AI Agents

elDoc provides AI Agents that can understand document content and perform document-related tasks based on natural-language instructions.

Unlike conventional GenAI Chat, which primarily returns information to the user, AI Agents can combine document understanding with controlled actions in elDoc. Depending on the task, an Agent can analyze multiple files, compare information, rename or reorganize documents, create new office files, and populate existing documents with information extracted from other sources.

AI Agents operate within the user's elDoc permissions and can access only the files and folders available to the corresponding user.

Contents:

Available AI Agent Capabilities

The following AI Agent capabilities are currently available in elDoc. This list is not exhaustive and continues to expand as new Agent capabilities are introduced.

AI Agent to Understand Documents

The Document Understanding Agent can read one or multiple files and analyze their content.

Users can ask the Agent to understand documents without having to manually open and review each file.

Typical requests include:

• explain what a document is about;

• identify important information;

• extract relevant facts;

• summarize document content;

• analyze several files together;

• answer questions based on the selected documents.

This document-understanding capability is also used as a foundation for other Agents, including document renaming, organization, comparison, and data extraction.

AI Agent to Rename Files

The Rename Files Agent can analyze a group of files, understand their contents, and generate meaningful file names based on the information contained in each document.

For example, a folder containing files with names such as:

scan001.pdf
scan002.pdf
document3.pdf

can be analyzed and renamed according to their actual content, for example:

Invoice_2026-0142_ACME.pdf
BankStatement_July_2026.pdf
SupplierAgreement_ExampleLtd.pdf

The Agent can process multiple files as part of a single request.

This capability is useful for:

• scanned document archives;

• bulk document imports;

• files received from external systems;

• repositories containing inconsistent file names;

• automatically generated file names that do not describe document content.

AI Agent to Reorganize Documents

The Reorganize Documents Agent can analyze multiple files and organize them into folders according to their content.

The Agent determines the meaning or category of each document and places files into appropriate folders based on the user's instructions.

For example, a user can request:

Organize these files by document type.

The Agent can analyze the files and create or use a structure such as:

Documents
├── Contracts
├── Invoices
├── Bank Statements
├── Purchase Orders
└── Correspondence

Files are then assigned to the corresponding folders according to their content.

Organization criteria can also be based on information such as:

• document type;

• company;

• year;

• project;

• subject;

• category;

• other information identified within the documents.

This allows large groups of unorganized files to be converted into a structured repository using natural-language instructions.

AI Agent to Create and Edit Documents

elDoc AI Agents can create new office documents and work with supported existing documents.

The Agent can create:

• text documents, including Microsoft Word and OpenDocument Text formats;

• spreadsheets, including Microsoft Excel and OpenDocument Spreadsheet formats.

This allows users to request creation of documents using natural language.

For example:

Create a spreadsheet containing the product name, manufacturer, weight, calories, protein, carbohydrates, and fat for these products.

or:

Create a document summarizing the information contained in these files.

The Agent can use information from documents available in elDoc as input when creating the new file.

AI Agents can also update supported existing office documents when the requested operation and file format allow it.

AI Agent to Compare Documents

The Document Comparison Agent can analyze multiple documents and compare their content.

The Agent can identify:

• differences;

• similarities;

• changed values;

• missing information;

• inconsistent information;

• information present in one document but absent from another.

For example, users can ask the Agent to compare:

• different versions of an agreement;

• specifications received from different suppliers;

• financial documents;

• policies or procedures;

• reports;

• product specifications;

• structured information contained across several documents.

The comparison is based on document content rather than only file names or binary file differences.

This allows the Agent to perform semantic comparison even when documents have different layouts or formats.

AI Agent for Visual Document Understanding

elDoc AI Agents can use Vision-Language capabilities to understand information contained in images and visually structured documents.

This allows Agents to analyze content such as:

• product specifications;

• product labels;

• nutrition facts;

• forms;

• scanned documents;

• photographs containing structured information;

• other image-based business documents.

The Agent can identify information directly from the visual content and use it in subsequent operations.

AI Agent to Compare Visual Data with Structured Data

Visual understanding can be combined with spreadsheet and document processing.

For example, the Agent can:

1. analyze an image containing a product label or specification;

2. identify relevant values;

3. open a spreadsheet containing expected or reference values;

4. compare the extracted values with the spreadsheet;

5. identify differences or inconsistencies.

A typical request could be:

Compare the nutrition information on these product images with the values in the product specification spreadsheet and identify any differences.

The Agent can therefore work across different content types within the same task.

