使用状況の測定と請求ロジック (フレックス プラン)

UiPath® Document Understanding^TM モダン プロジェクトを使用する場合、AI ユニットの消費について理解することが重要です。

ヒント:

• クラシック プロジェクトのライセンス情報については、クラシック プロジェクトの『Document Understanding ガイド』の「使用状況の測定と請求ロジック」をご覧ください。
• AI ユニットの消費を組織レベルおよびテナント レベルで追跡するには、「[AI ユニットの消費数の概要] ダッシュボード」をご覧ください。
• Document Understanding および AI Center のプロジェクトの AI ユニットの消費を追跡するには、「IXP の AI ユニット消費ダッシュボード」ページをご覧ください。

How charging works for modern projects

For Document Understanding modern projects, only one single AI unit is charged per page, regardless of the operations performed at runtime or the number of modern projects extracting data from the same page. The single per-page charge covers:

• Digitization, including OCR
• 抽出
• 分類

This applies to both robotic process automation (RPA) and REST API operations. For full REST API pricing — including how classic-project API calls are charged — see Metering and Charging in the API guide.

Through Document Understanding activity packages

For both DocumentUnderstanding.Activities and IntelligentOCR.Activities, one AI unit is charged per page, regardless of the operations performed (digitization, extraction, or classification).

The one AI unit per-page rule applies only when:

• For DocumentUnderstanding.Activities: a pre-defined generative or modern project classifier or extractor is in use.
• For IntelligentOCR.Activities: a modern project classifier or extractor is in use.

If you use any other classifier or extractor, the respective consumption logic from classic projects is added to the one AI unit per page. For more information, see Mixed projects.

Generative features — Activities 24.12-preview and above

This consumption applies in the following situations:

• PredefinedGenerative プロジェクトでバージョン 24.12-preview 以降のアクティビティを使用している
• PredefinedGenerative プロジェクトで Document Understanding API v1.1 を使用している

There are no additional AI units consumed when using Generative Classification and Generative Extraction. Only one single AI unit is charged per page, regardless of the operations performed, including Generative Classification and Generative Extraction.

注:

Using Generative Validation consumes an additional one AI unit for each processed page. This brings the total to two AI units consumed per page when generative validation runs.

Generative features — Activities up to 24.12-preview

This consumption applies in the following situations:

• Predefined プロジェクトでバージョン 24.12-preview より前のアクティビティを使用している
• Predefined プロジェクトで Document Understanding API v1.0 を使用している

In these versions, generative features incur specific charges per operation, as described in the following table.

Table 1. Generative features AI units consumption for activities versions up to 24.12-preview

生成 AI 機能	消費された AI ユニット
生成 AI 分類器	操作あたり 0.2
生成 AI による抽出	操作あたり 1.2
生成 AI による検証	操作あたり 1

モデルのトレーニング

There are no AI units charged when training and serving models. You can train models freely to improve performance, develop new capabilities, and experiment without incurring additional AI units charges.

Mixed projects (modern + classic)

When a project uses both classic and modern experiences, both pricing models apply based on their respective consumption.

The pricing model for classic projects applies in the following circumstances:

• Extraction and classification done using classic projects, regardless of the activity package used (IntelligentOCR or DocumentUnderstanding.Activities).
• Extraction and classification done using the out-of-the-box models included with the pre-defined project, regardless of the activity package used.
• REST API calls against classic-project models — classic per-capability rates apply, not the 1 AI unit per-page rule.

For full classic-project metering — including activity-based rates per OCR engine, classifier, and extractor — see the Document Understanding User Guide for Classic Experience. For classic-project REST API rates specifically, see Metering and Charging in the API guide.