> For the complete documentation index, see [llms.txt](https://docs.ibexa.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.ibexa.ai/tutorials/building-an-product-faq-assistant.md).

# Building a Product FAQ Assistant

| **Time to complete** | \~20 minutes                                                                                                                   |
| -------------------- | ------------------------------------------------------------------------------------------------------------------------------ |
| **Who this is for**  | Marketing team members, Content Editors, and anyone who wants to give their team instant, accurate answers about your product. |

## What You Will Build

By the end of this tutorial you will have a working **Product FAQ Assistant** - an chat assistant that answers questions about your product using *only* the official documentation you provide. It will not guess, make things up, or pull information from outside your content. Your teammates will be able to open the assistant from anywhere in the platform and ask questions like:

* *"What pricing plans do we offer?"*
* *"Does the product integrate with Salesforce?"*
* *"What file formats can users export to?"* ...and get accurate, sourced answers drawn directly from your FAQ documents.

## Prerequisites

Before you start, make sure the following are in place:

* At least one **AI model** has been connected and enabled. If you are unsure, go to **Organisation → AI Models** and check that a model is available. If not, ask your administrator or see [AI Models](/manual/organisation/ai-models.md).
* The **IAMP Knowledge Base** MCP Server has been added and is working. Go to **Organisation → MCP Servers** and confirm the server appears in the list with a green status indicator. If it is missing, ask your administrator to add it before continuing.
* You have your product FAQ content ready - this can be an existing document, a text file, or content you'll write directly in the platform.

## Overview of the Steps

1. Add your FAQ content to the Knowledge Base
2. Create a new agent and write its instructions
3. Connect the agent to your FAQ documents
4. Test the agent
5. Make it available to your team

## Step 1: Add Your FAQ Content to the Knowledge Base

The Knowledge Base is the document library your assistant will read from. Think of it as the single source of truth you hand to the AI.

### 1.1 - Open the Knowledge Base

Select **Knowledge Base** in the main navigation menu.

### 1.2 - Create a folder for your FAQ documents

Keeping your FAQ content in its own folder makes it easy to point the assistant to exactly the right material later.

1. In the left sidebar, select the **Organisation** section root (the top-level item).
2. Select the **Add** button (top-right of the right panel) and choose **New folder**.
3. Type a clear name, for example **Product FAQ**, and press **Enter** to confirm.

<figure><img src="https://3641047820-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fzi8TZ3KjOFqUxK0zOIrL%2Fuploads%2Fgit-blob-c7f438e836a141d36bd9239ae52bf8125ff2da65%2Fknowledge-base-add-menu.png?alt=media" alt="A Knowledge Base folder with the Add dropdown open, showing New folder, New document and Upload document"><figcaption><p>The Add dropdown</p></figcaption></figure>

### 1.3 - Add your FAQ document

Select the **Product FAQ** folder you just created, then choose one of the following options from the **Add** button:

| Option              | When to use it                                                                              |
| ------------------- | ------------------------------------------------------------------------------------------- |
| **New document**    | Your FAQ content doesn't exist yet, or you want to write/paste it directly in the platform. |
| **Upload document** | You already have a FAQ document on your computer (e.g. a `.pdf`, `.docx`, or `.txt` file).  |

**If you choose New document:**

1. Select **Add → New document**. A blank editor opens.
2. Give the document a clear title at the top, for example: *Product FAQ - General*.
3. Write or paste your FAQ content. Use clear question-and-answer formatting:

```
## Pricing

**Q: What pricing plans are available?**
We offer three plans: Starter, Professional, and Enterprise. ...

**Q: Is there a free trial?**
Yes, all plans include a 14-day free trial with no credit card required.

## Integrations

**Q: Does the product integrate with Salesforce?**
Yes. Native Salesforce integration is available on Professional and Enterprise plans. ...
```

4. Select **Save** when finished.

**If you choose Upload document:**

1. Select **Add → Upload document**.
2. Choose your file and select **Upload**. The document is created as a new **Draft**.

### 1.4 - Publish the document

A document must be **Published** before the assistant can read it.

1. Select your document in the left sidebar to open it.
2. Select the **Versions** tab.
3. Find the **Draft** version and select **Publish**. The status changes to **Published**. Your content is now available to agents. You can add multiple documents to the **Product FAQ** folder - for example, one for Pricing, one for Integrations, one for Technical Specs. The assistant will search across all of them

## Step 2: Create the FAQ Agent

Now you'll build the AI agent that will use the documents you just uploaded.

### 2.1 - Open the Agents page

Select **Agents** in the main navigation menu, then select **Add Agent** in the top-right corner. The **Agents Browser** opens. You can browse catalog templates here or start from scratch. For this tutorial, select **Add Custom Agent** to open a blank **Add Agent** page.

### 2.2 - Write the Instructions

This is the most important step - the **instructions** tell the agent exactly how to behave. You're essentially giving the AI its job description. In the **Instructions** field, enter the following. Feel free to adjust the product name and tone to match your brand:

```
You are a helpful Product FAQ assistant for [Your Company Name].

Your job is to answer questions about our product using only the information provided in the Knowledge Base documents. Do not use any external knowledge or make assumptions beyond what is written in those documents.

Follow these rules strictly:
- Only answer questions that are covered in the Knowledge Base.
- If a question is not answered in the documents, say clearly: "I don't have information about that in our FAQ. Please contact our support team at [support email]."
- Keep your answers concise, friendly, and factual.
- If a question matches multiple topics, answer all relevant parts.
- Do not make up features, prices, or policies that are not in the documents.
```

> **Why this matters:** The instruction "only use the Knowledge Base" is what keeps the assistant from guessing or hallucinating. Without it, the AI might answer based on general internet knowledge rather than your specific product details.

