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

# AI Agents

Configure your AI agents, their procedures, and their actions to handle your customers’ requests and involve your team.

{% hint style="success" icon="list-check" %}
**AT A GLANCE**

Fullwhere AI agents generate responses based on your instructions, examples, and the knowledge you provide.

With procedures, they can carry out multi-step processes, qualify an interaction, manage reservations, and forward a request to your team. Escalations and the fallback option cover situations that require human intervention.
{% endhint %}

## Choose what to configure

| Your need                                                              | Where to configure it                                                                                                                                    |
| ---------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Define the agent’s role, tone, and shared rules                        | [Writing instructions](#h_d4bb965222)                                                                                                                    |
| Provide examples for simple, recurring situations                      | [Custom responses](#h_d0f171a565)                                                                                                                        |
| Provide information about your services                                | [Knowledge bases](https://docs.fullwhere.com/documentation/documentation-en/automations/knowledge-bases.md)                                              |
| Handle a multi-step request that may require your agent to take action | [Procedures](https://docs.fullwhere.com/documentation/documentation-en/automations/ai-agents/procedures.md)                                              |
| Assign, notify, or populate attributes                                 | [Actions](https://docs.fullwhere.com/documentation/documentation-en/automations/ai-agents/actions.md)                                                    |
| Manage reservations                                                    | [Reservation management](https://docs.fullwhere.com/documentation/documentation-en/automations/ai-agents/actions.md#actions-de-gestion-des-reservations) |
| Involve the team in a specific situation                               | [Escalations and fallback](https://docs.fullwhere.com/documentation/documentation-en/automations/ai-agents/escalades-et-dernier-recours.md)              |
| Adapt instructions for each customer or location                       | [Dynamic variables](#h_3198d4840a)                                                                                                                       |
| Define which interactions are covered and validate the result          | [Test and deploy an agent](https://docs.fullwhere.com/documentation/documentation-en/automations/ai-agents/tester-et-deployer.md)                        |

A **workflow** defines when to use the agent. A **procedure** describes how to handle a situation. An **action** performs an operation in Fullwhere or in a reservation system at the point specified in the instructions.

## Access your agents <a href="#h_d7b3dc58a2" id="h_d7b3dc58a2"></a>

Open **Automations →** [**AI agents**](https://app.fullwhere.com/automation/agents). Select an existing agent to edit its configuration.

You can specialize your agents by interaction type, brand, or location, then define their scope in your [workflows](https://docs.fullwhere.com/documentation/documentation-en/automations/workflows.md).

{% hint style="info" icon="shield-check" %}
**ACCESS PERMISSIONS**

To edit the AI agents in your Fullwhere workspace, you must be a supervisor member with Administrator permissions.
{% endhint %}

## Create an agent <a href="#h_cdb8854904" id="h_cdb8854904"></a>

From the agents list, click the create button and give your agent a descriptive name, such as **Customer Relations Agent** or **Review Response Agent**.

Then configure its instructions and resources, and save your changes.

## Configure the agent <a href="#h_05580acb61" id="h_05580acb61"></a>

Start by defining the agent’s shared behavior and the examples it should follow. Then add the procedures and team handoff rules required for your business.

### Writing instructions <a href="#h_d4bb965222" id="h_d4bb965222"></a>

Writing instructions are divided into four fields:

| Field                       | What it should specify                                                                                   |
| --------------------------- | -------------------------------------------------------------------------------------------------------- |
| **Agent objective**         | The role to fulfill and the requests to handle.                                                          |
| **Writing instructions**    | The language, tone, whether to use formal or informal address, polite expressions, and wording to avoid. |
| **Customer context**        | Useful information for personalization, such as first name, organization, history, or satisfaction.      |
| **Additional instructions** | Any other shared rules to follow.                                                                        |

Use [dynamic variables](#h_3198d4840a) to adapt the instructions to the interaction. Specify what the agent should do if information is missing.

