Document Automation v.39 release

Review what's new and changed, and the fixes and limitations in Document Automation for the v.39 release.

What's new

Introducing custom IA générative Connexions de modèle support in Document Automation

You can now connect any LLMs, including open-source, Sur site, or custom models to Document Automation using the Connexions de modèle feature in Studio d\'Agent IA. This enhancement extends the bring your own key (BYOK) capability, enabling organizations to configure and use custom or Sur site IA générative models through secure, user-defined connections instead of using the default connections, such as OpenAI, Anthropic, or Google Gemini, available in Document Automation.

Important: If you are updating to v.39 or later release and using BYOK for IA générative providers in your existing learning instances, then perform the following steps:
  1. Create Connexions de modèle for the IA générative provider and then use it in the learning instance.
  2. Open the Extract data action of the learning instance and set None in the Additional settings option.

Utilisation des Connexions de modèle dans Document Automation

Pre-installed GPT‑5.1 and Claude Haiku 4.5 models for data extraction
When you process documents using the OpenAI or Anthropic IA générative option, GPT‑5.1 and Claude Haiku 4.5 models will be used for data extraction.
Note: These models will be available only in specific regions. For more information, see Matrice de prise en charge des régions.

Matrice de prise en charge des régions

Introducing Google Gemini models for data extraction

You can use the Google Gemini IA générative models for data extraction in Document Automation.

Google Gemini provides the following advantages:

  • Efficient processing of large, unstructured documents
  • Handle documents in both English and other languages
  • Handle complex document layouts
  • Faster processing of documents with better data extraction accuracy
Note: Bring your own key (BYOK) is not supported for Google Gemini models.

Créer une instance d\'apprentissage dans Document Automation

What's changed

Enhancements to version history in test mode

The following enhancements are made to the version history in test mode:

  • When you make changes to a learning instance and the version selected in the version history is the latest, you will be allowed to save the changes made to the learning instance.
  • When you make changes to a learning instance and the version selected in the version history is not the latest, you will see an error when you try to save the changes made to the learning instance. Click Go to latest version, make changes to the learning instance again, and then save the changes.
  • The version history drop-down will display all the available versions.

Voir l\'historique des versions d\'une instance d\'apprentissage

Fixes

The Control Room can now be installed using Microsoft SQL Server database version 2016.

Previously, an error was displayed when writing Document Automation data to the database.
You will now be able to select a table header that spans across multiple lines in the Advanced training settings.

Previously, you were not able to select the table header in certain scenarios.

Service Cloud Case ID: 02287894, 02287918
The Requêtes de recherche adaptatives feature is now supported on Microsoft SQL Server database version 2016.

Previously, this feature was not supported on Microsoft SQL Server database version 2016.
Automation Co-Pilot manager (AAE_Robotic Interface Manager) or administrator (AAE_Robotic Interface Admin) will see the correct error message when validating documents in the Automation Co-Pilot validator.

Previously, an incorrect error message was displayed in such a scenario.
When you enable the test mode option for a learning instance in the public mode, the learning instance will no longer be moved to the private mode and will continue to be in the public mode.

Previously, the learning instance was moved to the private mode in certain scenarios.

Service Cloud Case ID: 02273055, 02291142
Data extracted from tables using Service d\'extraction cloud and Robot de tâches now generates similar outputs.

Previously, the outputs were different in certain scenarios.
Table extraction using the Service d\'extraction cloud now works for User-defined learning instances.

Previously, table extraction did not work in such a scenario.
Advanced training settings such as table headers and end-of-table indicators are now preserved after applying the end-of-table value.

Previously, applying a new end-of-table indicator reset all advanced training settings.
You can now delete custom tables when editing or creating learning instances.

Previously, users were unable to delete custom tables in such scenarios.
The Learning Instances page now displays options in the Actions column correctly when the Control Room language is set to German or French.

Previously, the options were misaligned in such a scenario.
You will no longer see an error when moving learning instances from IQ Bot to Document Automation using the IQ Bot - DA Bridge package.

Previously, an error was displayed in such a scenario in certain instances.

Service Cloud Case ID: 02284063
You can now import a learning instance without errors when the .dw file contains a different version than the existing learning instance in the target environment, provided the test mode option is disabled in the existing learning instance.

Previously, an error was displayed in such a scenario.

Service Cloud Case ID: 02282032, 02288219
For a learning instance, if you keep the default table prompt used in the learning instance and add a custom prompt for any table field in the Requêtes de recherche adaptatives, the table prompt in the Requêtes de recherche adaptatives no longer resets and the custom prompt is retained.

After updating to the v.39 release, all blank or empty table prompts in Requêtes de recherche adaptatives will be treated as if the user clicked the Use Default option, resulting in the prompt values reverting to default table prompts. In such a scenario, you need to manually clear the table prompts in Requêtes de recherche adaptatives after the update to set the prompt values to blank.

When you move a learning instance that was used to process documents in German from IQ Bot to Document Automation using the IQ Bot - DA Bridge package and process documents in Document Automation, data containing certain German characters is extracted correctly.

Previously, certain German characters were not extracted correctly.

Service Cloud Case ID: 02231391, 02273919
You can now use a custom table name in Japanese, Korean, or Chinese when creating a custom table.

Previously, you were not able to use a custom table name Japanese, Korean, or Chinese.
When you restore a learning instance to any previous version from version history and reprocess a document, the document upload count is not increased.

Previously, the document upload count increased in such a scenario.
You can now add fields to existing Formulaires standards projects, create new models, and train the models without any errors.