Select **Next** to continue.

### 2.3 - Triggers

Triggers define when and how an agent can be started. Your FAQ assistant needs a **Run Now** trigger so that team members can start a conversation with it on demand.

1. Select **Add trigger**.
2. From the trigger type list, choose **Run Now**.
3. Leave the trigger settings at their defaults. Select **Next**.

### 2.4 - Tools

Tools give the agent the ability to search and read documents from the Knowledge Base. You need to connect the built-in Knowledge Base server and enable two tools.

1. Select **Select tools**.
2. Find **IAMP Knowledge Base** Server (the built-in MCP server).
3. Browse categories and in the tool list that appears, tick the following two tools:
   * **Get Document** - lets the agent retrieve the full content of a specific document.
   * **Search Documents** - lets the agent search across documents to find relevant answers.
4. Confirm your selection and return to the wizard. Select **Next**.

### 2.5 - Knowledge Base

This step grants the agent access to your FAQ documents.

| Field       | What to select                                           |
| ----------- | -------------------------------------------------------- |
| **Access**  | Select **Selected folders**                              |
| **Folders** | Select the **Product FAQ** folder you created in Step 1. |

> **Why "Selected folders" instead of "All"?** Selecting **Selected folders** and pointing to the Product FAQ folder means the assistant will *only* read from your FAQ documents. Selecting **All** would give it access to everything in the Knowledge Base - including internal documents that may not be appropriate to share in a FAQ context.

Select **Next**.

### 2.6 - Quality Metrics

Leave this step empty for now - you can configure quality monitoring later from the agent's **Edit** page. Select **Next**.

### 2.7 - General Properties

Fill in the agent's identity, and pick a model:

| Field           | What to enter                                                                                                                             |
| --------------- | ----------------------------------------------------------------------------------------------------------------------------------------- |
| **Name**        | `Product FAQ Assistant`                                                                                                                   |
| **Description** | `Answers questions about our product using official FAQ documents.`                                                                       |
| **Model**       | Choose the AI model your organisation has configured (e.g. *GPT-4o* via OpenAI). If you are unsure which to pick, ask your administrator. |

Leave the remaining fields (budget, maximum number of steps) at their defaults for now. Select **Submit**. You are taken to the new agent's detail page.

## Step 3: Test the Agent

Before making the assistant available to your whole team, take a moment to verify it behaves as expected.

### 3.1 - Open the test interface

From the agent's **detail page**, select the actions menu (top-right corner) and choose **Run test**. A chat interface opens with a welcome message.

### 3.2 - Ask a few test questions

Try questions that *are* in your FAQ documents, and some that are *not*:

| Test question                                                                          | What to expect                                                                        |
| -------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------- |
| A question covered in your FAQ (e.g. *"Do you offer a free trial?"*)                   | A clear, accurate answer drawn from the document.                                     |
| A question *not* in your FAQ (e.g. *"What's your refund policy?"* - if not documented) | The agent should say it doesn't have that information and direct the user to support. |
| An off-topic question (e.g. *"What's the weather today?"*)                             | The agent should politely decline and stay on topic.                                  |

### 3.3 - Refine if needed

If the agent gives an unexpected answer, the most common fixes are:

* **Strengthen the instructions**: go to the agent's **Edit** mode and make the "only use the Knowledge Base" rule more explicit.
* **Check document status**: go to Knowledge Base and confirm all documents are **Published** (not Draft).
* **Check Knowledge Base Locations** and confirm the correct folder is selected in the agent's Knowledge Base step (edit the agent to change it).

## Keeping the Assistant Up to Date

The assistant is only as good as the documents behind it. Make it a habit to update the Knowledge Base whenever your product changes.

| When this happens…                          | Do this                                                                                              |
| ------------------------------------------- | ---------------------------------------------------------------------------------------------------- |
| A feature changes or a new plan is launched | Edit the relevant document in the Knowledge Base, then **Publish** the new version.                  |
| You want to add a new topic                 | Create a new document in the **Product FAQ** folder and publish it.                                  |
| A document becomes outdated                 | Open the **Versions** tab and **Archive** the old published version, then publish the updated draft. |

> **Note:** Old published versions are archived automatically when you publish a new one - you don't need to archive them manually.

## Summary

You've just built a fully functional Product FAQ Assistant. Here's what you did:

1. ✅ Organised your FAQ content in a dedicated Knowledge Base folder.
2. ✅ Published your documents so the agent can read them.
3. ✅ Created an agent with clear, focused instructions.
4. ✅ Restricted the agent's knowledge to your FAQ folder only.
5. ✅ Tested the agent to verify it answers correctly. Your marketing and content teams can now get instant, accurate product answers without digging through shared drives or waiting for a colleague to reply.

## Next Steps

* **Add more documents** to the Product FAQ folder to expand coverage - pricing tables, integration guides, release notes.
* **Create topic-specific assistants** using the same approach - for example, a Brand Voice Assistant restricted to brand guideline documents only.
* **Monitor usage** via the agent's **Executions** tab to see how often it's being used and whether any runs are failing.