For detailed information about your services, connect a [knowledge base](https://docs.fullwhere.com/documentation/documentation-en/automations/knowledge-bases.md) to the response generation action in your workflow. Steps specific to a business case should be described in a [procedure](https://docs.fullwhere.com/documentation/documentation-en/automations/ai-agents/procedures.md).

### Custom responses <a href="#h_d0f171a565" id="h_d0f171a565"></a>

Custom replies are reference responses the agent can use to handle simple situations: the agent draws on their meaning and style, then adapts their wording to the conversation.

Organize them into topic groups, such as **Positive reviews**, **Reservations** or **Complaints**. Specify the situations in which the agent should use them.

#### Create a group <a href="#h_c3b4f2dad0" id="h_c3b4f2dad0"></a>

In Custom Replies, click **Create a group**, then enter the application criteria and sample replies. Each reply includes a label describing the situation and a reference text.

The text can include [dynamic variables](#h_3198d4840a). Confirm the group creation.

Example situation:

> The customer wants to change an existing reservation but hasn’t specified the new time.

Example reply:

> Hello {{customer\_first\_name}}, could you please provide the new date, time, and number of people? Our team will check availability before confirming the change.

#### Manage replies <a href="#h_65b32cb7a6" id="h_65b32cb7a6"></a>

From the Custom Replies table, edit a reply’s label and text, open its detailed editor, or rename a group.

Describe sufficiently distinct situations to avoid contradictory examples. For requests that require operations in Fullwhere, use a procedure with configured actions.

## Create procedures

Procedures [procedures](https://docs.fullwhere.com/documentation/documentation-en/automations/ai-agents/procedures.md) give the agent step-by-step instructions for handling complex situations, including carrying out specific actions in Fullwhere or external systems.

## Configure actions

Actions [actions](https://docs.fullwhere.com/documentation/documentation-en/automations/ai-agents/actions.md) are created from the **/** instructions menu. They appear as badges:

* **Transfer to team members** : define assignment, notifications, and how the agent should behave in subsequent messages;
* **Set attributes** : fill in the interaction fields;
* **Reservation management** : search for, create, modify, or cancel a reservation according to the permitted capabilities and the integration’s settings.

Writing “transfer the request” in an instruction does not replace adding and configuring a transfer action.

## Plan for escalations and fallback

In the [Escalations](https://docs.fullwhere.com/documentation/documentation-en/automations/ai-agents/escalades-et-dernier-recours.md)section, describe the situations in which your team should step in and add the corresponding actions directly to the instructions. Also configure the fallback for requests the agent cannot handle with its available resources.

For example: an explicit request to speak to someone, a complaint to be forwarded, or a question with no answer in the available knowledge.

## Use dynamic variables <a href="#h_3198d4840a" id="h_3198d4840a"></a>

Variables are replaced with interaction data at generation time. They let you tailor instructions, reference responses, and context for each customer.

In the new procedure, escalation, and fallback editors, click **/** then select **Insert a variable**.

A variable with no available data is replaced with empty text. For example, an email does not have a rating `{{rating}}`. Use wording that still sounds natural when information is missing.

