> ## Documentation Index
> Fetch the complete documentation index at: https://docs.brilo.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Tag your calls automatically

> Turn on AI Call Tags so Brilo labels every call by what it was about, and add tags of your own for the things you want to count, all from the agent AI Call Tags tab.

Your agent's **AI Call Tags** tab does two things. One switch lets Brilo label every finished call with
what it was about, and **Add AI Call Tag** lets you write tags of your own for the outcomes you want to
count. Both take minutes and both need no developer.

## Before you start

* An agent that is taking calls. See [Turn your agent on](/go-live/turn-your-agent-on).
* Two minutes to turn tagging on, a week of calls before the tags tell you anything interesting.
* You will be working in your agent's **AI Call Tags** tab.

## Turn on automatic tagging

<Steps>
  <Step title="Open the AI Call Tags tab">
    Open your agent and go to **AI Call Tags**.
  </Step>

  <Step title="Turn on AI Call Tag">
    Switch on **AI Call Tag**. Calls that finish from this point on are labelled automatically with what
    they were about.
  </Step>

  <Step title="Take a few calls, then look at your history">
    Tags need calls to be useful. Give it a day of real traffic, then open your call history and look at
    what came back.
  </Step>

  <Step title="Read the tags before you change anything">
    The first pass is diagnostic. The tags tell you what your callers actually ring about, which is
    routinely different from what you assumed when you wrote your instructions.
  </Step>
</Steps>

## Add a tag of your own

Automatic tags describe the call in general terms. A tag of your own is a yes or no question you write
once, which your agent then answers about every call. Use it for the outcomes you want to count, such as
whether a booking was agreed or whether the caller asked about price.

<Steps>
  <Step title="Open Add AI Call Tag">
    On the **AI Call Tags** tab, select **Add AI Call Tag** at the top right of the list.
  </Step>

  <Step title="Pick an emoji">
    Select the tile at the top of the panel to choose an emoji. It defaults to a tick, and it is what you
    will scan for in your call history.
  </Step>

  <Step title="Name the tag">
    Enter a **Tag Name**, for example `Scheduling`. Names have to be unique on the agent, and the name is
    what appears on the call.
  </Step>

  <Step title="Describe when it applies">
    In **What should this tag mean?**, describe the condition in plain language, for example
    `Customer agreed to a follow-up meeting`. Your agent applies the tag to calls where that was true and
    leaves it off the rest, so write one condition, not several.
  </Step>

  <Step title="Save it">
    Select **Save**. The tag applies to calls that finish from now on. Add as many as you need, and select
    any tag in the list to edit or delete it.
  </Step>
</Steps>

<Note>
  Your own tags are evaluated per call and only appear on the calls where the condition was true. An absent
  tag means the condition was not met, not that tagging failed.
</Note>

## You will know it worked when

* Recent calls in your call history carry labels describing what they were about.
* Your own tags appear, with their emoji, on the calls that met the condition you described.
* Your tags are listed on the **AI Call Tags** tab with their emoji, description, and the date you added
  them.

## What to do with the tags

The point of tags is not tidiness, it is the three decisions they let you make:

* **Fix your knowledge.** A large group of calls about something your agent handled badly tells you exactly
  which document to write next. See
  [What your agent can learn](/knowledge/what-your-agent-can-learn).
* **Add the action you are missing.** If a big group of calls is people asking to book, and you have no
  booking action, that group is revenue leaking. See
  [Book appointments](/actions/book-appointments).
* **Rewrite your instructions.** A group of calls that all ended badly usually shares one cause, and it is
  usually a missing rule. See [Write what your agent says](/agents/write-what-your-agent-says).

Look at the tags weekly for the first month. After that, monthly is enough.

## Automatic tags, your own tags, extraction, or summaries?

All four read the finished call, and they answer different questions.

| You want to                                                 | Use                                             |
| ----------------------------------------------------------- | ----------------------------------------------- |
| Group and count calls by topic, without deciding the topics | The **AI Call Tag** switch                      |
| Count a specific outcome you define, as yes or no           | **Add AI Call Tag**                             |
| Get specific named values for your systems                  | [Extraction](/reports/extract-call-information) |
| Read what happened without the transcript                   | [Summaries](/reports/personalised-call-summary) |
| Know whether the call went well                             | [Pulse Score](/reports/pulse-score)             |

Turning the switch on costs nothing and needs no setup, which makes it the right first step of the four.
Add your own tags once the automatic ones have shown you what your callers ring about.

