> For the complete documentation index, see [llms.txt](https://docs.pascom.net/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.pascom.net/en/setup/acd/ai-agents-pia-setup.md).

# Pia AI Agent Setup

This guide describes how to set up and configure the interactive Pia AI agent in your PASCOM phone system.

The **Pia** AI agent provides an interactive voice assistant. Pia answers incoming calls, analyzes the caller's request, prompts for required information, and routes the caller to the correct team. You configure the agent's behavior and dialog using prompts (text instructions).

{% hint style="info" %}
**Free Test Phase until the end of September 2026:** During the test phase, there are no costs for using Pia. To prepare for the subsequent live operation, the user interface already shows references to future per-minute billing.
{% endhint %}

## Create a Pia AI Agent

You can find the AI agent management section in the PASCOM Admin Dashboard.

1. Go to **Call Routing** > **AI Agents** in the menu.
2. Click **Add**.
3. Select the **Pia** type.

![Create Pia](https://2713225-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FVIw0BpSv358R2Pr2HflD%2Fuploads%2Fgit-blob-1be4bbfe7ec91d079a08d06b3eb6e392e8e43a07%2Fpia-create.en.png?alt=media)

## Configure Basic Data

Configure the basic properties of the agent in the **Pia Basic Data** section:

* **Display Name**: Enter a unique name for the agent (e.g., `Reception` or `Support Pre-qualification`).
* **Avatar**: Click **Change Avatar** to assign a visual profile picture to the AI agent.

## Configure the Pia Prompt

Enter the instructions for the AI agent in the **Pia Prompt** text box. These instructions determine how Pia greets callers and manages the conversation.

The following standard template is active by default for every newly created Pia agent:

```markdown
# Greeting

Greet the caller in a friendly manner and introduce yourself as an AI assistant.

Example: "Hello, you are speaking with Pia, the automated AI assistant of [Company Name]. How can I help you?"

# Routing

Determine the caller's request and route to the appropriate team:

- Technical issues or disruptions → Team Support

- Questions about invoices or contracts → Team Administration

- Interest in products or prices → Team Sales

- Anything else → Team Reception
```

![Pia Basic Data](https://2713225-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FVIw0BpSv358R2Pr2HflD%2Fuploads%2Fgit-blob-57f0310d77bdcba8b7dd565bf39acb49de4539c2%2Fpia-basedata.en.png?alt=media)

### Use Inline Tools and Labels

You can use integrated inline tools instead of describing routing and queries purely as text:

* **Routing to Teams**: Enter a slash character (`/`) in the prompt editor to open a menu of available commands. Select `/team` and define the target team (e.g., `/team Support`) to route the caller directly.
* **Collect Customer Information**: You can instruct Pia to query important data and save it in labels. For example: *"Ask for the customer number and save it in the customer number label."* The label must be created beforehand. Learn how this works in the [Configuring Labels](/en/setup/acd/labels.md) guide.

![Pia Inline Tools](https://2713225-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FVIw0BpSv358R2Pr2HflD%2Fuploads%2Fgit-blob-1d1b4526c13900886601a1c55ec57f0caa6aa97c%2Fpia-inlinetools.png?alt=media)

All labels set by Pia during the call display to the employee in the PASCOM app. This ensures the customer does not have to repeat their details.

## Model and Voice Settings

Select the language logic and voice under **Model / Voice (Provider)**.

### GDPR and Hosting

For data privacy and compliance reasons (ISO, GDPR), all models are hosted in European data centers (Microsoft Azure EU). Your data remains fully protected within the EU.

### Model Selection (GPT Realtime)

Various Realtime models are available:

* **Mini**: This model is sufficient for simple tasks such as classic call routing or standard IVR flows, and it operates very quickly.
* **1.5** and **2.0**: These larger models are suitable for highly complex prompts, nested conversation logic, and precise extraction of variable customer data.

### Speaking Speed & Interruption

* **Speaking Speed**: Use the slider to set the speaking speed of the AI voice (default value is `1.0`).
* **Interruption**: Turn on the **Can the caller interrupt the AI agent?** switch if the caller is allowed to interrupt Pia mid-sentence. When turned off, Pia always completes the current prompt section before responding to the caller.

![Pia Settings](https://2713225-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FVIw0BpSv358R2Pr2HflD%2Fuploads%2Fgit-blob-224895f507f4a97aeb0bbc623d089873f5df37a7%2Fpia-gptmodel.en.png?alt=media)

## Configure Call Result Labels

Use the **Label to save the call result** option to select a label that logs the status or result of the AI call. This allows precise tracking of calls in subsequent routing steps.

The system sets the following result values and triggers:

| Call Result Value                                     | Trigger                                                                 |
| ----------------------------------------------------- | ----------------------------------------------------------------------- |
| **Caller was transferred to an internal destination** | The AI successfully transferred the caller to a defined team or user.   |
| **Caller hung up during the conversation**            | The caller ended the call prematurely.                                  |
| **Total or agent CU limit exceeded**                  | The concurrent call limit (CUs) was reached.                            |
| **Daily limit reached**                               | The daily budget limit was reached.                                     |
| **Internal error**                                    | A technical problem occurred (e.g., network error, server unavailable). |

These labels attach to the call and are visible to users in the PASCOM app. This allows employees to immediately see how and if Pia processed the caller.

## Limits and Budget Control

* **Limit Concurrent Calls**: Turn on this switch to limit the maximum parallel lines (CUs) for this agent (e.g., to a maximum of 3 channels). When turned off, the agent uses all available CUs in the package. This protects you from unexpected costs in live operation.

![Pia Settings](https://2713225-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FVIw0BpSv358R2Pr2HflD%2Fuploads%2Fgit-blob-06318dd47a594beb61d28caab159e4fc439324a7%2Fpia-settings.en.png?alt=media)

## Integration in the Call Router (Best Practices)

Integrate Pia effectively into your call routing:

1. **Prepend Fixed Rules**: Filter clear criteria (such as holidays, business hours, or company shutdowns) beforehand using the classic action model in the call router. Do not route calls to Pia when your office is closed.
2. **Set up a Fallback**: Because Pia may not accept calls if CU limits or budget limits are reached, or during network disruptions, you must **always** define a fallback action at the end of the call router (e.g., forward to the regular office or reception). This ensures no caller ends up in a dead end.


---

# 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 dynamically by asking a question.

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

```
GET https://docs.pascom.net/en/setup/acd/ai-agents-pia-setup.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 `build a script that syncs our docs to a CMS` 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.