# Chat analysis: build an agent from your own conversations

**Chat analysis** reads the latest conversations your team already answered in one inbox and groups them into **intents**: the reasons people get in touch. Nothing is built or changed until you ask for a draft.

## Start an analysis

![The Analyse chats dialog](../../.gitbook/assets/guides/chat-analysis-dialog.jpg){ width="424" }

1. Open **Flow Agents** and click **Analyse chats**.
2. Pick the **Inbox**. The dialog shows how many recent conversations with a team reply qualify.
3. Click **Start analysis**.

Only conversations a person on your team replied to are read. The first run reads the latest 30. You can leave and come back: **Open the latest analysis of this inbox** takes you to the last result.

## Read the map

When the analysis is **Ready**, the top of the page counts **Conversations read**, **Intents found**, **Left out as noise**, any that **could not be read**, and the **Languages** used. Each intent is a card:

| On the card | What it shows |
|---|---|
| Share of the sample | For example *12 of 30 conversations* |
| **How customers put it** | Real phrasings from the chats |
| **Your team usually asks for** | Details your team collects for this request |
| **How your team usually answers** | The typical reply |
| **Open conversation #…** | Each conversation behind the intent |

Below the cards, **Conversations read** lists every chat with its language and outcome: **Answered**, **Handed to the team**, **Booked**, **No reply** or **Unclear**. Chats that were not real requests are marked as noise: **Another business's message**, **Automated reply**, **Spam** or **Too short to tell**.

**Hide from the map** drops an intent you don't want in the agent; **Show again** brings it back.

## Widen or rerun

* **Widen to 40 conversations** (then 50) reads more history and redraws the map.
* **Run again** starts a fresh analysis of the same inbox, for example after a busy month.

## Turn the map into a draft agent

Click **Draft agent**. You get a draft flow agent built around the main reasons customers write in, or a menu when the reasons are spread out. Smaller intents become FAQ drafts in AI Knowledge; hidden intents are left out. **Open draft agent** opens it in the designer.

The draft is connected to no inbox and answers nobody until you publish it. Check it in the Playground, and approve the FAQ drafts under **Review FAQ drafts**; see [Managing AI Knowledge](/flow-agents/managing-ai-knowledge).

## When something goes wrong

| Message | What to do |
|---|---|
| **No intents came out of these conversations** | Widen the sample or pick a busier inbox |
| **The analysis stopped before finishing** / **The draft could not be written** | Run it again a little later |
| **Chat analysis is not available for this account yet** | The feature is not on for your plan |

## Related

* [Creating a Flow Agent](/flow-agents/creating-a-flow-agent)
* [Building the conversation flow](/flow-agents/building-the-conversation-flow)
* [Managing AI Knowledge](/flow-agents/managing-ai-knowledge)
* [Testing and connecting an agent](/flow-agents/testing-and-connecting-an-agent)

---
Source: https://help.mampuai.com/flow-agents/chat-analysis