals  
Section: Linkubit help centre › The AI agent  
Updated: 2026-10-09

## In short

In AI agent › Agent and rules, Conversation goals say what to find out (by default: what they need, timeline, budget); Agent rules hold it back (block an action until a field is captured, send a topic to a person); Human escalation decides when a person takes over. Rules are always enforced, whatever the AI decides.

## Where

**AI agent › Agent and rules** (the screen is titled **AI settings**) — "What the agent may do, the rules it follows, and what each conversation should achieve." Five tabs: **General** (the AI pause, the hours the AI works, the reply target, how new conversations start, the holding message, lead qualification and who gets follow-ups), **Agent roles**, **Agent rules**, **Conversation goals**, **Human escalation**.

## Conversation goals

A goal is what the agent works towards: a **Stance** for how to behave, and the **Required fields** it finds out before the goal completes — see [Conversation goals and stance](https://www.linkubit.com/docs/conversation-goals-and-stance). If setup wrote no goal, the workspace starts with **Qualify the enquiry**: what they need, timeline and budget. Add the questions your sales team always asks — location, quantity, date of the event.

## Agent rules

On the **Agent rules** tab, **New agent rule** adds a rule, and each has one of three effects:

- **Block tool until fields present** — the agent may not use a capability until details are captured: "No meeting is written before a work email is captured".
- **Require human before tool** — the conversation is handed to a person before the agent may do it.
- **Tell the agent to always do something** — guidance it follows on every reply: "Mention free cancellation up to 48 hours before". Guidance only; the other two are enforced.

Topics that always go to a person — "a cancellation or a complaint" — are added on the **Human escalation** tab.

These are checked by Linkubit before anything is sent, not left to the AI's judgement — see [Rules an AI agent cannot break](https://www.linkubit.com/guides/rules-an-ai-agent-cannot-break).

> **How these settings work together.** Goals, fields, rules, escalation, scoring and workflows act on each other — a score band can qualify a lead, and a workflow on that can hand the conversation over. [How the AI agent works](https://www.linkubit.com/docs/how-the-ai-agent-works) explains each part and the chain between them, and [Why did the agent hand over or stop replying?](https://www.linkubit.com/docs/why-the-agent-handed-over-or-stopped-replying) is the diagnosis table.

## Human escalation

The agent always hands over when a customer sounds frustrated or upset — that check needs no rule and cannot be turned off. Add your own triggers on this tab. A person can also take over any conversation from the inbox: see [Take over a conversation from the agent](https://www.linkubit.com/docs/take-over-a-conversation).

## Lead qualification

With **Lead qualification: AI judges the class** on, the agent grades each lead high, medium or low from the conversation. **Follow-ups go to** sets who gets the follow-up tasks (default: the workspace administrator).

## Related
- [How the AI agent works](https://www.linkubit.com/docs/how-the-ai-agent-works)
- [Conversation goals and stance](https://www.linkubit.com/docs/conversation-goals-and-stance)
- [Human escalation: every trigger](https://www.linkubit.com/docs/human-escalation-explained)
- [Teach your AI agent your business](https://www.linkubit.com/docs/teach-your-agent)