### General variables <a href="#h_a055fec090" id="h_a055fec090"></a>

| Variable                        | Description                                                                                                                                                                          |
| ------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| `{{tags}}`                      | List of feedback tags, shown in brackets and separated by commas.                                                                                                                    |
| `{{content}}`                   | Text content of the customer feedback.                                                                                                                                               |
| `{{rating}}`                    | Rating given by the customer in the feedback.                                                                                                                                        |
| `{{language}}`                  | Target language of the feedback.                                                                                                                                                     |
| `{{source}}`                    | Source of the feedback (e.g., Google, Email, Instagram, etc.).                                                                                                                       |
| `{{class}}`                     | Class or category of the feedback (Review, Contact, Survey, Message)                                                                                                                 |
| `{{store}}`                     | Name of the relevant location (e.g., store, restaurant, branch…).                                                                                                                    |
| `{{store_signature}}`           | Signature associated with the location.                                                                                                                                              |
| `{{store_contact}}`             | Contact email address for the location.                                                                                                                                              |
| `{{brand}}`                     | Name of the brand associated with the response.                                                                                                                                      |
| `{{brand_signature}}`           | Brand signature.                                                                                                                                                                     |
| `{{brand_contact}}`             | Brand contact email address.                                                                                                                                                         |
| `{{contact}}`                   | Contact email used for the response: the location’s email if available, otherwise the brand’s.                                                                                       |
| `{{signature}}`                 | Signature used for the response: the location’s if available, otherwise the brand’s.                                                                                                 |
| `{{customer_first_name}}`       | Customer’s first name (if available).                                                                                                                                                |
| `{{customer_last_name}}`        | Customer name.                                                                                                                                                                       |
| `{{customer_full_name}}`        | Customer's full name (first name + last name).                                                                                                                                       |
| `{{customer_civility}}`         | Customer's title (e.g., Mr., Ms., Dr., etc.).                                                                                                                                        |
| `{{customer_feedback_count}}`   | Total number of feedback entries left by the customer.                                                                                                                               |
| `{{customer_purchase_count}}`   | Total number of orders placed by the customer.                                                                                                                                       |
| `{{customer_purchase_count_1}}` | Number of orders placed by the customer + 1 (useful for forward-looking phrasing).                                                                                                   |
| `{{customer_purchase_amount}}`  | Total cumulative amount of the customer's orders.                                                                                                                                    |
| `{{customer_average_rating}}`   | Average rating given by the customer.                                                                                                                                                |
| `{{purchase_amount}}`           | Amount of the relevant order.                                                                                                                                                        |
| `{{purchase_external_id}}`      | External order ID.                                                                                                                                                                   |
| `{{now}}`                       | Lets you provide your AI agent with the current date and time when generating a response. Specify the time zone to use in your instructions if your rule depends on a specific time. |

### Variables specific to satisfaction surveys <a href="#h_25b8fbde08" id="h_25b8fbde08"></a>

Use these variables to provide context to the agent generating responses to your satisfaction surveys.

| Variable                  | Description                                    |
| ------------------------- | ---------------------------------------------- |
| `{{survey_fields}}`       | List of all survey fields.                     |
| `{{survey_csat_smileys}}` | Results of *CSAT Smileys-type questions*.      |
| `{{survey_nps}}`          | Results of *NPS-type questions*.               |
| `{{survey_csat_stars}}`   | Results of *CSAT Stars-type questions*.        |
| `{{survey_notations}}`    | All rating fields (CSATs & NPS).               |
| `{{survey_choices}}`      | All choice fields (dropdown, checkbox, radio). |

### Variables specific to contact forms <a href="#h_d4588e34f9" id="h_d4588e34f9"></a>

Use these variables to provide context to the agent generating responses to your contact forms.

| Variable           | Description                             |
| ------------------ | --------------------------------------- |
| `{{site_choices}}` | All choice fields (dropdown).           |
| `{{site_texts}}`   | All text fields (text & purchase).      |
| `{{site_fields}}`  | All fields (dropdown, text & purchase). |

### Custom variables

To further customize your AI agents, you can define [custom variables](https://docs.fullwhere.com/documentation/documentation-en/automations/variables.md) whose values will be replaced based on the relevant brand or location.

## Test and deploy

Connect your agent to an action **Generate an AI response** in a [workflow](https://docs.fullwhere.com/documentation/documentation-en/automations/workflows.md#generer-une-reponse-ia), then validate the responses and action effects using test interactions.

Follow the guide [Test and deploy an agent](https://docs.fullwhere.com/documentation/documentation-en/automations/ai-agents/tester-et-deployer.md) to verify procedure thresholds, handoffs, attributes, and the fallback.

{% hint style="warning" %}
Disabling automatic sending keeps responses as drafts, but does not guarantee that the agent won’t execute actions. Separately verify assignments, notifications, attribute changes, and booking operations.
{% endhint %}

## Delete an agent

An agent used in a workflow cannot be deleted. First remove its use from the relevant workflows.


---

# 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://docs.fullwhere.com/documentation/documentation-en/automations/ai-agents.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.
