> For the complete documentation index, see [llms.txt](https://help.meltingspot.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.meltingspot.io/english/coach-ai/user-experience/conversations-with-the-ai-coach.md).

# Conversations with the AI Coach

Beyond displaying content, the AI Coach can engage directly with your users when two-way conversation is enabled (👉 more information [here](https://help.meltingspot.io/english/coach-ai/configuration/configuring-your-ai-coach.md)).

### **What the AI Coach can do**

It relies on your Spot's content (courses, documents, live sessions, forums, learning paths...) to build its responses and suggest them as complementary resources. Three types of requests:

* **FAQ**: explaining a concept, a feature, a product term.
* **Step-by-step guidance**: describing the actions to take, without ever performing them.
* **Recommendations**: suggesting best practices, possible choices, content to check out.

It never performs a concrete action and never makes a decision on the user's behalf.

{% hint style="info" %}
The AI Coach only relies on **published** content the user **has access to**. For example, a secret course, an invitation-only live session or a private channel is only used for members who have access to it.
{% endhint %}

### **How the user starts a conversation**

* **From displayed content**: to dig deeper into or validate what they just saw. The AI Coach primarily relies on this content.
* **On their own initiative**: direct click on the AI Coach, for example when facing difficulty somewhere not covered by an admin.
* **On the AI Coach's initiative (proactive)**: based on behavioral signals. 👉 more information [here](https://help.meltingspot.io/english/coach-ai/user-experience/proactive-messages.md).

### Finding past conversations

When the user clicks on the AI Coach, they first find the list of their past conversations with this AI Coach, rather than automatically resuming the last ongoing discussion.

For each conversation:

* a short, automatically generated title and summary,
* the date of the last activity.

{% hint style="info" %}
This list is specific to each user: they only see their own conversations, never those of other members.
{% endhint %}

A **+** button lets you start a new conversation at any time, without losing access to previous ones.

A conversation is automatically closed after **60 minutes of inactivity**. It stays in the history: the user can resume it at any time from the list.

#### Searching and reopening a conversation

{% hint style="info" %}
These features are being rolled out gradually.
{% endhint %}

* **Search**: a search field lets the user find a conversation by its title, by a suggested resource or by the content of a message.
* **Reopen last session**: a button lets the user go straight back to the last conversation.

### **How it adapts its responses**

The AI Coach personalizes its responses based on the user's role, groups, custom properties, the current page, resources already viewed, and the history of their exchanges on this Spot.

It responds in the detected language, using the tone(s) chosen by the admin in **Settings > AI Coach > Persona** (**Professional** by default; several tones can be combined). In all cases, it stays clear and caring, and never makes the user feel bad about a mistake. It favors concise responses.

It cites its sources discreetly: at most **3 recommended resources** per response, shown as cards.

### **When the AI Coach doesn't know how to answer**

If it lacks information on a product-related question, it clearly states so and offers, if configured, to switch to the fallback (👉 more information [here](https://help.meltingspot.io/english/coach-ai/configuration/configuring-the-ai-coachs-fallback.md)).

For clearly off-topic questions, it says so with a light tone rather than attempting an answer. It never invents information.

### Rating a response

The user can rate an AI Coach response with 👍 or 👎. This vote is offered on responses that recommend content (with a content card).

After a 👎, the user can pick a reason:

* Poor explanation
* False information
* Not following my request
* Incomplete response
* Chat history issue
* Personalization issue
* Other

They can also add a free comment (**1000 characters** max). The AI Coach then apologizes and offers the fallback, if configured.

{% hint style="info" %}
Once sent, the vote can no longer be changed.
{% endhint %}

### Good to know

* A message is limited to **8000 characters**.
* The user sends **one message at a time**: they wait for the AI Coach's response before sending another one.
* The response is displayed **progressively**. It can be copied.
* If an error occurs, a **Retry** button appears.
* Closing the AI Coach interrupts the response in progress.


---

# 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.meltingspot.io/english/coach-ai/user-experience/conversations-with-the-ai-coach.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.
