> For the complete documentation index, see [llms.txt](https://help.sipgate.de/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.sipgate.de/ai-agents/en/behavior/tipps-and-tricks-fur-ihren-agent.md).

# Tips & tricks for your agent

An agent is quick to set up and handles calls reliably. This article shows how to **optimize in a targeted way** and test the agent with **realistic tests** check.

### Optimize call handling

<details>

<summary><strong>Keep the greeting short</strong></summary>

A good greeting is clear, direct, and phrased naturally. The agent should briefly introduce itself and explain how it can help.

**Example:** “Hello, this is Lisa, the digital assistant from sipgate sona. How can I help you?”

Avoid:

* long greeting phrases
* multiple pieces of information at once
* unnecessary introductions

{% hint style="info" %}
Make sure to phrase the greeting in spoken language. The text will be read aloud exactly as it is written here.
{% endhint %}

</details>

<details>

<summary><strong>One question at a time</strong></summary>

Several questions at once make it harder to get the conversation started. Each question should refer to only one piece of information

Instead of: **“What is your name and what is this about?”**\
Better: **“What is your name?”, “What is this about?”**

</details>

### Run tests

<details>

<summary><strong>Test different conversation openers</strong></summary>

Callers phrase their concerns differently.

**Test:**

* direct concerns
* short answers
* unclear phrasing
* uncertain conversation openers

**Check:**

* does the agent reliably recognize the concern?
* does the conversation remain understandable?
* does the agent ask sensible follow-up questions?

</details>

<details>

<summary><strong>Test different wording</strong></summary>

The same concern can be described in different ways.

**Test:**

* Synonyms
* colloquial language
* alternative terms
* incomplete statements

The agent should react as consistently as possible.

</details>

<details>

<summary><strong>Simulate real conversation situations</strong></summary>

Intentionally include typical disruptions in tests.

**For example:**

* background noise
* interruptions
* rapid speech

**Check:**

* does the agent remain understandable?
* does it respond in a comprehensible way?
* does the conversation flow remain stable?

</details>

<details>

<summary><strong>Review conversations afterward</strong></summary>

Then analyze test calls in the transcripts.

**Pay attention to:**

* Are the answers understandable?
* Are the answers too long?
* Does the agent repeat itself unnecessarily?
* Is the conversation flow logical?

Long or unclear answers can often be improved with shorter customer questions and more precise content in customer questions or the knowledge base. Check whether information is phrased redundantly, too generally, or inconsistently.

</details>

<details>

<summary><strong>Test again after changes</strong></summary>

After changes to knowledge, scenarios, or call handling, short test calls should be made.

Proven effective are **3–5 test calls** after each major change.

</details>

### Check transfers & limits

<details>

<summary><strong>Test transfers</strong></summary>

Consciously test transfers several times and in different situations.

**Check:**

* is the transition announced clearly?
* does the transfer work reliably?
* does the handoff also work outside business hours?

</details>

<details>

<summary><strong>Test the agent's limits</strong></summary>

An agent does not solve every issue. What matters, however, is how it handles unclear or unsupported situations.

**Check:**

* does the agent communicate its limits clearly?
* does it ask sensible follow-up questions?
* does it offer suitable alternatives or transfers?

</details>

<details>

<summary><strong>Conduct external tests</strong></summary>

Tests by colleagues or external people often provide more realistic feedback than your own runs.

**Helpful questions:**

* What was unclear?
* Where did the conversation get stuck?
* Which answers seemed unnatural?
* What worked well?

</details>


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://help.sipgate.de/ai-agents/en/behavior/tipps-and-tricks-fur-ihren-agent.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