Previously, an error was displayed in such a scenario.
The Automation Co-Pilot validator now correctly displays a validation error when a required signature field is not found in Formulaires standards documents.

Previously, no error was displayed in such a scenario.
When you create a model in Formulaires standards, any leading or trailing spaces in table names will automatically be removed.

Previously, an error was displayed when training the model in such a scenario.
After performing a search for a learning instance, you can now create a learning instance in Community Edition if the limit allows.

Previously, users were not able to create a learning instance in such a scenario.

Limitations

The following limitations are applicable to Connexions de modèle:
  • Connexions de modèle using the Automation Anywhere vendor is not supported.

    Workaround: Use the built-in IA générative providers when creating learning instances.

  • In Connexions de modèle, the Fine-tuned and Grounded model types are not yet supported.

    Workaround: Use the Standard model type.

  • The following IA générative provider models in Amazon Bedrock are not yet supported:
    • Anthropic Claude
    • Nova Lite
    • Nova Pro
    • Nova Micro

    Workaround: Use the supported custom models.

  • Google Vertex AI models are not supported.

    Workaround: Use the supported custom models.

  • Only built-in models supporting OpenAPI schema support vision-powered IA générative data extraction. Custom models do not support vision-powered IA générative data extraction even when using OpenAPI schema.

    Workaround: Use the built-in models for vision-powered IA générative data extraction.

  • In test mode, when you revert to a previous version of the learning instance that was using a Connexion de modèle, the built-in OpenAI GPT IA générative provider is selected instead of the Connexion de modèle.
  • Connexion de modèle details are not visible in the test mode version history.
  • Importing or exporting of learning instances that are using Connexions de modèle is not supported.

    Workaround: Change to a built-in model and then import or export the learning instance. After importing the learning instance, change to the Connexion de modèle previously used.

  • Copying of learning instances that are using Connexions de modèle is not supported.

    Workaround: Change to a built-in model and then copy the learning instance. After copying the learning instance, change to the Connexion de modèle previously used.
If you are using v.39 (Build 42496) or an earlier release and using the Extraction de document package version 3.39.13, extraction will fail.

Workaround: Use Extraction de document package version 3.39.11 or an earlier version or update to v.39 release (Build 42649).

Document extraction will fail with the error “500: Internal Failure , Failed to initialize ABBYY’ when you change the default path (C:/ProgramData/AutomationAnywhere/GlobalCache) of Global Cache Location for your devices.
In the Advanced training settings option, you might see table fields listed in the Primary Column drop-down even when those fields are not selected for validation feedback.

Workaround: Avoid selecting any table field in the Primary Column drop-down if that field is not selected for validation feedback.
When you process documents containing multiple tables in Formulaires standards, the order of the tables extracted is reversed.
When you use a regular expression for a table field without defining an alias, the regular expression might not be considered for extracting the data.

Workaround: Define an alias for the table field before using a regular expression.
Limitations from previous releases
When you try to check in a process without having the required permissions, the audit log entry for this event will include Source as Control Room instead of Document Workspace and Item name as Unknown instead of <learning-instance-name>.
When you edit and check in a process linked to a public learning instance that is using Service d\'extraction cloud, the extraction robot might not be visible in dependencies, and document processing might fail.
Workaround: Use one of the following options:
  • Update to the v.39 release
  • Use the default process or copy the learning instance to create a new process and then publish the learning instance before processing documents
When you process documents with file names greater than or equal to 75 characters in test mode, you might not see improvements in the reprocessing time for such documents.
If you have disabled the OCR provider in the administrator settings and if you are using a language other than English for your Control Room, you will see an error to enable the OCR provider settings in English in the following scenarios:
  • When you create a learning instance that is using an OCR provider
  • When you change the OCR provider for an existing learning instance
When you use the Classificateur de documents actions (Classify, Classify documents, and Train Classifier) and Extract Data action in the Extraction de document package together in a robot, the robot will fail to execute.

Workaround: Ensure that you create separate robots when using any of the actions from the Classificateur de documents package and the Extract Data action of the Extraction de document package. If you need to execute these robots in a sequence, include these robots in an Automation Co-Pilot process.
If you copy a learning instance that is using a third-party parser configured in Document Automation and process documents using the copied learning instance, data extraction will fail.
When a user processes a document on a custom process learning instance, the validate document count is not updated post extraction. Also, if the user submits the document, the validate document count is updated with a negative value.
A user with the Automation Co-Pilot administrator permissions is unable to view the Document Automation tasks that are assigned or requested and are in pending or complete status.
When you are using the IQ Bot Pre-processor package actions and if the Output folder path contains Japanese characters, you will see an error when processing documents.

Workaround: Create an output folder in a folder path that does not contain any Japanese characters and provide the path in the output folder path field.

IQ Bot

The IQ Bot Sur site installer now displays the correct release version in the installation directory.

Previously, the release version displayed in the installation directory was incorrect.

Updates to the interface

Document Automation
The following screens are updated with the introduction of custom models to Document Automation using the Connexions de modèle:
  • Create learning instance option

    Create new learning instance updated screen

  • Create learning instance configuration

    Create learning instance updated configuration screen

  • Edit learning instance configuration

    Edit learning instance updated configuration screen

  • Additional settings option in Extraction bot

    Additional settings updated screen in Extraction bot

Introducing custom generative AI Model connections support in Document Automation

You can now use Google Gemini IA générative provider when creating learning instances:Create learning instance using the Google Gemini IA générative provider

Introducing Google Gemini models on AWS and GCP for data extraction in Document Automation