Classification Models

What classification models do

Classification is the process of identifying a piece of content, such as a document or email, based on certain traits. Classification models can automatically route content to the appropriate downstream workflow. For example, a classification model can identify whether an incoming document is a claim or an invoice.

In Appian, you can integrate classification into your processes using classification models in DocCenter.

The Classification tab in DocCenter has two pages: Models and Instances.

• Models: Lists all classification models in the environment. Click on a model to see all model versions and performance metrics.
• Instances: Lists each process execution of a classification model.

Classification model limits

DocCenter classification models have the following limits:

Configuration	Limit
Pages (standard classification)	Up to 1,000 pages. DocCenter processes large documents by splitting them into batches and sending each batch to the LLM separately.
Pages (Split Packets Into Individual Documents)	Up to 1,000 pages when Split Packets Into Individual Documents is enabled. When using Enable Vision Classification and Split Packets Into Individual Documents together, the limit is 100 pages.
Pages (Vision Classification)	Up to 1,000 pages when Enable Vision Classification is enabled. The page limit depends on the selected model's vision page limit. When using Enable Vision Classification and Split Packets Into Individual Documents together, the limit is 100 pages.
Categories per model	Up to 100 categories.

Supported file types

DocCenter classifies the following file types:

• CSV
• DOCX
• EML
• GIF
• HTML
• JPG, JPEG
• MSG
• PDF
• PNG
• TIFF
• XLSX (limited based on the complexity of the spreadsheet)

Create and configure a classification model

1. In the Classification page group, click Models.
2. Click Create Model.

3. Enter the Model Details:

   Field	Description
   Name	Name to uniquely identify model.
   Description	Description of model (optional).
   Key	Automatically generated string to uniquely identify model. Read only.

   

   Scroll or drag to pan · ⌘+/⌘− to zoom · Esc to close

4. Configure Additional Options for the extraction model:

   Field	Description
   Enable Vision Classification	Allow the model to better detect visual elements within the document to aid in classification.
   Log AI Usage	View detailed information on how this model uses AI-powered features. Must be enabled to generate suggested improvements, but may slow down overall classification times. When enabled, AI usage details appear in the AI Usage tab when viewing an instance.
   AI-Suggested Improvements	Get AI-generated suggestions on how to improve model accuracy. Enabling this option increases your AI action consumption.

5. Click NEXT.
6. In step 2, configure the model type and categories.
7. Choose how the model will classify content:
   • Select Individual Documents (default) to classify single documents, or
   • Select Split Packets Into Individual Documents to identify and split large PDF packets containing multiple documents. Each document type is classified and saved as a separate file for downstream processing.
8. Choose how many pages the model reviews before classifying:
   • Select Classify after reviewing the entire document (default) to read all pages.
   • Select Classify after reviewing a set number of pages and enter a value in the Pages field to limit how many pages the model reads. For most documents, the first 1–2 pages contain enough context for accurate classification. Use this option on large documents to reduce AI action consumption without affecting accuracy.

9. Add categories and descriptions:

   Field	Description
   Category	Enter the category name, such as Invoices.
   Description	Enter the category description, such as Documents we receive related to company purchase.

   

   Note:  When you select Split Packets Into Individual Documents, an Other category is automatically created by default. This category is used for any pages in the packet that don't match other defined categories.

10. In step 3, refine the prompt:

    Field	Definition
    Large Language Model	Choose the provider and model you'd like to use to classify data. The dropdown displays all preconfigured provider and model combinations available in your environment (for example, "Appian - Claude Sonnet 4.6 (Reasoning)" or "AWS - Claude Sonnet 4.6 (Reasoning)"). Some models prioritize speed, while others provide more advanced reasoning. Choosing the right model depends on your classification model's performance and accuracy needs. For models that split packets into individual documents, select a reasoning model for better accuracy in identifying document boundaries and types. To add more providers, see AI Services.
    Extended Thinking	Enable extended thinking to allow the model additional reasoning time before producing results. This can improve accuracy for complex classification tasks that require nuanced document analysis. Extended thinking appears as a toggle directly below the LLM dropdown. Note that enabling this option increases AI action consumption.
    Role	Tell the model what it's responsible for in this task. The role is pre-populated with text, but you can edit it as you see fit.
    Additional Instructions	Tell the model anything else it should keep in mind as it performs analysis and classifies items. For example, you could tell the model, "If you're unsure which category a document goes into, say so".
    Examples	Add inputs and outputs the AI model can use as an example of what you expect to receive as a response when presented with text.
    Generated Prompt	Preview the prompt that will be used when this AI model is called in a process.

    Note:  For single document classification, reasoning text is no longer included in responses by default, reducing AI action consumption on every classification call. To restore detailed reasoning, enable Extended Thinking and Log AI Usage. To review the model's judgment, check the AI Usage tab on the instance.

    

11. Click CREATE to proceed with creating your classification model.

When to split packets into individual documents

When creating a classification model, you can choose to split large PDF packets into individual documents. This option is useful when a single uploaded file contains multiple documents that need to be processed or extracted separately.

Consider enabling Split Packets Into Individual Documents when either of the following applies:

• Documents require different extraction models. If the documents in a packet have different fields to extract, they need to go through separate extraction models. For example, a packet might contain both invoices and purchase orders—related documents, but with different data to capture.
• Each document corresponds to a different case. If each document in a packet represents a distinct business transaction or entity, splitting allows each one to be processed independently. For example, a packet might contain several invoices from different providers, where each invoice needs its own downstream workflow.

