# AI Agent Reports

> How your chatbots and voicebots handled each conversation

## AI Agent reports

The **AI Agents** reports tell you how your chatbots and voicebots behaved: how many conversations they handled, what they consumed, and what happened in each one.

There are three of them, picked from the dropdown in the header:

- **Chatbot**: How many interactions each chatbot had, day by day.
- **Voicebot**: The same for voicebots, plus how many minutes they used.
- **Voicebot History**: Call by call: who called, how long it lasted, and how it ended.

The toolbar, the filters and the download work as they do across [the rest of the reports](https://docs.audara.io/en/reportes/). All three open on **today's** data, and you widen the period from there.

> **Note**
> The date filter here has four options, not five: **Today**, **Yesterday**, **Last 7 days** and **Range**. The year option is not among them.

## Chatbot and Voicebot

These two are summaries and read the same way: one row per bot per day, with the interaction count. If a bot had no activity on a day, that day does not show up.

![The Chatbot report: **Date**, **Chatbot** and **Interactions**. A bot active on several days appears once per day.](https://docs.audara.io/reportes-bots/imagenes/chatbot.jpg)

*The Chatbot report: **Date**, **Chatbot** and **Interactions**. A bot active on several days appears once per day.*

The **Voicebot** one carries those same three fields and adds **Usage (min)**, the minutes that voicebot burned that day.

![The Voicebot one adds **Usage (min)**. Read that column against the interaction count: they do not always rise together.](https://docs.audara.io/reportes-bots/imagenes/voicebot.jpg)

*The Voicebot one adds **Usage (min)**. Read that column against the interaction count: they do not always rise together.*

Filters on these two are minimal, the date and the bot, because there is not much else to filter in a summary.

> **In practice**
> A voicebot's **Usage (min)** is meant to be read next to the interaction count, not instead of it. Lots of minutes with few interactions means long conversations, and a long conversation with a bot is rarely good news: it is usually someone failing to make themselves understood.

## Voicebot History

This is the detailed report. One row per call, and it is where you actually understand what is happening on voice.

![Above, the four totals for the period. Below, one row per call with its times and how it ended.](https://docs.audara.io/reportes-bots/imagenes/historial-voicebot.jpg)

*Above, the four totals for the period. Below, one row per call with its times and how it ended.*

### The four numbers on top

Above the table sit four totals that sum up the period:

- **Total calls**: How many the voicebot handled.
- **ATT**: How long a conversation with the bot lasted on average.
- **Sent to campaign**: What share ended up going to a human.
- **Abandons**: What share dropped without getting anywhere.

The ratio between the last two is the indicator that matters: how many the bot resolved on its own. If nearly all of them go to a campaign, the bot is working as a receptionist rather than an agent.

### The columns

| Column | Description |
| --- | --- |
| Detail | Opens the full conversation |
| Initiated | How the call started: Incoming, Outgoing or Transfer, shown as an icon |
| Date | The day of the call |
| Start conversation | The time it began |
| End conversation | The time it ended |
| Total time | How long it lasted |
| Schedule | Whether it landed in Business hours or Off hours |
| Voicebot | Which of your voicebots took it |
| Bot tags | The marks the bot itself left during the conversation |
| Contact phone | The number of whoever called |
| End action | How it ended |
| Destination name | Where the call went, when it went somewhere |

The downloaded file also carries the **Schedule result**, which is not shown on screen.

> **Note**
> A column full of dashes is not an error. **Destination name** only has a value when the call went somewhere, and **Bot tags** only when the flow tags anything. A dash means that did not apply to that call.

### How the call ended

The **End action** column is the most informative one in the report, and its values are worth knowing:

- **Call hang up**: The conversation closed right there, passed to nobody.
- **Inactivity**: The person stopped answering and the bot closed.
- **Transfer**: The call was transferred.
- **Goto queue**: It went to an inbound campaign, meaning to a human agent.
- **GotoIVR**: It went back to a voice menu.
- **Goto extension**: It went to a specific extension.
- **Goto conference**: It entered a conference room.
- **Goto voicemail**: It ended in a mailbox.
- **Voicebot limit**: The conversation ran out of turns before it was resolved.

When the ending was a handoff, the **Destination name** column says which campaign, IVR or extension.

> **Important**
> **Voicebot limit** is worth going through one by one. It is not that the person left, nor that the bot resolved anything: it is that the conversation hit its turn cap. Every one of those is a call left half done, and they usually point at a spot in the flow where the bot and the person go round in circles.

### The filters

![This report's filter panel, the fullest of the three.](https://docs.audara.io/reportes-bots/imagenes/historial-filtros.jpg)

*This report's filter panel, the fullest of the three.*

On top of the date, it filters by **Voicebot**, **Initiated**, **Schedule**, **End action**, **Bot tag** and **Contact phone**.

The **Bot tag** one is worth learning: if your voicebot marks conversations as things unfold, this filter lets you isolate exactly those. How those tags get set is covered in [Voicebot](https://docs.audara.io/en/voicebots/).

### One call in detail

The **Detail** button opens the full conversation, with what was said on both sides in order and the time of each turn.

![The conversation turn by turn. It opens with the user's **Start**, which is the call coming in, and closes with the time it was considered finished.](https://docs.audara.io/reportes-bots/imagenes/detalle-conversacion.jpg)

*The conversation turn by turn. It opens with the user's **Start**, which is the call coming in, and closes with the time it was considered finished.*

If the call was tied to a CRM contact, a tab with their details shows up as well.

This is where the real audit happens. An abandon in the table is a figure; in the detail you see the question the bot could not answer.