# What is an execution?

> 

Source: https://help.kustomer.com/en_us/what-is-an-execution-HyLX58boGg

Last updated: 2026-10-05T17:13:05.610Z

An execution is an attempt by an automation to handle a single event by carrying out an action according to instruction you provide.

This article explains what counts as one execution, what happens inside it, and when that execution becomes billable.

**In this article:**

*   [What counts as an AI execution](#defining-ai-executions)
*   [What happens inside an AI execution](#types-of-executions)
*   [How conversations relate to AI executions](#how-executions-trigger-in-conversations)
*   [How executions relate to billing](#how-executions-accumulate)

## Defining AI executions

Each time an AI execution triggers, it takes an action to get to the desired output. For example, the **Score** attribute's execution can score a customer's satisfaction rating based on the criteria you define. The specific action taken to perform that execution, such as reading the message and then examining and scoring it based on the criteria you define, are all a part of a single execution.

The final product is a score assigned to the conversation as requested, the actions taken to get to that output all count as one execution.

### Types of executions

[AI attributes](https://help.kustomer.com/en_us/understanding-ai-attributes-rydWQsOtGg) execute based on the instructions you write. AI attributes use generative AI to compute output based around these three types:

*   **Classifier.** Sorts the resource into labels you define.
*   **Score.** Scores the resource against a weighted rubric and stores a weighted score.
*   **Extractor.** Pulls out or summarizes text from the resource.

[AI QA](https://help.kustomer.com/en_us/understanding-ai-qa-ryHdt6utfe) scores a conversation against criteria you write. Its only trigger is **Conversation Done**.

## How executions accumulate

One conversation can generate multiple executions based on how you have configured AI to be used throughout.

**The volume of executions is controlled by trigger selection and configured conditions.**

An AI attribute or AI QA rule accumulates one execution per qualifying event. If a trigger has no or broad filtering conditions (e.g., channel, tag, conversation status), every matching event will generate an execution. These events can include spam or low-signal conversations.

To limit execution volume, [add conditions](https://help.kustomer.com/en_us/using-matches-all-and-matches-any-in-filters-HkZ7kmAbg) to the attribute's configuration.

## How executions trigger in conversations

Each inbound event can start executions based on its triggers, so a single conversation with a customer can generate several executions over its lifetime.

For example, let's assume these are all active:

*   Issue Category, a classifier, trigger **Conversation Created**.
*   Order Number, an extractor, trigger **Message Count** at 2 inbound messages.
*   CSAT Score, a score, trigger **Conversation Done**.
*   AI QA, trigger **Conversation Done**, with a condition that the channel is email.

Under these conditions, the following executions are possible:

1.  The conversation is created.
    1.  Issue Category evaluates.
    2.  That is one execution.
2.  The customer sends a second message.
    1.  Order Number evaluates once.
    2.  That is a second execution.
    3.  A conversation that ends after one inbound message never gets this evaluation.
3.  The conversation is marked Done, and the channel is email.
    1.  CSAT Score and AI QA both evaluate.
    2.  Each evaluation adds one execution, for a total of two additional executions.

If the channel is not email, CSAT Score can also still evaluate, but the AI QA evaluation is skipped because its conditions were not met.

### When executions do not accumulate

Executions do not accumulate when they encounter an error that prevents completion.

As examples:

*   5xx errors
*   Rate limit errors