If neither criterion applies—for example, if the packet contains a single document type and all pages belong to the same case—you can use standard Individual Documents classification instead.

Test a classification model

Before you deploy anything to production, you want to know how well it meets your expectations. You can test a classification model in the DocCenter application, review its accuracy metrics, and make any adjustments you'd like before the model classifies any real data.

To test a model:

1. In the Classification page group, click Models.
2. In the list, find and click on the model you want to test.
3. Click RUN CLASSIFICATION.
4. Choose the type of document you want to classify:
   • Existing Document: Choose from a file in Appian.
   • New Document: Upload a file.
   • Add Text: Enter plain text.
5. Click SUBMIT.
6. Click the Instance tab, where your test run appears.
7. Click the instance that represents your test to learn more, or glance at the Value column to quickly see which category the model chose for your content.

To see the model's reasoning why it chose a certain category, open the instance.

Scroll or drag to pan · ⌘+/⌘− to zoom · Esc to close

If the model categorized your content incorrectly, you can correct it. Click Reconcile Instance in the related action menu.

After running a few tests, if you notice that the model isn't performing well enough for your purposes, you can create a new version to try tweaking the prompt or other details.

Testing models that split packets into individual documents

If the classification model is configured to automatically identify and split documents, the test results display additional information:

• Document Types: All document types identified within the packet.
• Page Ranges: The start and end pages for each identified document.
• Original Document: A link to view the original unmodified packet.
• Classification Reasoning: The AI's reasoning for each document type classification.

You can verify that the model correctly identified document boundaries and page ranges. If the splits are inaccurate, you can reconcile the instance to correct the page ranges and document types before using the model in production.

View AI-Suggested Improvements

AI-Suggested Improvements uses AI to analyze your classification model's performance and generate recommendations for improving accuracy. If you enabled AI-Suggested Improvements for the model, the suggested improvements appear on the Testing tab and when you update a model. Developers can view detailed suggestions about how to update a model's configurations to improve accuracy.

When you've implemented the suggested improvements, click the checkmark to mark the item as done. Click MARK AS DONE at the top of the list to mark all items as done. Marking an item as done removes it from the list, so you and other developers won't see it anymore.

Processing versions

Each model version carries a processing version that determines how DocCenter processes documents internally. When a newer processing version is available, a banner appears on the model version page. To upgrade, see Upgrade a processing version.

Create a new model version

1. Click on Classification page group and select Models.
2. Find the model you want to use as the basis for your new version. Then:
   • Click the overflow menu in the model's row, and click New Version.

   OR

   • Click the model name to open its summary. In the menu of options at the top of the page, click New Version.
3. On the Create Classification Model Version page, edit the following details:
   • Version Description: Provide information about the updates you're making in this version.
   • Options: Add or remove options.
   • Prompt: Edit the Role and Additional Instructions fields as needed.
   • Examples: Add or remove relevant examples for this model to use as a reference.

   Tip:  Refer to Create and configure a classification model for details on configuring the model.

4. Click SAVE to save this version for future updates, or SAVE AND PUBLISH to save your changes and use this version now.

All new classification processes run against the latest published model version.

Restore a model version

Follow these steps to select and restore an earlier version of your classification model:

1. Navigate to the Classification page group and select Models.
2. Click on the model you wish to restore a previous version for.
3. Go to the Versions tab.
4. Review the version history and select the older version you wish to restore.
5. Click on the options button next to the desired version.
6. Select Restore Version.

Add triggers

Classification is often one part of a larger business workflow. To automate downstream actions after classification, you can create triggers that initiate other process models when classification activities are complete. For example, after the model classifies documents as either Invoices or Purchase Orders, you can kick off the process that extracts the data from those documents.

To create a trigger:

1. Go to the classification model where you want to add a trigger.
2. Open the Triggers tab.
3. Click CREATE TRIGGER.
4. Enter the following configurations:
   1. When this activity completes: Select the activity you want to base the trigger on. Currently, Reconciliation is the only available option.
   2. Run this process model: Choose the process model to start when the activity is complete.
   3. Trigger this process to run: Synchronously or Asynchronously.
5. Click ADD TRIGGER.

Every time the classification model runs, your chosen process model will automatically start when the chosen activity is complete.

Scroll or drag to pan · ⌘+/⌘− to zoom · Esc to close

To edit a trigger:

1. Go to the classification model where you want to edit a trigger.
2. Open the Triggers tab.
3. Click Update Trigger .
4. Change the configurations as desired.
5. Click UPDATE TRIGGER.

To delete a trigger, open the Triggers tab and click Delete Trigger in the list.

Requirements for triggered processes

For classification triggers to work effectively, the process model must meet the following requirements:

• Process variable: The process model must have a single parameterized process variable named instance with AIA Classification Instance as its type.
• Permissions: The process model object must have View permissions for the AIA Operations group.
• Re-run handling: The process model must have flows that account for re-runs or updated runs.

You can only create one trigger per classification model. If you need to trigger more than one process model, create a wrapper process model containing all needed processes, and call that wrapper process model from the trigger.

Feedback