Product Image
      ↓
Visual Understanding
      ↓
Extracted Values
      ↓
Spreadsheet Data
      ↓
Comparison
      ↓
Differences / Results

This capability is useful for quality control, verification, reconciliation, and document review scenarios.

AI Agent to Populate Documents from Visual Data

Information extracted from images can also be written into new or existing office documents.

For example, an Agent can:

1. analyze product specification images;

2. extract selected fields;

3. create a new spreadsheet;

4. add one row for each product.

The resulting spreadsheet could contain:

The same approach can be used to populate an existing spreadsheet or text document.

This enables workflows such as:

Images / Documents
       ↓
AI Understanding
       ↓
Structured Information
       ↓
New or Existing
Word / ODT / Excel / ODS

As a result, AI Agents can perform not only information extraction but also end-to-end document transformation.

Working with Multiple Files

elDoc AI Agents are designed to work with multiple files as part of the same task.

For example, a user can select a group of documents and request:

Read these files and rename them according to document type, company, and date.

or:

Analyze these documents and organize them into appropriate folders.

or:

Compare all supplier specifications and create a spreadsheet containing the main differences.

The Agent processes the selected files as a working set and uses their content to determine the required actions.

This is particularly useful for large document repositories where performing the same task manually for every file would require significant effort.

Supported Content

elDoc AI Agents can understand different types of document content.

Supported AI-processing scenarios include:

Images

Various image formats can be processed using vision-capable models.

This includes scanned documents, photographs, screenshots, product labels, forms, specifications, and other image-based content.

PDF Documents

Both digitally generated and image-based PDF documents can participate in AI processing.

Image-based PDFs can be processed using OCR or Vision-Language capabilities where required.

Microsoft Office Documents

AI Agents can work with supported Microsoft Office files, including document and spreadsheet formats.

OpenDocument Files

Supported OpenDocument formats can also be processed, including text documents and spreadsheets.

Documents Created by AI Agents

elDoc AI Agents can create supported office files directly within File Management.

Currently supported output categories include:

• text documents;

• spreadsheets.

These can be created in formats such as:

• Microsoft Word;

• OpenDocument Text;

• Microsoft Excel;

• OpenDocument Spreadsheet.

The generated files become normal elDoc files and can subsequently be:

• opened;

• edited;

• versioned;

• shared;

• indexed;

• included in workflows;

• processed by other AI capabilities.

Combining Agent Capabilities

AI Agent capabilities can be combined as part of a larger task.

For example:

Read Files
    ↓
Understand Content
    ↓
Classify Information
    ↓
Rename Files
    ↓
Organize into Folders

Another task can involve:

Analyze Product Images
    ↓
Extract Specification Data
    ↓
Read Reference Spreadsheet
    ↓
Compare Values
    ↓
Create Results Spreadsheet

Or:

Read Multiple Documents
    ↓
Compare Content
    ↓
Identify Differences
    ↓
Create Summary Document

This ability to combine document understanding with actions is one of the main differences between AI Agents and conventional question-and-answer Chat functionality.

AI Agents and GenAI Chat

AI Agents are invoked directly from the AI Chat dialog in the AI File Management module.

Users do not need to select a specific Agent manually or construct predefined commands. Instead, they describe the required task using standard natural language.

For example:

Rename these files based on their content.

Sort these documents into folders by document type and year.

Read these agreements and summarize the main differences.

Extract the product information from these images and add it to this spreadsheet.

Compare these product labels with the specification table and identify inconsistencies.

elDoc interprets the request, determines the required AI Agent capabilities and tools, analyzes the relevant documents, and performs the requested task within the user's available permissions.

This allows users to work with AI Agents through the same conversational interface used for GenAI Chat, without needing to understand the underlying Agent implementation or tool orchestration.

Security and Permissions

AI Agents operate within the existing elDoc security model.

An Agent can access and perform operations only on files and folders for which the current user has the corresponding permissions.

For example:

• reading document content requires access to the document;

• renaming requires the corresponding File Management permission;

• creating documents requires permission to create content in the destination folder;

• editing an existing file requires editing permission;

• moving or reorganizing files requires the applicable File Management permissions.

AI Agents therefore do not bypass File Management access control.

The same security rules apply whether an operation is performed manually by the user or through an AI Agent acting on the user's request.

Current AI Agent Capabilities at a Glance

The currently available capabilities can be summarized as:

Extensible Agent Architecture

elDoc AI Agents are based on an extensible agent and tool architecture.

New Agent capabilities can therefore be introduced as additional document-processing, retrieval, automation, and integration functions become available within the platform.

The overall objective is to allow AI Agents to work with enterprise documents end-to-end - from understanding and extraction to comparison, transformation, organization, and action.