## What else you can do here

* Score call quality automatically alongside the tags. See [Pulse Score](/reports/pulse-score).
* Pull named values out of the same calls. See
  [Pull the details out of every call](/reports/extract-call-information).
* Send your tags to your own system with everything else after the call, where they arrive as `call_tags`.
  See [Send data to your systems](/actions/send-data-to-your-systems).
* Manage the same tags from your own code. See the [Call Tag API](/api-reference/endpoint/call-tag).

## If it did not work

**No tags appeared.** Check the switch is on for the agent whose calls you are looking at, and remember it
applies to calls that finish after you turned it on. Existing calls are not tagged retrospectively.

**My own tag never appears.** Either the condition was genuinely not met, or the description is too vague
to judge. Rewrite it as one plain condition a person could answer yes or no to, and check it on a test
call. See [Test your agent yourself](/go-live/test-your-agent-yourself).

**Brilo says the name already exists.** Tag names have to be unique on the agent. Open the existing tag
from the list and edit it rather than adding a second one with the same name.

**The tags are too generic to be useful.** Very short calls give little to work from. Look at the calls
over about thirty seconds, where there is a real conversation to read. For anything more specific, add a
tag of your own.

**Everything is tagged the same.** That may be true. A dental practice where four out of five calls are
appointments is a dental practice, not a tagging failure. Look at the smaller groups, because that is where
the surprises are.

## FAQ

<AccordionGroup>
  <Accordion title="Can Brilo label my calls automatically?">
    Yes. Open your agent, go to **AI Call Tags**, and turn on **AI Call Tag**. Every call that finishes
    from then on gets labels describing what it was about, with no setup beyond the switch.
  </Accordion>

  <Accordion title="Can I define my own call tags?">
    Yes. On the agent's **AI Call Tags** tab, select **Add AI Call Tag**, then give it an emoji, a unique
    **Tag Name**, and a plain description of when it applies. Your agent evaluates that condition on every
    call and tags the ones where it was true.
  </Accordion>

  <Accordion title="Are old calls tagged when I turn this on?">
    No. Tagging happens as each call ends, so it applies to calls from that point forward. Your existing
    history stays as it is, for the automatic switch and for tags you add yourself.
  </Accordion>

  <Accordion title="Why can't I save my call tag?">
    Three things are required: a **Tag Name**, a description in **What should this tag mean?**, and a name
    no other tag on that agent already uses. Brilo names the missing one in the error it shows.
  </Accordion>

  <Accordion title="What is the difference between call tags and extraction fields?">
    Call tags answer yes or no, which is what you want for grouping and counting calls. Extraction fields
    return a value you named, such as a date or a budget, which is what you want for sending data into
    another system. See [Pull the details out of every call](/reports/extract-call-information).
  </Accordion>

  <Accordion title="How many tags of my own can I add?">
    Brilo does not cap them, but every tag is one more judgement your agent makes about each call. Start
    with the three or four outcomes you would actually report on, and add more once those are earning
    their place.
  </Accordion>

  <Accordion title="Do call tags cost anything extra?">
    No. Tagging is part of how Brilo processes a finished call. See
    [What counts towards my usage](/account/what-counts-towards-my-usage) for what does affect your
    allowances.
  </Accordion>

  <Accordion title="Do my call tags reach my own systems after the call?">
    Yes. Matched tags arrive in the `call_tags` array on the post call webhook, each with a stable
    machine readable key derived from the tag name. See
    [Send data to your systems](/actions/send-data-to-your-systems).
  </Accordion>

  <Accordion title="What should I do with the tags once I have them?">
    Look for the biggest group your agent handles badly and fix that one thing, whether it needs a document,
    an action, or a rule. Repeat weekly for a month. That is the whole value of tagging.
  </Accordion>
</AccordionGroup>

## Related

* [Pulse Score](/reports/pulse-score) for how well each call went
* [Pull the details out of every call](/reports/extract-call-information) for values you define
* [Personalised call summaries](/reports/personalised-call-summary) for prose a person reads
* [What your agent can learn](/knowledge/what-your-agent-can-learn) to fix what the tags reveal
* [Send data to your systems](/actions/send-data-to-your-systems) to send tags onward